Changes between Version 12 and Version 13 of WARPLab/Benchmarks


Ignore:
Timestamp:
Feb 9, 2015, 2:19:05 PM (9 years ago)
Author:
welsh
Comment:

--

Legend:

Unmodified
Added
Removed
Modified

  • WARPLab/Benchmarks

    v12 v13  
    55In order to get a better understanding of WARPLab performance, we have added [wiki:Reference/Utility/#wl_benchmark wl_benchmark] so that you can perform some basic performance testing on WARPLab setup. 
    66
    7 For the WARPLab 7.3.0 release, we have provided some Ethernet performance benchmarks since one of the biggest bottlenecks is moving data into and out of the node. 
    8 
    97
    108== Ethernet Performance ==
    119
     10Since one of the biggest performance bottlenecks is moving data into and out of the node, we have provided some benchmarks on Ethernet performance.
     11
    1212As part of our on-going development to make WARPLab better and faster, we have introduced a new [wiki:WARPLab/Framework/Modules#Transport Transport] (ie, the [source:ResearchApps/PHY/WARPLAB/WARPLab7/M_Code_Reference/classes/wl_transport_eth_udp_mex.m WARPLab MEX UDP Transport]) to boost communication performance between a host and a node.  For more information on how to install and set up the WARPLab MEX UDP Transport, please refer to the [wiki:WARPLab/QuickStart QuickStart Guide].
    1313
    14 === Hardware and Software ===
    15 
    16   * Netgear ProSafe 5-port Gigabit Desktop Switch (GS105)
    17   * PC Hardware
    18     * Dell Optiplex 790
    19     * Intel Core i7-2600 3.4 GHz
    20     * 16 GB RAM
    21     * Window 7 Professional SP1 64-bit
    22     * Intel 82579LM Gigabit Network Connection
    23       * Configured to support Jumbo Packets of up to 9014 Bytes
    24   * PC Software
    25     * Matlab 7.12.0 (R2011a)
    26     * Microsoft Visual C++ 2010 Express (compiler for MEX)
    27     * Microsoft Windows 7 SDK
    28 [[BR]]
    29 === Procedure ===
    30 
    31   * Download Reference bit stream w/ iMPACT
    32   * Open MATLAB
    33     * Set current folder within MATLAB to the M_Code_Reference folder of the appropriate WARPLab release
    34     * Run the following commands
    35       * wl_setup
    36         * NOTE:  This allows us to configure the Transport type and the Ethernet Packet Size
    37       * nodes = wl_initNodes(1)
    38       * results = wl_benchmark( nodes, 'write_iq', 10000, 'max' );
    39       * results = wl_benchmark( nodes, 'read_iq', 10000, 'max' );
    40 [[BR]]
    41 === Results ===
    42 
    43 ''' WARP v3 Hardware '''
    44 
    45 Over 10,000 trials using 32768 samples per trial, we have the following Read IQ and Write IQ performance of the different Transports:
    46 
    47 [[Image(WARP_v3_Ethernet_Performance.png)]]
    48 
    49 {{{
    50 ------------------------------------------------------------------------------------------------------------------------------------
    51 |  ID |  WLVER |    Serial # |   Transport | Pkt Size (B) | Num Samples | Num Trials | Avg Write IQ per sec |  Avg Read IQ per sec |
    52 ------------------------------------------------------------------------------------------------------------------------------------
    53 |   0 |  7.3.0 |  W3-a-00027 |  WL Mex UDP |         1464 |       32768 |      10000 |             210.4994 |             232.6682 |
    54 |   0 |  7.3.0 |  W3-a-00027 |    Java UDP |         1464 |       32768 |      10000 |              17.0187 |              16.9429 |
    55 |   0 |  7.1.0 |  W3-a-00027 |    Java UDP |         1464 |       32768 |      10000 |              16.9390 |              16.6640 |
    56 ------------------------------------------------------------------------------------------------------------------------------------
    57 |   0 |  7.3.0 |  W3-a-00027 |  WL Mex UDP |         8960 |       32768 |      10000 |             445.8885 |             466.6237 |
    58 |   0 |  7.3.0 |  W3-a-00027 |    Java UDP |         8960 |       32768 |      10000 |              82.0430 |              71.2319 |
    59 |   0 |  7.1.0 |  W3-a-00027 |    Java UDP |         8960 |       32768 |      10000 |              73.1989 |              44.6655 |
    60 ------------------------------------------------------------------------------------------------------------------------------------
    61 }}}
    62 
    63 
    64 
    65 
    66 ''' WARP v2 Hardware '''
    67 
    68 Over 10,000 trials using 16384 samples per trial, we have the following Read IQ and Write IQ performance of the different Transports:
    69 
    70 [[Image(WARP_v2_Ethernet_Performance.png)]]
    71 
    72 {{{
    73 ------------------------------------------------------------------------------------------------------------------------------------
    74 |  ID |  WLVER |    Serial # |   Transport | Pkt Size (B) | Num Samples | Num Trials | Avg Write IQ per sec |  Avg Read IQ per sec |
    75 ------------------------------------------------------------------------------------------------------------------------------------
    76 |   0 |  7.3.0 |         N/A |  WL Mex UDP |         1464 |       16384 |      10000 |             128.1127 |             133.0076 |
    77 |   0 |  7.3.0 |         N/A |    Java UDP |         1464 |       16384 |      10000 |              32.9602 |              31.1658 |
    78 |   0 |  7.1.0 |         N/A |    Java UDP |         1464 |       16384 |      10000 |              32.5807 |              22.0276 |
    79 ------------------------------------------------------------------------------------------------------------------------------------
    80 }}}
    81 
     14  - Ethernet Performance Benchmarks
     15    - [wiki:WARPLab/Benchmarks/WARPLAB_7_3_0 WARPLab 7.3.0]
     16    - WARPLab 7.4.0 benchmarks are the same as WARPLab 7.3.0
     17    - [wiki:WARPLab/Benchmarks/WARPLAB_7_5_0 WARPLab 7.5.0]