Hi all,
First, here are details of the two Alphas concerned:
Name: news
DU 4.0B
AS2100
4/275 dual processor
256M mem
HSZ40 RAID controller for news spool
FDDI (DEFPA)
Name: leaf
DU 4.0B
AS4000
5/300 single processor
256M mem
SWXCR 200 RAID controller for news spool (everything else spread over
several other disks)
Full duplex switched Fast Ethernet (DE500)
Both runnng inn-1.5.1 and using innfeed 0.10.1.4 (although using a
different version of innfeed makes no difference)
'news' takes numerous external news feeds, and is then responsible
for forarding them on to our internal news machines, of which 'leaf'
is one.
'news' is capable of sustaining an incoming rate of over 550,000
articles a day from its various feeds, but 'leaf' appears to peek at
350,000, which gives us a continuous backlog.
'leaf' will backlog even when it has no readers, and s doing nothing
but accepting incoming news.
We don't think that it is a network problem, as we get a sustainable
ftp transfer rate between the two machines of 25Mb/s.
The root of the problem seems to be that the CPU on 'leaf' has no
idle time (as shown by top), and is spending a considerable amount of
time using system time, almost certainly inside innd.
Anyone got any idea where the bottleneck could be? Is the RAID
controller on 'leaf' sucking up all the CPU time?
----------------------------------
E-Mail: Tony Gale <gale_at_minotaur.dra.hmg.gb>
Date: 17-Jun-97, Time: 10:56:18
Automated Cookie:
I'm still waiting for the advent of the computer science groupie.
Received on Tue Jun 17 1997 - 19:35:12 NZST