December 2009 doc.: IEEE 802.11-09/1317r1
Internet Traffic Modeling
Date: 2009-12-09
Authors:
Name Affiliations Address Phone Email
Sai Nandagopalan Broadcom San Diego 858 521 2192 nsai@broadcom.com
Vinko Erceg Broadcom San Diego 858 521-5885 verceg@broadcom.com
Submission Slide 1 Vinko Erceg, Broadcom
December 2009 doc.: IEEE 802.11-09/1317r1
Background
• In Nov 2009 TGad meeting in Atlanta, people wanted a simple model to
characterize internet traffic for TGad evaluation methodology document
• The people wanted to see if they could have similar requirements on this
traffic modeling as was done in TGn few years back
• So we have considered this new model for FTP and HTTP traffic
Submission Slide 2 Vinko Erceg, Broadcom
December 2009 doc.: IEEE 802.11-09/1317r1
Why Model Internet Traffic?
• Why model traffic?
– The traffic model is the key for determining the performance of the
system. The more accurate is the traffic model the better is the system
quantified in terms of its performance.
– Traffic model in the evaluation methodology document should focus
on capturing the accents of the application which posts special demand
on the system performance.
• What is being modeled here?
– We are modeling TCP and Web browsing
• This document proposes to changes that were presented in
IEEE 802.11-09/1216r1.
– Why?
– Almost all of us use standard simulators such as OPNET, NS2,
OMNET etc.
– These simulators are rich enough to have most types of TCP inbuilt
with all the TCP parameters.
– Additionally, we want to include reference to IEEE 802.11n
evaluation methodology document
Submission Slide 3 Vinko Erceg, Broadcom
December 2009 doc.: IEEE 802.11-09/1317r1
Modeling TCP Traffic
• Almost all the file transfers between individual devices
that are connected via IEEE 802.11 and its extensions
happens through TCP.
– This is the recommended protocol for all traffic that passes through the
IP stack
• We consider a simple file transfer.
– The file size is fixed at 10 GByte. In this profile the AP is connected to
the backbone. Since the packet size is limited by Ethernet, we set it as
1500 bytes
Submission Slide 4 Vinko Erceg, Broadcom
December 2009 doc.: IEEE 802.11-09/1317r1
TCP Procotol Configuration
TCP Configuration Parameters in Simulators
MSS Ethernet (1500 bytes)
Receive Buffer (Bytes) 65535
The table on the right outlines the Receive Buffer Adjustment None
TCP configuration in all simulators. Delayed ACK Mechanism Segment/Clock based
All these parameters are available. Max. ACK Delay (Sec) 0.05
These were the same parameters that Slow Start Initial Count (MSS) 1
Fast Retransmit Enabled
were used in IEEE 802.11n
evaluation methodology document. Duplicate ACK Threshold 3
Fast Recovery Reno
Window Scaling Enabled
Selective ACK (SACK) Disabled
ECN Capability Disabled
Segment Send Threshold Byte Boundary
Active Connection Threshold Unlimited
Karn’s Algorithm Enabled
Nagle Algorithm Disabled
Initial Sequence Number Auto Complete
Initial RTO (sec) 3.0
Min RTO (sec) 1.0
Max RTO (sec) 64
RTT Gain 0.125
Deviation Gain 0.25
RTT Deviation Coefficient 4.0
Timer Granularity 0.25
Submission Slide 5 Vinko Erceg, Broadcom
December 2009 doc.: IEEE 802.11-09/1317r1
Modeling HTTP Traffic (1)
• HTTP traffic characterization is quite complex and is a
function of
– Structure of WWW pages and
– Nature of human interaction
• Based on the above properties, it is bursty and is modeled
by ON/OFF sources
First Packet of Session Reading Time Reading Time Last Packet of Session
Parsing time
– ON periods represent the webpages being transferred from the server
to client and the OFF periods refers to the reading time of the user
– Also a webpage consists of multiple subpages and each main page
reference generates the request of those subpages
Submission Slide 6 Vinko Erceg, Broadcom
December 2009 doc.: IEEE 802.11-09/1317r1
Modeling HTTP Traffic (2)
• The parameters of HTTP file Component Distribution Parameters PDF
transfer are: Mean = 10710 bytes 1 ln x 2
fx exp ,x 0
– SM: Size of main object in page 2 x 2
Main
Truncated
SD = 25032 bytes 2
object
Lognormal
Min = 100 bytes 1.37, 8.37
size (SM) Max = 2 Mbytes
– Nd: Number of embedded objects (before truncation)
if x>max or xmax or xmax, discard and regenerate a
distributions new value for x
x
– Approximate mean data rate of Reading time
(Dpc)
Exponential Mean = 30 sec f x e ,x 0
= 0.033
(10KB+48KB)/31 = 12.31 kbps
x
Parsing time
Exponential Mean = 0.13 sec f x e ,x 0
(Tp)
7.69
Submission Slide 7 Vinko Erceg, Broadcom
December 2009 doc.: IEEE 802.11-09/1317r1
Conclusions
• We have outlined the models of TCP traffic and HTTP
traffic to be used for evaluation methodology document
for TGad
– The TCP traffic is the same as what was done in the evaluation
methodology document in IEEE 802.11n
– The HTTP traffic is represented by few distributions that have been
taken from extensive studies of references [1-6].
Submission Slide 8 Vinko Erceg, Broadcom
December 2009 doc.: IEEE 802.11-09/1317r1
References
• [1] P. Barford and M Crovella, "Generating Representative Web Workloads for Network and
Server Performance Evaluation" In Proc. ACM SIGMETRICS International Conference on
Measurement and Modeling of Computer Systems, pp. 151-160, July 1998.
• [2] S. Deng. “Empirical Model of WWW Document Arrivals at Access Link.” In Proceedings of
the 1996 IEEE International Conference on Communication, June 1996.
• [3] R. Fielding, J. Gettys, J. C. Mogul, H. Frystik, L. Masinter, P. Leach, and T. Berbers-Lee,
"Hypertext Transfer Protocol - HTTP/1.1", RFC 2616, HTTP Working Group,
ftp://ftp.Ietf.org/rfc2616.txt, June 1999.
• [4] B. Krishnamurthy and M. Arlitt, "PRO-COW: Protocol Compliance on the Web", Technical
Report 990803-05-TM, AT&T Labs, http://www.research.att.com/~bala/papers/procow-1.ps.gz,
August 1999.
• [5] B. Krishnamurthy and C. E. Wills, "Analyzing Factors That Influence End-to-End Web
Performance", Computer Networks: The International Journal of Computer and
Telecommunications Networking, Volume 33 , Issue 1-6, pp. 17-32, June 2000.
• [6] H. K. Choi, J. O. Limb, "A Behavioral Model of Web Traffic", Proceedings of the seventh
International Conference on Network Protocols, 1999, pp. 327-334, November 1999.
Submission Slide 9 Vinko Erceg, Broadcom