[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference 7.286::fddi

Title:FDDI - The Next Generation
Moderator:NETCAD::STEFANI
Created:Thu Apr 27 1989
Last Modified:Thu Jun 05 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2259
Total number of notes:8590

963.0. "FDDI benchmark speeds" by DECIDE::MOFFITT () Thu May 20 1993 12:03

    Help...
    
    A couple of months ago I saw a VAXmail message cross my screen
    concerning an impromptu benchmark (at Interop, I believe) of FDDI
    speeds between two Alpha platforms. Details are fuzzy but as I
    remember, the test was IP to IP (not at the driver level) and speeds
    measured in the 90 Mbps range in standard mode and somewhere over
    130Mbps in Digital (duplex) mode.
    
    For the life of me, I can't locate that message. Was I halucinating or
    did others see it also? If someone kept a copy would it be possible to
    post it here or send it to me at DECIDE::MOFFITT.
    
    thanks in advance,
    tim moffitt
T.RTitleUserPersonal
Name
DateLines
963.1I wouldn't call it a "bench mark" - just a bricks demoJEDI::CAUDILLKelly - NaC Tech Support - 264-3320Thu May 20 1993 12:2316
    No, you were not halucinating.  At spring Interop we had a GIGAswitch
    and connected to it were two Flamingos (with DEFTA) running OSF/1 and a
    dual-process Cobra (with two DEFAAs) running VMS.
    
    There are two, currently seperate, IP performance reports that were
    given out at that demo.  One is about the results of some performance
    improvements made to TCP/IP on DEC OSF/1 AXP.  The other is about the
    same improvements made to TCP/IP services for OpenVMS AXP (aka UCX V3).
    
    I don't know first hand about either operating system running on any
    other alphas, but both OSF/1 on a Flamingo and UCX on VMS on a (even
    single processor) Cobra can easily saturate an FDDI with TCP or UDP.  
    
    Maybe someone else can post pointers to the reports.  Maybe they are
    already posted in here somewhere.  If not, I have a couple ideas of
    people who should be able to post them.
963.2It's still thereSTKHLM::TORGNYMon May 24 1993 04:2834
            <<< MRKTNG::USERD:[NOTES$LIBRARY]UNIX-PARTNERS.NOTE;3 >>>
                               -< UNIX-PARTNERS >-
================================================================================
Note 477.28               ALPHA NEWS AND MISCELLANENOUS                 28 of 28
SMURF::POEGEL                                        27 lines   9-MAR-1993 16:14
     -< High Performance TCP/IP for OSF/1 Alpha AXP Workstations (White  >-
--------------------------------------------------------------------------------
Subject: High Performance TCP/IP for OSF/1 Alpha AXP Workstations (White Paper)

   From:   ERLANG::DSAP01::hawe "Bill Hawe"
   Subj:   Alpha/OSF TCP/IP performance paper we are publishing at InterOp this week

   Abstract:

     Today, Digital offers a winning solution for open high speed
     networking. Applications can now run at full FDDI speed, 100 megabits
     per second. The combination of Alpha AXP workstations, the DEC
     FDDIcontroller/TURBOchannel network adapter, and DEC OSF/1 Operating
     System software represents a signification price/performance 
     advantage for users. This white paper describes details of these 
     system components and explains how each one affects overall system 
     performance. Performance measurement results are also provided. The
     system components include:

	- The world-class Alpha AXP workstation series

	- A new low-cost FDDI network adapter

	- State-of-the-art OSF/1 operating system software with 
          BSD 4.3 Reno TCP/IP

   Location:  HUMAN::ALPHA_PUBLIC:HIGH_PERF_TCPIP_DEC_OSF1_AXP_WS.PS


963.3Big help!DECIDE::MOFFITTTue May 25 1993 15:113
    Yessir, that's the ticket! Thanks for the pointer!
    
    tim m.
963.4OpenVMS was sighted flying at 170 Mb/s on FDDIMUDDY::WATERSThu May 27 1993 16:3615
    The folks who set up the network demos for DECUS report the following
    tidbits.  These people were not intentionally benchmarking the systems
    for publication, so our best speed may be better yet.

    Using a DEC 4000/620 with two DEFAA Futurebus+ FDDI controllers, a
    VXT 2000 X Window terminal, OpenVMS AXP V1.5 and TCP/IP Services for
    OpenVMS V3.0, they report two bricks demos running over UDP/IP to
    two DEC 3000/500 workstations at over 170 Mb/s, and over 105 Mb/s
    using TCP/IP.

    Between the two DEC 3000/500 workstations with one DEFTA FDDI apiece
    and DEC OSF/1 AXP V1.2, bricks runs at full FDDI speed over either
    TCP/IP or UDP/IP.  This has been reported before.

    Bricks traffic was flowing through a GIGAswitch for these measurements.
963.5OSI performance on FDDI?JULIET::HATTRUP_JAJim Hattrup, Santa Clara, CAThu May 27 1993 23:423
    
    Any FDDI OSI data rates?   Any expectations relative to other 
    protocol data rates?
963.6Desciption of DECUS network and GIGAswitch/FDDI demoJEDI::CAUDILLKelly - NaC Tech Support - 264-3320Fri May 28 1993 11:2051
    
                  The network at Spring DECUS '93 in Atlanta

The DECUS network is an extended LAN consisting of multiple ethernet segments
bridged to an FDDI backbone. The backbone FDDI ring and several dedicated
FDDI rings are connected by a GIGAswitch.

Network components are housed in a Master Distribution Frame (MDF), located
in the Networks booth, and in six Intermediate Distribution Frames (IDFs)
distributed around the DECUS exhibit floor. IDFs are cabled to the MDF in 
a star configuration. 

The MDF contains a GIGAswitch, two FDDI concentrators (DECconcentrator 500s), 
and two FDDI to Ethernet bridges (DECbridge 620s), connected in a dual ring 
configuration. IDFs contain one FDDI concentrator, one or more DECbridge
500s or 620s, a terminal server, and LAN Bridge 100 operating as a LAN 
traffic monitor. The GIGAswitch, concentrators, and bridges in the MDF and 
IDFs form the show backbone FDDI ring. 

DECUS systems configured for FDDI are cabled directly to the concentrators
in the nearest IDF or the MDF. Ethernet systems are connected to one of the
ethernet segments created by the DECbridge 500s and 620s.

Additionally, in the Networks booth, there are two racks containing DECNIS
routers.  One DECNIS is connected directly to the FDDI ring ... it operates 
as the designated router for the network, and provides wide area DECnet, OSI, 
and IP connections. The other is connected to an ethernet segment and will 
be used to demonstrate new DECNIS down line loading and management features.


                             GIGAswitch/FDDI demo

Also connected to the GIGAswitch is a DEC 4000 Model 620 AXP departmental 
server and two DEC 3000 Model 500 AXP workstations.

The DEC 4000/620 has two DEFAA Futurebus+ FDDI controllers, and each controller
has a dedicated FDDI connection to the GIGAswitch. The 4000/620 is running 
OpenVMS AXP V1.5 and TCP/IP Services for OpenVMS V3.0.  Serving as a "console" 
for the 4000/620 is a VXT 2000 X Windows terminal. The VXT will be used to 
show Out-Of-Band management (OBM) on the GIGAswitch, MCC management of the 
GIGAswitch, bridges, and concentrators, and the bricks demo described below.

The two DEC 3000/500 AXP workstations each have one DEFTA FDDI controller and
a dedicated FDDI connection to the GIGAswitch.  The 3000/500 workstations are 
running DEC OSF/1 AXP V1.2.  They will be running the bricks demo.

The bricks demo is an FDDI performance demonstration. We can demonstrate
the DEC 4000/620 sending UDP/IP at over 170Mbps thru the GIGAswitch to the 
two DEC 3000/500 workstations.  Alternatively, TCP/IP can be sent at over 
105Mbps. Any of these systems can be shown sending TCP/IP or UDP/IP at full 
FDDI speeds. Full duplex FDDI can also be demonstrated.
963.7I haven't run bricks of DECnet/OSI on AXP - yetJEDI::CAUDILLKelly - NaC Tech Support - 264-3320Fri May 28 1993 11:2710
    re: .5, I have great faith in the DECnet/OSI performance, but I have
    not been able to spend the time to try it - sorry :-(
    
    I have lots of excuses...  DECnet/OSI for OpenVMS AXP field test kits
    were not available until too late in our staging schedule... I did not
    want to have DECnet/OSI on the OSF systems and not the VMS system - not
    that I care too much about being fair, I just didn't want that
    difference to take attention from the performance demo and the GIGAswitch.
    
    Maybe next time (ie Interop).