Docstoc

U160-Benchmarks-AMI

Document Sample
U160-Benchmarks-AMI Powered By Docstoc
					Benchmark Results
MegaRAID Ultra160 Controllers
Express 500, Elite 1600, & Enterprise 1600
Introduction
         Benchmark tests are used by nearly everyone when
         initially evaluating various products. There are many
         different benchmarking utilities available on the market,
                                                                                Saccharine VS
                                                                                       Sugar
         some freeware and others for a nominal fee. These
         tests include WinBench and Winstone from ZD Labs,
         IOmeter from Intel, HD Tach from Testa Labs and                SYNTHETIC VERSUS REAL-WORLD
         SCSIBench32 from Adaptec.                                      BENCHMARKING
                                                                        This issue has been the cause of much
W HAT IS A BENCHMARK?                                                   turmoil and consideration in the past. What
         A computer benchmark is a program that measures one            is defined as a synthetic benchmark is
                                                                        software that has been coded specifically
         or more performance attributes of a computer system,           to evaluate one or more aspects of a
         and provides a relative score to allow comparison              computer's performance. That's how this
         between different computer systems. It's important to          paper defines a computer benchmark.
                                                                        However, there is another method of
         distinguish between the software that generates results        evaluating system performance that relies
         and the results themselves. Some computer                      on real-world applications: logically, we
         benchmarks exist to measure specific system                    term this "real-world benchmarking".

         resources, such as CPU or memory performance,                  Real-world benchmarking came about
                                                                        mostly through games. Doom was one of
         others focus on a more general level of hardware               the first games that incorporated a method
         performance, such as video performance, while still            for determining frame rates in the game. It
                                                                        was soon discovered that this frame rate
         others provide an overall score for the entire computer
                                                                        tracker was a great way to benchmark a
         system.                                                        computer's video subsystem. Duke Nukem
         There are four general categories of benchmark tests:          3D provided a similar frame rate counter,
                                                                        then Quake...and the rest is history. Today,
         •     Application-based tests, which run real                  more and more 3D games include some
                                                                        form of frame rate tracking to allow for
               applications and time them. These are sometimes          benchmarking.
               referred to as 'Real-world Benchmarks.' (see             There are also real-world benchmarks
               sidebar)                                                 based on such applications as 3D Studio.
                                                                        Finally, there are hybrid's such as ZD Labs'
         •     Playback tests, which use logs of system calls           Winstone utility: a synthetic benchmark
               made during specific application activities (e.g.
               graphics calls or disk usage) and play them back in isolation.
         •     Synthetic tests, which approximate application activity in specific subsystems.
         •   Inspection tests, which do not attempt to mimic application activity, but instead
             directly exercise specific subsystems.
         Application-based tests give you the best measure of how your entire system
         performs in the real world. Where it's feasible, playback technology gives you the
         most realistic way to measure individual subsystems in isolation. Synthetic tests
         remain the state of the art for measuring processor performance. Likewise,
         inspection tests have their place in verifying bug-free behavior and measuring
         performance on an operation-by-operation basis.
A Closer Look At The Tests Used
IOMETER
          IOmeter is Intel's benchmarking utility. As stated on their website:
             IOmeter is an I/O workload generator and performance analysis tool for single
             and clustered servers. It works like a dynamometer for server I/O: it measures
             the I/O performance of the system while stressing it with a controlled workload.
             You can configure IOmeter to simulate the workload of any real-world
             application or benchmark, or create a completely artificial workload to stress
             the system in specific ways. IOmeter will accurately and reliably reproduce any
             workload you specify, while gathering data such as throughput, latency, and
             CPU utilization.       By running the same workload on multiple system
             configurations, you can determine the strengths and weaknesses of each and
             select the optimum configuration for your needs. *

          AMI has used IOmeter for many years, as it is accepted industry-wide, by OEMs and
          integrators alike, as a standard benchmarking tool for RAIDed systems. IOmeter is
          best used when evaluating server level systems, although the results obtained can
          provide valuable insight into possible workstation and desktop performance. Version
          1999.10.20 of IOmeter has been used for all the tests performed at AMI.

SCSIBENCH32
          This utility from Adaptec is designed primarily for                    Agreement &
          desktop SCSI benchmarking. SCSIBench32 is simple
          to setup and easy to use. Although it is not designed for           Interoperability
          benchmarking of server level systems, the numbers               WHAT ABOUT ADAPTEC?
          obtained can give a glimpse into possible performance
                                                                          These benchmarks were all obtained using
          in higher end systems. SCSIBench32 is normally                  controllers available to the public using
          bundled with Adaptec's EZ-SCSI suite of software,               normal means. There were two separate
                                                                          benchmark utilities used: IOmeter and
          designed for Desktop SCSI management. The version               Adaptec's      own     SCSIBench32    Utility.
          used in these tests was Version 5.0.                            Although attempts were made to use the
                                                                          utility from Adaptec for all tests due to its
A Glance at the Controllers                                               ease of use, this was impossible. AMI's and
                                                                          Mylex's controllers generated expected
          Which controllers were part of these tests? In order to         results with SCSIBench32, however when
          prevent confusion by including too many different               attempting      to   test  Adaptec's    own
                                                                          controllers, the utility was unable to
          manufacturers, only AMI, Adaptec and Mylex controllers          recognize them. Every attempt was made
          were included in these tests. In addition, an attempt           to have the utility recognize the Adaptec
          was made to limit, where possible, the controllers to only      controllers, including contacting Adaptec's
                                                                          own tech support, but no solution was
          those which provide support for Ultra160. These are the         found.       In order to provide some
          controllers which were included in the various tests            comparison with the Adaptec products,
          conducted by AMI:                                               IOmeter was considered to be the next best


          AMI:
              •    Express 500
              •    Elite 1600
              •    Enterprise 1600

          Adaptec:
              •    2100S
              •    3200S
              •    3400S

          Mylex:
              •    AcceleRAID170
Spec by Spec Comparison
        The following quick reference chart shows some simple comparisons of the controller
        specifications. A more detailed specification comparison can be found on AMI's
        website at www.ami.com or obtained from your local AMI sales representative or
        authorized AMI reseller/distributor.


                                              AMI                              Adaptec                 Mylex
         Spec:                   Express    Elite      Enterprise                                     AcceleRAID
                                                                    2100S       3200S      3400S
                                   500      1600         1600                                            170
         Technology              Ultra160   Ultra160     Ultra160   Ultra160    Ultra160   Ultra160     Ultra160
         PCI bus (bit) /          32 / 33   64 / 66      64 / 66    32 / 33     64 / 33    64 / 33      32 / 33
         PCI bus speed (MHz)
         Hardware / software     hardware   hardware     hardware   firmware    hardware   hardware
                                                                                                        firmware /
         XOR                                                                                            hardware
         PCI card size             Half      Half          Full      Half         3/4        Full         Half
         (Full or Half size)
         Number of channels         1          2            4          1           2          4            1
         Online capacity           Yes       Yes          Yes        Yes          No         No           Yes
         expansion
         Immediate RAID            Yes       Yes          Yes        Yes         Yes        Yes           Yes
         availability
         Bootable CD-ROM           Yes       Yes          Yes        Yes         Yes        Yes           Yes
         support
         NT clustering support     Yes       Yes          Yes        Yes         Yes        Yes           No
The Benchmark Results
         These are the benchmark results obtained by AMI in controlled environments. Since
         different benchmark tests were used, the configuration of the controllers changed
         slightly from test to test. During each individual test, however, all controllers
         participating in that test were configured as close to identically as possible, given the
         differences in implementation of features between manufacturers.

FIRST GLANCE: AMI VS AMI
         Test Used: SCSIBench32
         Controllers: AMI Express 500, Elite 1600, & Enterprise 1600

                                   RAID 0 Sequential Throughput

                  42200                                              42101
                  42000

                  41800
                  41600                              41524

                  41400                41300
                  41200

                  41000
                  40800

                                    Express 500    Elite 1600    Enterprise 1600




                                   RAID 1 Random Throughput

                 11500
                                                                      11392
                 11400
                 11300
                 11200                               11141
                 11100
                 11000
                                      10885
                 10900
                 10800
                 10700
                 10600
                                    Express 500    Elite 1600    Enterprise 1600




                                     RAID 5 Random Throughput

                 11800                                               11717
                 11700                               11653
                 11600
                 11500
                 11400
                 11300
                                      11205
                 11200
                 11100
                 11000
                 10900

                                     Express 500    Elite 1600    Enterprise 1600
ENTRY LEVEL PART ONE: AMI EXPRESS 500 & M YLEX ACCELERAID 170
         Test Used: SCSIBench32
         This is the only test where controller default settings were left in effect. In other
         words, both controllers were taken directly 'from the box' and placed into the system
         for testing without any additional configuration. This was done to simulate what many
         channel customers will do when using RAID controllers - ie. Plug and Play.

                                          Same Sector Throughput Comparison

                                         40500
            Throughput (kb/s)




                                         40000

                                         39500


                                         39000
                                                       Logical Drive 1                    Logical Drive 2

                                  AcceleRAID 170           40000                              40000
                                  Express 500              40259                              40131
                                                                         Logical Drives




                                                Random Throughput Comparison

                                         15000
              Throughput (kb/s)




                                         12500



                                         10000
                                                       Logical Drive 1                    Logical Drive 2

                                  AcceleRAID 170           12226                              10735
                                  Express 500              13500                              13500

                                                                         Logical Drives




                                           Sequential Throughput Comparison

                                         55000
             Throughput (kb/s)




                                         45000

                                         35000

                                         25000
                                                       Logical Drive 1                    Logical Drive 2

                                  AcceleRAID 170           30000                              48000

                                  Express 500              35000                              50000

                                                                         Logical Drives
ENTRY LEVEL PART TWO: AMI EXPRESS 500 & ADAPTEC 2100S
         Test Used: IOmeter
         RAID 5 Sequential Performance

                     120                                                            45                      42.10
                                         108.16
                                                                                    40
                     100
                                                        81.96                       35
                      80                                                            30




                                                                         MB/sec
            MB/sec

                                                                                    25
                      60
                                                                                    20
                      40                                                            15
                                                                                    10
                      20                                                                                                      4.10
                                                                                     5
                          0                                                          0
                                          64k Sequential Read                                              64k Sequential Writes
                                   AMI Express 500     Adaptec 2100S                                AMI Express 500        Adaptec 2100S




         RAID 5 Random Performance

                     35                                                            8
                                         30.40                                                              6.93
                     30                                                            7

                     25                                                            6
                                                                                   5
           MB/sec




                     20                                17.11             MB/sec
                                                                                   4
                     15
                                                                                   3
                     10                                                                                                      1.85
                                                                                   2
                     5                                                             1
                     0                                                             0
                                         64k Random Reads                                                   64k Random Writes
                                 AMI Express 500     Adaptec 2100S                                 AMI Express 500     Adaptec 2100S




         RAID 0 Sequential Performance

                     120.00                109.47                                          80.00
                                                                                                                   68.99
                                                                                           70.00
                     100.00                              87.16
                                                                                           60.00
                      80.00
                                                                                           50.00
            MB/sec




                                                                                  MB/sec




                      60.00                                                                40.00
                                                                                                                                27.96
                                                                                           30.00
                      40.00
                                                                                           20.00
                      20.00                                                                10.00
                          0.00                                                             0.00
                                           64k Sequential Read                                                 64k Sequential Writes

                                     AMI Express 500     Adaptec 2100S                                    AMI Express 500       Adaptec 2100S
M ID RANGE HEAD TO HEAD: AMI ELITE 1600 & ADAPTEC 3200S
         Test Used: IOmeter
         RAID 5 Sequential Performance

             250.00                                       40.00
                                                                       33.65
                                                          35.00
                          192.72
             200.00
                                                          30.00


             150.00                                       25.00
                                                                                      18.93
                                                          20.00
             100.00                        77.11          15.00

                                                          10.00
              50.00
                                                           5.00

               0.00                                        0.00
                              64k Sequential Read                       64k Sequential Writes


                       AMI Elite 1600     Adaptec 3200S            AMI Elite 1600    Adaptec 3200S




         RAID 0 Sequential Performance

             300.00                                       120.00       111.14
                          257.86
             250.00                                       100.00


             200.00                                        80.00

                                          143.33
             150.00                                        60.00

             100.00                                        40.00                       29.49

              50.00                                        20.00

               0.00                                         0.00
                              64k Sequential Read                        64k Sequential Writes


                       AMI Elite 1600     Adaptec 3200S            AMI Elite 1600     Adaptec 3200S
HIGH END PRIZE FIGHT: AMI ENTERPRISE 1600 VS ADAPTEC 3400S
         Test Used: IOmeter
         RAID 5 Sequential Performance

                                      187.68
                     200.00                                                    40.00
                                                                                                33.65
                                                                               35.00
                     150.00                                                    30.00
                                                                               25.00
            MB/sec




                                                                      MB/sec
                     100.00                                                    20.00                          17.35
                                                    76.31
                                                                               15.00
                      50.00                                                    10.00
                                                                                5.00
                       0.00                                                     0.00
                                      64k Sequential Read                                      64k Sequential Writes

                              AMI Enterprise 1600     Adaptec 3400S                     AMI Enterprise 1600    Adaptec 3400S




         RAID 0 Sequential Performance

                                     369.20                                                      95.12
                     400.00                                                    100.00
                     350.00
                                                                                80.00
                     300.00
                     250.00
            MB/sec




                                                                      MB/sec
                                                                                60.00
                     200.00                         140.17
                     150.00                                                     40.00                         29.05
                     100.00
                                                                                20.00
                      50.00
                       0.00                                                      0.00
                                      64k Sequential Read                                      64k Sequential Writes

                              AMI Enterprise 1600    Adaptec 3400S                      AMI Enterprise 1600    Adaptec 3400S

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:8
posted:8/27/2012
language:
pages:8