On Wed, 2003-10-08 at 13:40, Peter Childs wrote: > On Wed, 8 Oct 2003, Anselm Martin Hoffmeister wrote: > > > > > What does the load average in TOP mean? One of our user is trying > > > > to run a HUGE spreadsheet in open office. Running TOP I see that up > > > > to 97% of the CPU usage gets allotted and the CPU "Load Average" is > > > > showing averages of 1, 5 & 15 minutes ranging between 3.8 ~ 1.2 I > > > > want to know what this figure mean? Is this indicating that the > > > > processor is getting loaded beyond the capacity (and hence slow > > > > down??) > > > > > > > > > > Its the average number of processes wanting to use a processor in the > > > last x minutes. I think. So a load average of 1 could mean that 1 > > > processes has been using the processor solidly for the last 1 minute > > > or it could mean that 6 processors have each done 10 seconds > > > processing for the last minute. > > > > > Also processes waiting for disk are counted but not idle ones hence > > > if your io is the bottle neck load will go up and up while your > > > processor is actually doing nothing (or practically nothing) Any load > > > over 50 is bad and probably means you need more processors but is not > > > unlikly if you have 50 people all tring to compile at once on the > > > same computer. > > > > Did you ever try to format an IDE Soft RAID? > > I remember having a load average of 250 ca., half an hour constant load. > > Of course on a not-yet production system. > > (And Softraid was, because of low budget, better choice than no RAID at > > all). > > For sure, / is not on /dev/ataraid :-) > > Are we saying then that the load average is a completly > meaningless number a bit like bogomips.... > Speed is a matter of perception. > > While I'm here I am think of bigging some server(s) in the next > few months. Hmm > > We need to run.... > > LTSP (of course) > Postgres (Large busy database) > Our Software (in house written in Qt) > Jabber > Web > EMail > > Evetually we also need to run a web server... > > Must be 24x7 operation with NO available down time. > > ThinClients are currently planned at ST320 > http://www.optomathinclient.com/ST320.html using LTSP (We managed to find > a local supplier who is happy to reflash the Windows Off) > Office is about 30 desks. I won't say people bacause its all > shifts we have closer to 100 users.... > Any ideas would be most helpful. > > Peter Childs > Peter, I am not even running anything close to what you are planning. About 20 desks (12~15 average online users) running mainly Office applications. See my other post on configuration and the time. Most of the users are happy but this one user who has huge files for work. Problem is in convincing him why it will be worthwhile to shift to LTSP+OpenOffice when on his much smaller P2 system running Win98 the same file opens / closes in less than 20 seconds???!!! And top reports 97% of CPU and 25% of memory allocated to this task!! Other than this other process queues of mail / browsing go on without a hitch. -- Sudev Barar
Learning Linux ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _____________________________________________________________________ Ltsp-discuss mailing list. To un-subscribe, or change prefs, goto: https://lists.sourceforge.net/lists/listinfo/ltsp-discuss For additional LTSP help, try #ltsp channel on irc.freenode.net
