Alpha Tulip performance

Wai-Sun Chia (waisun@msa.dec.com)
Sat, 27 Feb 1999 11:57:59 +0700


I've got an Avanti (AS200 4/166) and an Alcor (AS600 5/300) on RH5.2
running as Squid siblings (I have another i386 sibling). Traffic are
pretty heavy on the network I/O, cpu load is low to medium.

My problem is that for the Alpha squids, ping time on the local subnet
increases with time. i.e. network throughput slowly but surely
decreases; from a ping RTT of 3ms to 3000ms from another PC on the same
subnet. This 1000 fold degradation typically takes from a couple of
hours to maybe a day depends on traffic. After which the alpha boxes are
totally inaccessible either via ftp or telnet. This sucks because one of
them is my main ftp server!

The workaround is to do a network reset (S10network restart), and things
are zippy again (for now)! I'm sure this behavior is alpha-related as
the i386 box has not exhibited this behavior.

I'm now running a daily cronjob to reset the network! This is not to say
the least embarrasing as this installation is supposed to show off to my
colleagues and management the stability and reliability of Linux running
backend services unattended....

Anybody seen this?
Help!

BTW, I'm running 2.2.2pre4 as 2.2.2 doesn't compile. I've heard that
2.2.2-ac4 does compile on Alpha but I haven't tried it yet...

-- 
squidster

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/