Re: souping up a UDB

Eric Youngdale (eric@sub2317.jic.com)
Sun, 3 Nov 1996 16:17:52 -0500

>I've been under the impression that it isn't likely a CPU bottleneck
>problem, but either a protocol latency, a protocol inefficiency (NFS
>@1k blocks), or a network latency that causes this loss in throughput.

If I were using NFS then yes this could explain some of it.
I was using ftp, and my understanding is that the daemon is pushing
data packets with the understanding that an acknowledgement would come
at some later time. Thus it should be shoving out other packets while
waiting for the response for the first one, all in order to increase
throughput.

Come to think of it, I do have to go through a router to get to the
local machine I was using, so perhaps a little additional overhead is to
be exptected.

>Isn't DS1 (aka T1) throughput 1.5mbps (x10e6 not 1024^2) or about
>150kBps?

This is close to where the ftp transfer rates were peaking when I was
going to a nearby machine which was outside of the cable company domain. I
think I actually got peak rates closer to 110-120 Kbytes/sec in the middle of
the night when the network was relatively idle. I had a cron job to benchmark
ftp transfers every 15 minutes 24 hours a day, and I let it run for about a
week so I could see where things actually maxed out.

-Eric

--
To unsubscribe: send e-mail to axp-list-request@redhat.com with
'unsubscribe' as the subject.  Do not send it to axp-list@redhat.com



Feedback | Store | News | Support | Product Errata | About Us | Linux Info | Search | JumpWords
No Frames | Show Frames

Copyright © 1995-1997 Red Hat Software. Legal notices