| T.R | Title | User | Personal Name
 | Date | Lines | 
|---|
| 1765.1 |  | STAR::STOCKDALE |  | Thu Aug 03 1995 13:41 | 10 | 
|  | As far as I know, there is nothing the VMS driver has to do.  All that is
needed is to have V2.0 of the firmware in the DEMFA and connect it a port
that supports full duplex.
Not that there is any good reason to use full duplex on a DEMFA, unless you
can get excited over a 10 to 15 mbit improvement.  (The best I ever saw was
around 110-115 mbits/sec - its hard to tell the DEMFA is running full-duplex
since the difference is so small).
- Dick
 | 
| 1765.2 |  | STRWRS::KOCH_P | It never hurts to ask... | Thu Aug 03 1995 14:22 | 4 | 
|  |     
    I agree with .1, but even though it doesn't really increase thruput,
    wouldn't it have an effect on latency given you could be doing
    simultaneous xfer & rcv?
 | 
| 1765.3 |  | STAR::STOCKDALE |  | Fri Aug 04 1995 16:28 | 5 | 
|  | Right, there should be some difference there, although on a VAX I doubt
the difference would be measurable (not enough computes and the driver
implementation is less than optimal).
- Dick
 | 
| 1765.4 |  | TPOV07::TONYCHENG |  | Tue Aug 08 1995 05:25 | 7 | 
|  |     
    I got trouble because i told my customer, our FDX function should boost
    FDDI throughput. Do we have the DEMFA performance analysis report ?
    Where can find it ?
    
    Regards
    Tony Cheng   
 | 
| 1765.5 |  | STRWRS::KOCH_P | It never hurts to ask... | Tue Aug 08 1995 08:19 | 3 | 
|  |     
    The question is, was this testing only done with a VAX 6600? Was it
    rerun using an Alpha system? Or is it a limitation of the controller?
 | 
| 1765.6 |  | STAR::STOCKDALE |  | Tue Aug 08 1995 18:34 | 1 | 
|  | Its a DEMFA hardware limitation.
 | 
| 1765.7 |  | STEVMS::PETTENGILL | mulp | Mon Aug 21 1995 21:11 | 17 | 
|  | >    I got trouble because i told my customer, our FDX function should boost
>    FDDI throughput. Do we have the DEMFA performance analysis report ?
>    Where can find it ?
"boost FDDI throughput" relative to what?
We measured an increased thruput over not-FDX when sending/receiving a high
percentage of back-to-back 4500 byte packets.  The only way to accomplish this
is with MSCP server traffic to a lot of disks.  Under these circumstances its
possible to move from the FDDI and/or GIGAswitch backplane limit of 97-98
megabits up to the DEMFA DMA limit of 115-120 megabits.  The DEMFA was designed
a long time ago before full duplex was considered so it provided 10-20% added
bandwidth to ensure that it wasn't the bottleneck.
We also did some testing to see if we could demonstrate reduced latency, but
I don't believe that we came up with anything convincing.  Its difficult to
do this was VAX systems since the clock resolution is so course (10ms).
 |