Experimental Characterization of

Document Sample
Experimental Characterization of Powered By Docstoc
					      Experimental Characterization of Home Wireless
            Networks and Design Implications
                                   Konstantina Papagiannaki‡ , Mark Yarvis† , W. Steven Conner†
                                        ‡: Intel Research Cambridge, †: Intel Corporation

   Abstract— Anecdotal evidence suggests that home wireless              (AP), several PCs, and increasingly, consumer electronic (CE)
networks may be unpredictable despite their limited size. In this        devices. Today a home wireless network is primarily used to
work, we deploy six-node wireless testbeds in three houses in the        provide access to a wired Internet connection; communication
United States and the United Kingdom. We examine the quality of
links in home wireless networks and the effect of (i) transmission       occurs to and from an AP. Future home wireless networks are
rate, (ii) transmission power, (iii) node location, (iv) type of         expected to feature different types of traffic (such as multi-
house, and (v) 802.11 technology. We provide empirical evidence          media streaming) which may require efficient communication
suggesting that homes are challenging environments for wireless          between any two devices in the home. Such an environment
communication. Wireless links in the home are highly asymmetric          can be enabled using 802.11 ad hoc mode, 802.11 Wireless
and heavily influenced by precise node location, transmission
power, and encoding rate, rather than physical distance between          Distribution System (WDS) frames, or the proposed 802.11e
nodes. In our measurements, many links were unable to utilize the        Direct Link Setup (DLS) capability [3]. An evaluation of
maximum transmission rate of the deployed 802.11 technology,             typical link characteristics in the home environment is crucial
and a few provided no connectivity at all. These results suggest         to understanding the behavior of both traditional AP-based as
that creating an AP-based topology with maximum coverage and             well as future network topologies.
throughput in this environment is challenging. Our findings have
implications on the design of future home wireless networks and             Using small networks of devices deployed in three homes,
requirements for future wifi-enabled consumer electronic devices.         two in the United States and one in the United Kingdom, we
We show that coverage and performance can be improved using a            study the properties of wireless links in home environments
multi-hop topology, implying that mesh capabilities may actually         using a pure measurement approach. We examine the impact
be needed in consumer electronics for seamless connectivity              of transmission rate and transmission power on link quality, in
across the home.
                                                                         terms of success rate and throughput. We show that despite the
                                                                         small size of home wireless networks, connectivity between
                          I. I NTRODUCTION                               any two wireless devices is not guaranteed or necessarily
    Wireless networks have become increasingly popular due               predictable, regardless of transmission power or rate. We also
to ease of deployment and low cost compared to wired                     show that small changes in antenna orientation and node
networks. However, the transmission principles in wireless               location can have a dramatic and unpredictable impact on the
communications are dramatically different than those of wired            connectivity of the network. Our results span both 802.11a and
networks. A recent study of wireless access points deployed              802.11b technologies and do not show strong correlation with
over a metropolitan area demonstrates significant challenges to           the physical distance between nodes. These results suggest that
performance and connectivity [1]. Similarly, deployment of a             a typical home user cannot depend on common sense alone in
wireless network in an enterprise environment, while relatively          deploying a high-performance wireless network.
well understood, typically requires a site survey to engineer               The above results have a direct impact on home wireless net-
a network with proper coverage and capacity. Comparatively               work design. Infrastructure mode wireless networks typically
little is known about the properties of home wireless networks,          deployed in homes require one or more access points, through
beyond anecdotal evidence [2]. Even less is known about                  which all other nodes communicate. Our characterization of
optimal design of such networks. Unfortunately, previous                 connectivity in the home suggests that most of the possible
studies of campus/metropolitan-area networks and enterprise              access point deployment locations fail to provide full coverage
networks are hardly applicable to home networks, which tend              of high-performance connectivity throughout a typical home.
to be much smaller (both in total size and contiguous space),            To find the optimal location for access point deployment, a
have a single access point, are almost entirely indoors, and             homeowner could resort to a site survey. Design of medium
have no IT staff to perform site surveys. In this paper we               or large-scale wireless networks typically relies on a site
consider factors that impact the design and performance of               survey or tools that model radio signal propagation, taking into
home wireless networks.                                                  account a detailed floor plan along with construction materials
    Before investigating network design, we first attempt to              and placement of household objects [4], [5]. These solutions
measure the characteristics of home wireless networks. A                 are far too costly and time-consuming to be applied in the
typical home wireless network consists of an access point                home. In addition, aesthetics and the location at which the
                                                                         Internet enters the home are usually overriding concerns for
  ∗   Other names and brands may be claimed as the property of others.   the home user when selecting the location for an access point.
                                                                                                                TABLE I
A home user is much more likely to put an access point in
                                                                                     D ESCRIPTION OF HOMES USED IN EXPERIMENTAL TESTBEDS .
the corner of an office than the middle of a living room, even
if the later location is known to provide optimal performance.                        Label      Size (f t2 )   Construction    # Floors   # Nodes
A second option would be to deploy multiple access points.                           ushome1       2,500           Wood            2          6
However, this approach typically requires the wired network                          ushome2       2,600           Wood            3          6
                                                                                     ukhome1       1,500        Brick / steel      3          6
to be extended to multiple points within the home. Thus, this
option would eliminate the key advantage of home wireless
networking: low installation cost. Our results suggest that more
flexible topologies can provide a more appropriate alternative                    •   Transmission power, denoted by txpower.
in home environments.                                                            •   Transmission rate, denoted by txrate.
   The measurements collected from our 6-node testbeds show                      •   Node location.
that across all three houses there is always at least one
pair of nodes that cannot directly communicate. Moreover, a
                                                                               A. Experimental Setup
substantial number of pairs cannot communicate at the highest
rate supported by the deployed 802.11 technology. Using these                     We deploy six wireless nodes inside each home3 . Nodes are
measurements we study the ability of alternative topologies                    located in different rooms, wherever computing or consumer
to alleviate these problems. More precisely, we study the                      electronic devices might be found in the home. For 802.11b
impact of (i) AP, (ii) direct, and (iii) mesh topologies. We                   experiments the nodes are small form-factor PCs with Netgear
show that the AP topology typical in home networks today                       MA701 compact flash 802.11b wireless cards. The nodes
is highly sensitive to the location of the AP and in many                      run Linux kernel version 2.4.19 and the hostap driver [8].
cases leads to sub-optimal performance, due to the mandatory                   For 802.11a experiments the nodes are laptops with NetGear
use of the AP as a relay. A direct topology eliminates the                     WAG511 CardBus 802.11a/b/g cards running Linux kernel
need to transmit each packet twice and also allows nearby                      version 2.4.26 and the MIT madwifi-stripped driver [9]. All
nodes to utilize a high-rate channel encoding, which would not                 radios have omnidirectional antennas, and could be considered
otherwise be possible if the access point is far away. However,                comparable to the radios that are likely to be integrated
in a direct topology not all nodes will be able to communicate                 in future consumer electronics, e.g. cheap radios with basic
with each other. On the other hand, a mesh topology can                        functionality. Lastly, all three testbeds are homogeneous; each
help nodes overcome the limitations of their environment,                      node consists of the exact same hardware to limit the impact
establishing either direct connections or multi-hop connections                of hardware peculiarities on the obtained results.
through neighboring nodes as appropriate. Our results show                        Our network setup is common among all experiments.
that solutions enabling mesh topologies [6] can help eliminate                 All nodes utilize an unused frequency that is at least five
poor quality links in the home while establishing paths for the                channels away from the next occupied 802.11 frequency. To
communication of nodes that are otherwise unreachable. To the                  facilitate our experiments, we utilize the 802.11 Independent
best of our knowledge, our work is the first one to identify                    Basic Service Set (IBSS) mode, which allows all nodes to
and quantify such challenges in home wireless communication                    communicate directly. However, this configuration does not
and to demonstrate the need for mesh capabilities (e.g., IEEE                  constrain the usefulness of our results to ad hoc (mesh)
802.11s [7]) on wifi-enabled consumer electronics.                              topologies, as will be described in Section V. Each node is
   The remainder of the paper is structured as follows. In                     instructed to run an experiment toward every other node in
Section II we present the experimental methodology followed                    turn. Our experiments are designed to assess: (i) success/loss
throughout the paper. In Section III we assess the quality                     rate, and (ii) throughput under different combinations of txrate
of wireless links inside the house in terms of reachability.                   and txpower. We further alter the node location for specific
Throughput measurements are presented in Section IV. We                        experiments in order to quantify the impact of exact node
study the impact of alternative topologies on the performance                  location, antenna orientation, and obstacles. Experiments are
of a home wireless network in Section V. We present related                    carried out during the night to avoid interference from moving
work in Section VI and conclude in Section VII.                                people and facilitate reproducibility. Except where explicitly
                                                                               stated, each result represents a single experimental run, due to
              II. E XPERIMENTAL E NVIRONMENT                                   the highly time consuming nature of our experiments. Instead,
   Our experiments are intended to assess the quality of wire-                 we rely on the validating runs presented in the following
less links in home environments. We evaluate three homes, two                  subsections to lend credence to our results.
in the United States and one in the United Kingdom. High-                         1) Reachability: The reachability experiments assess link
level details of the different homes hosting our experiments                   quality between each pair of nodes in the home network in
can be found in Table I1 . Our experiments are designed to                     terms of success/loss rate, and rely on a series of UDP probe
investigate the impact of the following parameters:                            packets sent from every node to every other node. Each probe
   • Type of house, e.g. size, construction material.                          packet lists the source node, as well as its number in the
                                                    2                          series. The size of the probe packet and the duration of each
   • Wireless technology used: 802.11a or 802.11b .

  1 Inushome2, the bottom floor was partially below grade.                        3 The nodes are denoted as node-2 through node-7 in the remainder of the
  2 Thewireless cards used also support 802.11g, which has not been reported   paper. Node-1 controls each experiment, instructing each node in turn about
upon in this paper due to the poor 802.11g support in the driver.              the type and parameters of the experiment.
sub-experiment are configurable. In all experiments, link-layer
retransmissions were disabled, the probe size was 1472 bytes,
and the duration of each sub-experiment was 60 seconds, with
a frequency of one packet every 500 ms. Each individual
wireless link is assessed independently, and no simultaneous
transmissions take place inside the network.
   2) Throughput: The throughput experiments were run to
assess the quality of layer-3 communication within the home
network. Throughput was assessed for both TCP and UDP
protocols. For our throughput experiments we use the same
basic methodology which relies on measurements from all
pairings of the six nodes in the testbed. Throughput is mea-
sured by the netperf traffic generator using 1472 byte packets.
Each node initiates a netperf connection to every other node
(in turn) and measures the throughput achieved over a 60
                                                                     Fig. 1. Matrix of probe packets successfully delivered between each pair
second time interval. Unlike the reachability experiments,           of nodes in ushome2 at 30mW and 2Mbps (ushome2 − layout1 − link −
the throughput experiments are conducted with link layer             validation1).
retransmission enabled (maximum of 3 retries), which is likely
to alleviate the effect of short term degradation in link quality.
                                                                     the loss and throughput rates observed for each wireless link
B. Basic Methodology                                                 under each test. We consider this an essential step of our
                                                                     measurement methodology, to identify whether a change in the
   Each reachability experiment quantifies the loss rate ob-          collected measurements, across a specific experiment, reflects
served by each wireless link, as well as the sequence of             the effect of the studied parameter (e.g. txrate or txpower) and
successes and failures. We graphically present the obtained          not the inherent variability in the measurements themselves.
measurement matrix in Figure 1 as collected in ushome2 when
                                                                        1) Reachability: In Figure 2 we present the results obtained
txrate is 2 Mbps and txpower is 30 mW. Every row in Figure
                                                                     across two experiments with the same setup. Each experiment
1 corresponds to probes sent from a specific source node. In
                                                                     results in four subplots, where txrate is either 2 Mbps or 11
each subplot, a bar denotes the successful reception of a probe
                                                                     Mbps and txpower is either 30 mW or 1 mW. Each subplot
by the destination node.
                                                                     contains the performance of individual wireless links in terms
   From Figure 1 we note that in ushome2 and under the
                                                                     of their loss rate in each direction4 (Figure 1 contains the
selected transmission power and rate, communication from
                                                                     source data presented in the lower left plot of Figure 2(a).).
node-2 to node-7 is extremely limited; most probe packets
                                                                     While not exactly identical, the performance shown in each
were lost. In addition, we see that while node-2 has a low
                                                                     subgraph of Figure 2(a) is similar to that of Figure 2(b). Links
delivery success rate to node-6, the quality of the link in the
                                                                     that are poor or asymmetric in one run, tend to also be poor
reverse direction is significantly better (nearly 100% success).
                                                                     or asymmetric in the next. Thus, network performance does
Such link asymmetry has been reported in previous perfor-
                                                                     not change significantly from one run to the next. We ran the
mance studies of wireless networks [1], [10] and was found
                                                                     same validation test in each home and found this result was
to be quite common in the home environments studied in this
                                                                     easily reproducible.
paper. Throughput experiments simply output the rate achieved
                                                                        To determine whether 60 seconds is sufficient to obtain
for each pair of nodes under the different txrate, txpower
                                                                     an accurate view of the quality of the wireless link, we
combinations, along with the respective packet loss rate.
                                                                     also measured links over a longer period of time. Using a
   Given the large amount of experimental data collected for
                                                                     transmission rate of 11 Mbps and a transmit power of 30 mW,
this study and the number of different dimensions, we tag the
                                                                     we performed the same experiment in ushome1 for a time
experiments with the following 4-tuple: home-layout-type-
                                                                     span of 20 minutes (instead of 1 minute). We then compared
experiment. The first tuple can be ushome1, ushome2, or
                                                                     the success rates derived using the entire time series with
ukhome1. The second tuple reflects the positions of the differ-
                                                                     the success rates that would be estimated by the first 120
ent nodes and is of the form layout1,2,.., etc. The third tuple
                                                                     samples (i.e. 60 seconds). In Figure 3, each point represents
denotes whether the experiment assesses link loss/success rate,
                                                                     the two success rate measurements for each unidirectional link.
UDP, or TCP throughput. The last tuple denotes the goal of
                                                                     In each case, the success rate measured in 60 seconds was
the experiment, for instance in the next section we present
                                                                     a reasonable estimate of the success rate over the 20 minute
results from our validation experiments.
                                                                     period. Thus, 60 seconds is long enough to assess the medium-
                                                                     term properties of each link under the tested conditions.
C. Validation
  To validate whether our experimental results represent ac-            4 An easy way to read the performance summaries in Figure 2 is to observe

tual link characteristics or a transient effect, we run two          how busy they are; a busy plot implies poor performance, a mostly empty
                                                                     plot indicates no loss across the majority of node pairs. Similar summaries
experiments with the exact same node deployment (denoted             are used to capture throughput results, in which case absent bars indicate zero
by layout1) and at the same time of day. We then compute             throughput along a particular link direction.
                                                                                 Fig. 3. Comparison of success rate results for 120 and 2400 sample lengths.
                                                                                 The straight line is used as reference of equality (y=x) (ushome1−layout1−
                                                                                 link − long, txrate=11Mbps, txpower=30mW).


                                                                                 Fig. 4. Loss rate as a function of time of day for ushome1 (txpower=30mW,
                                                                                 txrate=11Mbps). First bar is node-4 to node-6, second bar is node-6 to node-4.
                                                                                 (ushome1 − layout1 − link − day, txrate=11Mbps, txpower=30mW)

                                                                                 experiment. Indeed, Figure 6 confirms such a statement.
                                                                                    In Figure 6 we see that good quality links typically achieve
                                                                                 the highest throughput supported, while bad links typically ob-
                                                                                 tain low throughput. We note that the set of reachability mea-
                                     (b)                                         surements is taken before the set of throughput measurements,
                                                                                 with a one hour time lag between the two measurements of
Fig. 2. Loss rates for each pair of nodes in experiments (a) ushome2 −           a given link. Links with occasional losses may have variable
layout1 − link − validation1, and (b) ushome2 − layout1 − link −
validation2. Average loss from top-left to bottom-right: (a) 0.42, 0.37, 0.35,   performance. In addition, throughput experiments were carried
0.35, (b) 0.33, 0.38, 0.29, 0.29.                                                out with link level transmission turned on (maximum of 3
                                                                                 retries), while reachability experiments were performed with
                                                                                 no link layer retransmission. Throughput provides a more
   We must also consider the effect of time of day on link per-                  accurate picture of these losses, resulting in intermediate
formance. Recall that experiments were typically performed                       throughput values. From Figure 6, however, it is clear that high
at night to avoid interference from household activity. To                       loss rate on any direction leads to very poor UDP throughput
determine if results obtained at other times of day would vary                   along that direction. In case of TCP throughput, if high loss
significantly, we performed a 60-second link test for a single                    is observed on the forward direction then TCP throughput is
node pair in ushome1 once per hour for 24 hours. As shown                        very low (link 3-6). If high loss rate is observed on the reverse
in Figure 4, while link quality may fluctuate somewhat with                       direction, then TCP throughput is not zero but significantly
time, a ”good” link tends to remain ”good” (and a ”bad” link                     smaller (links 2-3, 5-6, 5-6) than what could be achieved
remains ”bad”), despite small deviations over time. To avoid                     on a loss-free link on both directions. We note that such
any complications from time of day specific behavior, we tried                    great sensitivity is not evident in our measurements when
to collect comparable data at the same time of day.                              the transmission rate is 2 Mbps due to the robustness of the
   2) Throughput: The validation process for the throughput                      encoding scheme (also seen in Figure 5).
experiments involves the following steps. First, we compare                         3) Experimental Setup: One last aspect of the experimental
the UDP results obtained across two different runs on two                        setup that needs to be validated has to do with the hardware
sequential days with the same node configuration. We found                        itself. Wireless measurements are likely to be influenced by the
that results were easily reproducible (Figure 5). Second, we                     hardware and software used. The results presented henceforth
compare whether the throughput rates achieved agree with                         are specific to the platform we use. However,we must validate
the success rates reported from the preceding reachability                       whether specific observed characteristics are an artifact of the
                                                                                  Fig. 6.   Relationship between loss rate and UDP/TCP throughput
                                     (a)                                          (ushome2 − layout − ∗ − validation1, txrate=11Mbps, txpower=1mW).

                                                                                  A. Overall characteristics of a home network
                                                                                     Using 802.11b radios, a full set of measurements like those
                                                                                  presented in Figure 1 was collected for four combinations of
                                                                                  transmission power and rate. In Figure 7(a) we present our
                                                                                  results for all combinations from ushome2. The deployment
                                                                                  of nodes in the individual homes is schematically shown in
                                                                                  Figure 8. We refer to the initial layout of the nodes in each
                                                                                  home as layout1.
                                                                                     As expected, in most cases link loss rates are higher when
                                                                                  the encoding rate is higher and somewhat lower when the
                                                                                  power level increases. While each home represents a small
                                                                                  space, wireless connectivity is not always omnipresent. Across
                                     (b)                                          all rates and power levels, a large number of asymmetric
                                                                                  links are present. In most experiments, at least one pair of
Fig. 5. UDP throughput achieved across validation runs, (a) ushome2 −             nodes has greater than 30% loss. And, as seen in Figure
layout1 − udp − validation1, (b) ushome2 − layout1 − udp −
validation2. Average throughput in Mbps from top-left to bottom-right: (a)        7(a), while the increase in transmission power improves some
1.01, 2.92, 1.01, 2.89, (b) 1.06, 2.92, 1.02, 3.02.                               links, the overall problem is not eliminated. This initial set
                                                                                  of experiments demonstrates that lossy links are likely to
                                                                                  be found inside a home, and in some cases, loss cannot be
behavior of specific nodes. To discount such a possibility, all                    eliminated by reducing the transmission rate or increasing the
three testbeds comprise the exact same hardware and run the                       transmission power. On the other hand, such changes do not
exact same versions of the respective software. Before we                         appear to affect the quality of links with low loss rates.
started the described experiments we had to replace specific
nodes that appeared to mis-behave, sourcing all asymmetric                        B. Small changes in antenna orientation and location
links for instance. After each individual testbed was assessed
to function properly, we rotated nodes and radios around the                         There are several reasons why particular node pairs may
house (node-2 moved to the position of node-3, node-3 to the                      not be able to communicate. The location of the nodes and
position of node-4, etc.), and observed whether there were                        the orientation of their antennas impact the obstacles in their
specific trends in the results. Our findings validated that all                     direct path, and thus multi-path fading and signal attenuation.
nodes are equivalent in terms of quality.                                         To evaluate these effects in the home, we modify layout1
                                                                                  by rotating all nodes 180o , such that their antennas face
                                                                                  the opposite direction (such a change may have also slightly
                   III. R EACHABILITY R ESULTS                                    changed their exact location). We call this deployment layout2.
   We now evaluate the home wireless environment along five                        We perform the same series of experiments on layout2 and
dimensions: (i) txrate, (ii) txpower, (iii) node location, (iv)                   present the results in Figure 7(b).
house type, (v) physical layer5 . Due to space limitations, we                       We observe that a small change in node location and
present results for the most interesting experiments across                       orientation can have a significant impact on link quality. The
homes and settings. All experimental results from all houses                      number of links with a loss rate above 50% in layout2 is
are available at [12].                                                            smaller than the one in layout1. Since the distance between
   5 We have also looked at external interference (e.g., from a microwave oven)   nodes does not change significantly, and since the change
in a preliminary version of this work [11].                                       observed between layout1 and layout2 is much greater than
                        (a)                                                (b)                                              (c)
Fig. 7. Loss rate for node pairs for (a) layout1, (b) layout2, and (c) layout3 in ushome2 (ushome2 − ∗ − link − placement). Average loss from top-left
to bottom-right: (a) 0.42, 0.37, 0.35, 0.35, (b) 0.19, 0.29, 0.31, 0.29, (c) 0.31, 0.38, 0.28, 0.28.

normal variation (Section II-C), exact node placement must                     Our results thus far demonstrate that the performance of
be a key contributor to performance.                                        a home wireless link tends to be most affected by the set
  Our findings for ushome2 are summarized in Figure 8(b).                    of objects between the two endpoints, rather than physical
The leftmost figure denotes the node pairs that experience                   distance or transmission power.
the worst connectivity (links with greater than 95% loss)
in layout1. In the middle we identify links with the worst                  E. Comparison between home networks
connectivity in layout2. Under the new configuration, the set of                Across homes, results differ substantially. Figure 10 presents
nodes that cannot communicate has changed. Similar findings                  loss rates for layout1 across all three homes. In layout1, the
were obtained for ushome1 and ukhome1 as shown in Figure                    largest home, ushome2, has the worst performance (Figure
8(a) and (c), with ushome1 experiencing the most dramatic                   10(b)), and the smallest home, ukhome1 has the best per-
changes in performance between layout1 and layout2.                         formance, particularly at low transmit power (Figure 10(c)).
C. Large changes in node placement                                          Results for ushome1 (Figure 10(a)) are significantly different
                                                                            from the results for ushome2 even though they are similar in
   The previous section considered the impact of small changes              size. Ushome1 and ushome2 differ in that ushome2 is a three-
in node location. We now move each node in ushome2 from                     story building, while ushome1 has two floors. Nonetheless, the
its positions in layout1 to a different location within the same            impact of the number of floors is not evident for ukhome1,
room (rightmost plot of Figure 8(b)). We call this configuration             which allowed an almost fully connected wireless network.
layout3. Loss rates measured for layout3 are shown in Figure                While some of the observations above could suggest that
7(c). We observe that layout3 leads to a more “concentrated”                distance or size play a significant role in performance, the
loss area than layout1.                                                     overall results presented here and in Section III-D demonstrate
   The above results clearly demonstrate the challenges of                  that loss rate cannot be predicted based on such features. The
home environments on the design and performance of home                     key parameter is precise node location and orientation, rather
wireless networks. Node positioning has a dramatic impact on                than home size or distance between nodes.
network connectivity, and ”randomly” selecting the location of
a node will not ensure its connectivity. Moreover, ”randomly”
selecting the location for an access point does not necessarily             F. The impact of the physical layer: 802.11a
ensure a fully connected network. Consider use of an AP                        While the preceding data was collected using the IEEE
topology in our testbed, with one testbed node replaced by                  802.11b physical layer, other physical layers may possess
an access point. For the example of layout3, node locations                 different characteristics. In this section, we consider the per-
2, 5, 6 and 7 would not be good choices for an AP, as they                  formance of the IEEE 802.11a physical layer in the home.
would not have good connectivity to all other nodes.                        As described in Section II, we deploy laptops with 802.11a
D. The relationship between link quality and distance                       wireless cards in the same locations as the 802.11b nodes
                                                                            and perform the same series of connectivity experiments. Each
   In Section III-A we demonstrated that home wireless links                experiment is completed with the same transmission power: 30
tend to be highly asymmetric. The presence of asymmetry                     mW. We considered four different link encoding rates: 6 Mbps,
suggests a loose relationship between distance and link quality.            18 Mbps, 36 Mbps, and 54 Mbps. Two node deployments
In this section we look into this question in more detail.                  were used, where layout1 is the initial deployment, and in
   Figure 9 presents the loss rate between node pairs for                   layout2 nodes are rotated by 180o . The loss rates in ushome1
layout2 in each home versus the distance between the nodes.                 for layout1 and layout2 are reported in Figure 116 .
Clearly there is no correlation between physical distance and
wireless link quality in these home networks. This result holds               6 802.11a experiments were not possible in the UK home due to local
across homes and across txrate and txpower settings.                        regulations and limitations of the card’s IBSS implementation.
                            (a)                                                   (b)                                          (c)
Fig. 8. Abstract home floorplans and location of links with greater than 95% loss rate at 1 mW and 11 Mbps under different configurations: (a) ushome1
for layout1, layout2, (b) ushome2 for layout1, layout2, and layout3, and (c) ukhome1 for layout1, layout2 (∗ − ∗ − link − homes).

                            (a)                                           (b)                                            (c)
Fig. 9. Loss rate for each pair of nodes against their distance for (a) ushome1, (b) ushome2, and (c) ukhome1 under layout2 (∗ − layout2 − link − homes).

                         (a)                                              (b)                                               (c)
Fig. 10. Loss rate for each pair of nodes for layout1 in (a) ushome1, (b) ushome2, and (c) ukhome1 (∗ − layout1 − link − homes). Average loss from
top-left to bottom-right: (a) 0.23, 0.26, 0.31, 0.31, (b) 0.42, 0.37, 0.35, 0.35, (c) 0.02, 0.09, 0.06, 0.07.

   As might be expected, the characteristics of 802.11a wire-                   and 802.11b. In ushome1 the 6 Mbps 802.11a links were
less links in the home are not entirely unlike 802.11b wireless                 much more reliable than either the 2 Mbps or 11 Mbps
links. Packet loss rates generally increase as the link encoding                802.11b links. Thus, one would expect 802.11a to provide
rate increases. Many links are lossy, and some links are                        better throughput in the home. However, the 54 Mbps link
highly asymmetric. In some cases it is possible to create a                     encoding performed very poorly between more than 40% of
nearly loss-free network at low data rates, but only at specific                 all node pairs. Thus, unless nodes are very optimally placed
node locations and orientations. As with the 802.11b results,                   in the home, it is unlikely that 54 Mbps will be attained (a
network quality appears to be sensitive to small changes in                     finding which is also confirmed in the next section). While
node position and orientation, as seen in layout1 and layout2.                  one might expect the 802.11a MAC to perform better in equal
Finally, we have previously confirmed that 802.11a link loss                     environments by design, lower levels of interference from non-
rates do not correlate with the distance between node pairs                     802.11 devices in the 5 GHz band may also contribute to the
[11].                                                                           superior performance of 802.11a in the home environment.
   While the 802.11a results are similar to the 802.11b results,
one difference is quite clear. In the home, 802.11a links                                         IV. T HROUGHPUT RESULTS
appear to have a rather ”binary” behavior, despite the lack                        In the previous section we looked into the characteristics of
of link-layer retransmissions. Link loss rates in the 802.11b                   home wireless networks in terms of reachability. The results
experiments take on a much wider variety of values.                             demonstrate that despite the small size of home networks,
   Figure 12 provides a summary comparison between 802.11a                      there may always exist nodes that cannot communicate with
                                                                              Fig. 12. Cumulative density function of loss rates under IEEE 802.11b and
                                                                              IEEE 802.11a in ushome1 (layout1) - the 11a 6Mbps is not visible because
                                                                              there is no link with loss under this configuration (ushome1 − layout1 −
                                                                              link − 11a/11b).

                                                                              An increase in transmission power, however, does not appear
                                                                              to improve quality in ushome1. Looking at the equivalent
                                                                              TCP throughput rates (Figure 13(b)) we notice a significant
                                                                              degradation in performance. This effect is expected, given
                                                                              that TCP is designed to react aggressively to dropped packets,
                                                                              lowering the overall throughput observed between two nodes
                                                                              with occasional packet losses.

                                                                              B. Fixed transmission rate using IEEE 802.11a
                                                                                 In Figure 12 we conjectured that home users may not be able
                                                                              to make full use of the maximum transmission rate offered by
                                                                              IEEE 802.11a due to increased loss rates. In this subsection,
                                                                              we revisit this issue by measuring the throughput achieved by
                                                                              all node pairs in ushome1 under the 6, 18, 36, and 54 Mbps
                                                                              transmission rates.
                                   (b)                                           UDP throughput across transmission rates is shown in
Fig. 11. Loss rate for each pair of nodes for ushome1 under IEEE 802.11a,     Figure 14(a)). For each node pair, as the link rate increases
with two different node orientations, (a) layout1 and (b) layout2 (ushome1−   throughput increases, until the loss rate begins to dominate
layout1/layout2 − link − 11a). Average loss from top-left to bottom-right:
(a) 0, 0, 0.1, 0.42, (b) 0, 0.06, 0.06, 0.44.                                 and throughput decreases. In ushome1, as the transmission
                                                                              rate increases beyond 18 Mbps, many links drop to zero
                                                                              throughput. Only half of the links inside ushome1 can make
all other nodes in the network. Moreover, such communication                  use of the 54 Mbps transmission rate.
problems cannot always be rectified by increasing the trans-                      The effect of packet loss is even more dramatic on TCP
mission power or decreasing the transmission rate (and hence                  (Figure 14(b)). As expected, in most cases the effect of packet
employing more robust encoding).                                              loss (and perhaps increased overhead) results in lower through-
   Throughput is an additional dimension to the connectivity                  put than UDP. Increasing the data rate to 54 Mbps results in
problem. Despite reachability, a link between two nodes may                   extremely poor performance: no TCP connection exceeds a
experience non-negligible loss rates, and thus suffer from                    throughput of 10 Mbps, and half are actually zero. We note,
intermittent connectivity and low throughput rates; a case that               however, that the performance observed for txrate=18Mbps
will be more evident for TCP traffic that is designed to react                 and txpower=30mW is superior to the performance of 802.11b
to observed losses. In this section we look at the achieved                   when txrate=11Mbps at the same txpower level.
UDP and TCP throughput on our testbed networks with and
without automatic link rate adaptation.                                       C. Autorate
                                                                                 In our previous experiments, the link-layer transmission
A. Fixed transmission rate using IEEE 802.11b                                 rate was fixed. In reality, wireless cards can adjust their
  Figure 13(a) presents the UDP throughput rates achieved                     encoding rate according to the quality of the wireless channel,
by different node pairs for different combinations of txrate                  i.e. “autorate”. The autorate functionality is implemented on
and txpower in ushome2. We note that each node pair with a                    wireless cards such that if a high transmission rate cannot
poor quality wireless link (Figure 10(b)) has zero throughput.                be effectively supported, the card can fall back to a lower,
                                 (a)                                                                         (a)

                                 (b)                                                                         (b)
                                                                          Fig. 14.     (a) UDP and (b) TCP throughput rates for ushome1 under
Fig. 13.     (a) UDP and (b) TCP throughput rates for ushome1 under
                                                                          different combinations of txrate and txpower (802.11a, ushome1−layout1−
different combinations of txrate and txpower (802.11b, ushome1−layout1−
                                                                          udp/tcp − 11a). Average throughput from top-left to bottom-right: (a) 4.43,
udp/tcp − validation1). Average throughput in Mbps from top-left to
                                                                          8.64, 9.98, 9.23, (b) 3.93, 8.74, 5.85, 4.75.
bottom-right: (a) 1.12, 3.23, 1.1, 3.14, (b) 0.93, 2.41, 0.94, 2.34.

                                                                          Consequently, the effective throughput achieved is under 1
more robust (in terms of encoding) transmission rate. In this             Mbps.
section we explore the impact of autorate on our results.                    One should bear in mind that autorate is likely to improve
Unfortunately, since autorate algorithms are not standard, our            performance only for cases in which the minimum rate pro-
results are specific to the particular implementation in our               vides some connectivity. A node pair with zero throughput
cards.                                                                    across all configurations will not benefit from autorate. Indeed,
    We enabled autorate on all wireless cards in our testbed              in Figure 15 links with zero throughput with txrate=2Mbps
in ukhome1 and ran a series of experiments to compare                     do not improve with autorate (e.g., link 5 → 4). On the other
UDP throughput when txpower=30mW with and without a                       hand, some links with non-zero throughput at txrate=2Mbps
fixed transmission rate, presented in Figure 15. We note that              and zero throughput at txrate=11Mbps, achieve non-zero
the link from node-2 to node-4 achieves zero throughput                   throughput with autorate enabled.
over an 11 Mbps link, while optimal throughput is achieved
over a 2 Mbps link. Autorate allows nodes to adjust the
                                                                                              V. I MPACT OF T OPOLOGY
link encoding rate to environmental conditions. Indeed our
results demonstrate that node-2 is capable of identifying the                The previous sections demonstrated that homes represent
appropriate transmission rate for the link to node-4 and using            a challenging environment for wireless networks, particularly
it. On the other hand, the link from node-5 to node-6 achieves            when used to support bandwidth-demanding applications, such
a much lower throughput using autorate than when the link                 as video streaming. Performance, in terms of both success rate
rate is fixed at 11 Mbps. In fact, the measured throughput                 and throughput, varies widely across links, and asymmetry
over the autorate link is similar to the throughput over the              is common. Performance of a given link can be difficult to
2 Mbps link. Apparently, the use of autorate in this case                 predict, since the exact position and orientation of individual
drops the transmission rate to 2 Mbps and does not recover.               nodes has a greater effect than the distance between nodes.
                                                                       to forward packets on behalf of other nodes. Thus, a wide
                                                                       variety of different topologies are possible. Typically a routing
                                                                       algorithm is used to evaluate various paths between each pair
                                                                       of communicating nodes and selects the one that optimizes
                                                                       a particular metric. In particular, mesh networking can allow
                                                                       direct communication between nodes, two-hop communication
                                                                       through a third node, or more hops when necessary. While
                                                                       each hop introduces an additional bandwidth penalty (since
                                                                       each hop contends for the same channel, as in the AP-based
                                                                       case), multiple high-quality links may be better than a single
                                                                       low-quality link. Previous work has shown that such multi-
                                                                       hopping strategies could offer higher throughput multi-hop
                                                                       connections, when the quality of the direct link is poor [13].
Fig. 15. Effect of autorate for ukhome1 at txpower=30mW (ukhome1 −     Multi-hopping mechanisms and protocols are currently being
layout1 − udp − autorate). Average throughput in Mbps top to bottom:   explored within the IEEE 802.11s task group [7].
1.31, 3.84, 3.73.                                                         In Figure 16 we present an example node layout to demon-
                                                                       strate schematically the topologies considered in this section.
                                                                       Each link is labelled with the link loss rate, as might be
   In such an environment, network topology can have a                 measured from our reachability test. Figure 16(a) demonstrates
significant impact on network performance. In this section, we          the “direct” topology that could be employed for the commu-
use the results from Sections III and IV to evaluate the impact        nication between nodes; every node communicates with every
of three possible topologies: AP-based communication, direct           other node using the direct path. Note that direct communica-
communication, and multi-hop (or mesh) communication.                  tion between node-6 and node-7 has a 80% loss rate in this
                                                                       example. Figure 16(b) presents an AP-based topology, with
A. Topologies                                                          node-5 acting as the AP. In this case, communication between
   Today most home wireless networks utilize an AP-based               node-6 and node-7, through node-5 achieves a 90% loss rate.
topology. Each packet originating from a node in the wireless          Lastly, Figure 16(c) presents an alternative topology, based
network is first transmitted to an access point. The AP then            on multi-hopping. If routing is determined using Dijkstra’s
forwards the packet to the destination, which may be located in        algorithms and link weights reflect loss rate, then the optimal
the wireless network or may be accessible through some other           routing for node-6 to reach every other node inside the network
(typically wired) interface. This topology is reasonable (though       is shown with dark lines. In this case, node-6 can reach node-7
not necessarily optimal) when most traffic flows between                 through node-2 with 100% success.
wireless stations and the Internet through an AP.                         The question we address in the remainder of this section
   AP-based topologies are less optimal for traffic that flows           is the impact of topology on home wireless network perfor-
between stations within the wireless network. And, as the              mance. If, for example, a mesh topology could significantly
number of wireless-enabled devices in the home increases               increase overall throughput in home networks, a change in the
the amount of intra-wireless-network traffic will also increase.        design and functionality of radios for consumer electronics
Examples of such traffic include computer-to-printer traffic, as         would be warranted.
well as communication between consumer electronic devices.
Such traffic pays a two-hop penalty, since all packets transmit-
                                                                       B. Methodology
ted by the source must be retransmitted by the AP, and the AP
contends for the same channel as the originating node when                To assess the potential gain from alternative topologies in
forwarding packets. This approach is particularly sub-optimal          the home, we need a means to map the throughput values mea-
if the communicating devices are relatively close, while the           sured through our experiments to the throughput that would be
AP is relatively far away. Moreover, given our evaluation of           achieved by different node pairs under alternative topologies.
wireless link characteristics, high-bandwidth communication            The fundamental difference between our measurements and
in home networks between the source and destination nodes              the throughput that would be achieved under the AP or mesh
and the AP are not assured. In addition, choosing a location           topology is that specific nodes now use intermediate nodes
for an AP that maximizes network throughput is non-trivial.            for their communication. The throughput achieved by node-A
   Direct communication between wireless stations is an alter-         to node-C through node-B can be estimated using the airtime
native to AP-based communication that eliminates the two-hop           consumed by each individual transmission. For this purpose,
penalty. These topologies are most relevant for traffic between         we make the pessimistic assumption that all nodes are within
two devices within the home. However, our results from the             the same collision domain.
previous sections demonstrate that home networks may not be               In our current measurements, the airtime consumed by node-
able to guarantee the interaction between every two devices in         A’s transmission to node-C is the inverse of the throughput
the home.                                                              of the link from node-A to node-C, or thr1 , which is the
   Mesh networking provides a third alternative to direct and          amount of time the wireless channel is busy. If node-A uses
AP-based communication. Mesh networks allow any node                   node-B as a relay, then an individual transmission now leads to
               (a) direct communication                    (b) AP topology - AP at node-5                           (c) mesh topology
Fig. 16.   Example topologies for the interactions sourcing at node-6. Link weights represent loss rates along the direct path.

                                                                                                                   TABLE II
two transmissions: one from node-A to node-B, and one from
                                                                                       M ULTIHOP THROUGHPUT FROM NODE -2 TO NODE -dst THROUGH
node-B to node-C, with respective airtimes of thr1 , thr1 .
                                                 A,B    B,C                                   NODE -rtr WHEN txrate=11Mbps, txpower=30mW
Consequently, the throughput that would be observed by node-
                                                                                                 (ukhome1 − layout1 − udp − multihop).
A toward node-C through AP node-B can be expressed as:
                                            1                                                 1st hop    2nd hop     multihop     estimate    abserr   relerr
                      thrS,D =        1            1                      (1)    dst    rtr   (M bps)    (M bps)     (M bps)      (M bps)    (M bps)    (%)
                                   thrS,A   +   thrA,D                            3      4     5.04        6.29       3.14          2.79       0.35      10
                                                                                  3      5     5.18        2.93       1.83          1.87       0.04       2
   The above formula describes the airtime for a transmission                     3      6     5.82        6.30       3.05          3.02       0.02       0
between node-A and node-C through node-B as the sum of the                        3      7     6.11        6.27       3.10          3.09       0.00       0
air times for the transmissions between node-A and node-B,                        4      3     6.25        6.08       3.34          3.08       0.25       7
                                                                                  4      5     5.07        5.49       3.14          2.63       0.50      16
and node-B and node-C.                                                            4      6     6.02        6.15       3.19          3.04       0.14       4
   We validate this formula experimentally and use it to assess                   4      7     6.11        6.12       2.57          3.05       0.48      18
the impact of alternative topologies in the studied homes. For                    5      3     6.26        5.20       2.91          2.84       0.06       2
                                                                                  5      4     5.57        4.14       2.28          2.37       0.09       4
the validation step we measure the multihop effect as observed                    5      6     5.99        5.86       3.06          2.96       0.09       3
from node-2 in ukhome1 when txrate is 2Mbps or 11Mbps.                            5      7     6.11        5.15       2.37          2.79       0.42      17
Our experiments cover the full set of combinations for destina-                   6      3     6.26        6.12       3.24          3.09       0.14       4
tion and relay nodes, and results are presented in Table II for                   6      4     5.57        6.15       3.17          2.92       0.24       7
                                                                                  6      5     5.02        3.63       2.09          2.10       0.01       0
the case in which txrate=11Mbps. Each experiment measures                         6      7     6.11        5.65       3.25          2.93       0.31       9
the throughput of the two single-hop paths comprising the                         7      3     6.26        6.13       3.23          3.09       0.13       4
multi-hop path. Then the source and relay nodes are configured                     7      4     5.41        6.15       3.13          2.87       0.25       8
                                                                                  7      5     5.01        4.60       2.60          2.39       0.20       7
appropriately to facilitate multihop communication, and the                       7      6     5.95        6.05       2.97          2.99       0.02       1
end-to-end throughput is measured. Note that the experimental
measurement of the two single-hop throughputs preceded the
measurement of the multihop throughput, and therefore slight
                                                                                 throughput achievable by different pairs of communicating
variations should be expected.
                                                                                 nodes, when a particular testbed node is selected to act as
   The absolute and relative errors for txrate=11Mbps are
                                                                                 an AP. For instance, assuming that the AP is located at the
shown in Table II. The absolute error (the magnitude difference
                                                                                 position of node-2 we quantify the throughput achieved from
between the measured and estimated values in Mbps) incurred
                                                                                 node-X to node-Y given that it has to multihop through node-
using Eq. 1 is small. The distribution of the relative errors (the
                                                                                 2. Our results are presented in Figure 18.
percentage difference between the measured and estimated val-
                                                                                    Figure 18 clearly demonstrates the impact of the location
ues) incurred across both rates and for all the possible multihop
                                                                                 of the AP on the network performance. Only one of the six
connections initiating from node-2 is further shown in Figure
                                                                                 node locations in the testbed (node-3) leads to a network in
17. We notice that 80% of the relative error values obtained
                                                                                 which no pair of nodes experiences zero throughput. For any
when txrate=2Mbps are below 5%. The respective number
                                                                                 other choice of AP location (of the six possible locations),
for txrate=11Mbps is slightly greater (10%), which is a direct
                                                                                 35%-80% of the links are unusable (i.e. zero throughput).
consequence of the fact that a slight change in the conditions
across the measurement of thrA,B , thrB,C , thrA,C has a                            Next we compare the network performance in each home
more significant impact. Given the results in Table II and Fig-                   for the direct topology, the best AP topology, and the optimal
ure 17, we conclude that the “airtime” metric is appropriate for                 multi-hop topology. Note that the best topology for each home
the investigation of the performance implications of alternative                 is determined using the same method shown in Figure 18.
topologies in the home.                                                          Multihop routing is derived using Dijkstra’s algorithm, where
                                                                                 link weights are proportional to measured loss rate. While this
                                                                                 routing scheme determines optimal routes, it is important to
C. Findings                                                                      note that it does not necessarily provide an optimal routing
  Previously we conjectured that random selection of the                         algorithm. The reader interested in mesh routing can refer to
location of the Access Point (AP) in the home network is not                     [6] and references therein.
guaranteed to lead to a high performance wireless network.                          Results for each of ushome1, ushome2, and ukhome1 are
Using the collected throughput measurements we quantify the                      presented in Figure 19 and brief overall summaries are given
                        (a)                                                  (b)                                             (c)
Fig. 19. The impact of different topologies on UDP throughput (a) ushome1, (b) ushome2 and (c) ukhome1 (txrate=11Mbps, txpower=30mW, ∗ − layout1 −
udp). Throughput statistics can be found in Table III.

                                                                                                              TABLE III
                                                                                       P ERFORMANCE STATISTICS FROM ALL THREE HOMES UNDER
                                                                                      ALTERNATIVE TOPOLOGIES   (∗ − layout1 − udp − multihop).

                                                                                   Home      Metric           Direct   best AP-topology   Multihop
                                                                                   ushome1   Minimum Thr.        0             0             2.2
                                                                                             Average Thr.      3.15          3.07           3.79
                                                                                             Aggregate Thr.    94.58         92.26         113.88
                                                                                   ushome2   Minimum Thr.      1.23          1.14            2.3
                                                                                             Average Thr.      2.89          2.81           3.55
                                                                                             Aggregate Thr.    86.88         84.43         106.61
                                                                                   ukhome1   Minimum Thr.        0             0             2.2
                                                                                             Average Thr.      4.24          3.16           4.35
                                                                                             Aggregate Thr.   127.36         94.88         130.41

Fig. 17. Relative error in the estimation of multihop throughput (ukhome1−
layout1 − udp − multihop).
                                                                               as the best AP topology. In many cases, the mesh topology
                                                                               greatly surpasses the best AP topology, simply by allowing
                                                                               direct links to be used, without requiring that they be used in
                                                                               the cases where direct links provide low (or no) throughput.

                                                                                                      VI. R ELATED W ORK
                                                                                  Several recent studies have evaluated large wireless net-
                                                                               works deployed across university campuses. Kotz and Essien
                                                                               [14] studied a 476 access point wireless network deployed
                                                                               across a large campus, focusing primarily on user traffic
                                                                               characteristics rather than link performance measurements.
                                                                               Other studies have investigated the characteristics of wireless
                                                                               links in sensor networks. Zhao and Govindan [10] measured
                                                                               the link characteristics of 60 sensor nodes deployed in an
Fig. 18. The impact of AP location (txrate=11Mbps, txpower=30mW,               office building, an outdoor park, and a parking lot. The study
ushome1 − layout1 − udp − multihop).
                                                                               finds that many links operate in a ”gray area” with difficult-
                                                                               to-predict intermediate loss rates and performance.
                                                                                  The viability of multi-hop wireless mesh network topologies
in Table III. The AP-based topology performs the worst, with                   has been demonstrated for large outdoor and office networks
lower average and aggregate throughput in nearly every case,                   in recent trial network and testbed deployments. Aguayo,
and lower minimum throughput than both direct and multihop                     et al. [1] deployed and maintained a 50-node community
topologies in ushome2. As expected, the direct communication                   wireless mesh network used by students and faculty across
topology does eliminate the two-hop penalty (except in the                     a large university campus. In this deployment, each node in
case that either the source or destination is the AP). Thus,                   the network was able to communicate with other nodes and
many additional node pairs are able to achieve the highest pos-                a small number of gateways to the Internet by using multi-
sible throughput. However, the direct topology also disables                   hop routing and forwarding. Draves, et al. [6] used a 23-
communication between many node pairs. The mesh topology                       node wireless testbed deployed in an office environment to
neither improves nor harms the throughput of the links with                    measure the performance of multi-hop mesh routing protocols
the highest throughput. At the same time, the mesh topology                    and metrics. Their study demonstrates that selecting multi-hop
is, in every case, able to provide at least as much throughput                 routes that minimize end-to-end airtime based on link-level
measurements results in high-throughput performance in both           While not typically controllable by a home user, we found
single-radio and multi-radio mesh networks.                        that topology had the largest impact on overall network perfor-
   While it is not unexpected that wireless link performance       mance in the home. Using throughput measurements collected
will vary when deployed across large geographic areas, our         from all three homes, we have shown that the location of the
study focuses specifically on the characteristics of home           access point can have a dramatic effect on the performance of
networks and demonstrates that variations in link quality are      a wireless network. In many cases, a given AP deployment
very common even when wireless networks are deployed               will not yield a connected network. Since AP deployment
within the relatively small area of a home. Moreover, this         is typically determined by the point of entry of the Internet
study demonstrates that the benefits of wireless mesh network       service and aesthetic concerns, more flexible topologies may
topologies are not limited to wireless deployments spread          be more appropriate.
across large campuses or office buildings. Mesh networking             We considered two other types of topologies (e.g. direct
also improves the performance and reliability of small home        and mesh) and found that mesh topologies offered significant
wireless networks.                                                 benefits in the home. By selecting the topology according to
   In a previous study, we provided early evidence of signifi-      measured link characteristics, such as loss rate, a mesh can
cant variability and asymmetry in home network link quality        provide more uniform connectivity while also allowing high-
[11]. We have extended this early work in several important        performance direct links where available. Used alone or in
respects: (1) all new data sets, validating our previous results   combination with the other wireless configuration parameters,
and claims, (2) measurement of throughput using TCP and            mesh topologies can increase the performance of wireless
UDP transports, (3) evaluation of the impact of automatic          networks in the home. These results suggest a need for support
rate selection, and (4) comparison of the impact of flexible        of mesh networking capabilities in wireless-enabled consumer
topologies on the performance of home wireless networks.           electronic devices.

                     VII. C ONCLUSIONS                                                          R EFERENCES
                                                                    [1] D. Aguayo, J. Bicket, S. Biswas, G. Judd, and R. Morris, “Link-
   Using six-node testbeds deployed in three different houses           level measurements from an 802.11 mesh network,” in Proc. of ACM
in the United States and the United Kingdom we studied                  SIGCOMM, Portland, OR, Sept. 2004.
the properties of home wireless networks. We showed that            [2] W. Rothman and T. Bradley, “Wi-fi versus your walls,” This Old House
                                                                        Magazine, vol. 90, pp. 71–73, July/August 1995.
despite a home’s relatively small size, the home is not a           [3] IEEE 802.11 Task Group e, “Draft supplement to standard for telecom-
benign networking environment, and omnipresent connectivity             munications and information exchange between systems–lan/man spe-
is not guaranteed. Homes tend to feature wireless paths with a          cific requirements–part 11: Wireless medium access control (mac) and
                                                                        physical layer (phy) specification. medium access control (mac) quality
variety of obstacles which may render wireless communication            of service (qos) enhancements,” IEEE std 802.11e draft 13.0, Jan. 2005.
impossible between node pairs.                                      [4] S. J. Fortune, D. M. Gay, B. W. Kernighan, O. Landron, R. A.
                                                                        Valenzuela, and M. H. Wright, “Wise - a wireless system engineering
   Our results demonstrate that while wireless links inside             tool,” IEEE Computational Science and Engineering, vol. 2, no. 1, pp.
homes tend to be stable over time there is significant per-              58–68, Mar. 1995.
formance variation across links, and many links are highly          [5] A. Verstak, J. He, L. T. Watson, N. Ramakrishnan, and C. A. Shaffer,
                                                                        “S4w: Globally optimized design of wireless communication systems,”
asymmetric. In home environments, precise node location                 in Proc. of Parallel and Distributed Processing Symposium, Ft. Laud-
is perhaps the single most important factor determining the             erdale, FL, Apr. 2002.
quality of wireless communication. Indeed our results clearly       [6] R. Draves, J. Padhye, and B. Zill, “Comparison of routing metrics
                                                                        for static multi-hop wireless networks,” in Proc. of ACM SIGCOMM,
confirm that distance has no impact on the quality of wireless           Portland, OR, Sept. 2004.
links in the home, while small changes in antenna orientation       [7] IEEE 802.11 Task Group s, “Ieee 802.11 ess mesh
and node location can dramatically change the performance of            networking par (802.11s),” May 2004. [Online]. Available:
an individual link.                                                 [8] J. Malinen, host AP driver for Intersil Prism2/2.5/3,
   We have shown that the configuration parameters at the      
user’s disposal today, power control, physical layer selection,     [9] J.     Bicket,     madwifi      Stripped     /   Click    Wifi    Driver,
and link rate, are not effective in achieving high-performance     [10] J. Zhao and R. Govindan, “Understanding packet delivery performance
connectivity throughout a home. We found little difference              in dense wireless sensor networks,” in Proc. of ACM Conference on
between network performance at a moderate transmit power                Embedded Networked Sensor Systems (SenSys), Los Angeles, CA, Nov.
and the maximum transmit power. Links in our IEEE 802.11a          [11] M. Yarvis, K. Papagiannaki, and W. Conner, “Characterization of 802.11
and 802.11b networks had similar overall link characteristics,          wireless networks in the home,” in Proc. of 1st Workshop on Wireless
although the performance of 802.11a appears to be slightly              Network Measurements (WiNMee), Riva del Garda, Italy, 2005.
                                                                   [12] K. Papagiannaki, M. Yarvis, and W. Conner,
better in the home. In both 802.11a and 802.11b networks,     ˜kpapagia/infocom06 techreport.html.
operating at the highest possible data rate tended to eliminate    [13] S. Lee, S. Banerjee, and B. Bhattacharjee, “The case for a multi-
connectivity on a subset of links. Thus, the user is forced             hop wireless local area network,” in Proc. of Conference of the IEEE
                                                                        Communications Society (INFOCOM), Hong Kong, Mar. 2004.
to either choose a lower data rate, sacrificing performance         [14] D. Kotz and K. Essien, “Analysis of a campus-wide wireless network,”
for connectivity, or use an auto rate selection algorithm.              in Proc. of International Conference on Mobile Computing and Net-
Unfortunately, auto rate selection may also fail to achieve the         working, Atlanta, GA, Sept. 2002.
highest possible data rate. Moreover, auto rate selection cannot
help if connectivity is poor at all rates.