[Info-vax] OT: TCP: trading latency for bandwidth

[John Wallace]

On May 17, 9:15 pm, JF Mezei <jfmezei.spam... at vaxination.ca> wrote:
> someone wrote:
> >> By deficiencies, are you referring to the slow performance?  I've always found the TCP/IP stacks on VMS to be incredibly sluggish.  It's not just a bandwidth problem, it's very slow (relatively speaking) at reading and sending very small packets.  The situation on a platform like Linux is quite different.  It's very tragic.
>
> At one point, I thought I was having IP performance problems on the link
> and was using FTP as a means to check upload and download speeds (since
> VMS doesn't have a working  flash which is used for most speed tests).
>
> ftp tests showed poor performance.
>
> I did the same tests on my xserve (OS-X) and low and behold, the
> performance difference was very drastic and make full use of the line
> speed  (showing the line had no problems).
>
> I had already set a large TCPIP$FTP_WINDOW_SIZE (if I recall correctly)
> and it didn't seem to make a difference.
>
> There may be some paarmeters that can be set in TCPIP$ETC file
> (systemconfig or something akin to this), but out of the box, the
> performance was significantly lower than on OS-X (which came from freeBSD)
>
> It is also possible that the FTP application was not properly ported and
> fails to use any performance enhnacements possible on VMS.
>
> The end result is that the peformance was sigificantly slower on VMS
> than on OS-X.

Back in the days before dslreports had their crash, they had some
interesting writeups on the behaviour of TCP slow start across
differering implementations. Maybe they still do e.g. there's a brief
mention at http://www.dslreports.com/speed.

Anyway, unless a bulk data speed test runs for many seconds AND the
link is sufficiently error free that there is no packet loss  AT ALL
at low level, it is entirely possible that the results of an arbitrary
speed test can be dominated by the behaviour of TCP slow start rather
than by the actual usable bandwidth between sender and receiver. Not
saying it always will be, just that it can happen.

DSLreports tweak test tells you a lot about your IP settings but
didn't comment on "slow start" incompatibilities as such last time I
looked (a long time ago).

I like the maxspeed test at www.numion.com; you can get a graph from
it which potentially illustrates "slow start" and packet loss effects.