Re: Ram use



On Sat, 2009-06-20 at 16:13 -0600, Karl Larsen wrote:
I was curious to determine if Kubuntu uses more RAM than Gnome.
Since I have both on my laptop in version 8.10 I decided to try to
measure this. I used top to get the data and I saved the part needed to
gedit which you can get on both kubuntu and gnome. I took data at 10
minutes after boot-up. Here is the data:

You need to be careful about how you interpret these kind of results,
especially on a demand paged virtual memory system like Linux. The more
critical measure for responsiveness (which I think is the critical
measure for a desktop system)is likely the paging rate and especially
the pageout rate. Paging is very expensive when compared to memory
access time. vmstat is a crude, but useful tool to measure this sort of
metric. If you run:

vmstat 5

watch the si and so columns, especially the so column which is the page
out rate to swap. That is a measure of pages being written to the swap
device(s). iostat which is part of the sysstat package can also be
useful.

A more useful test would be to run vmstat 5 or iostat and then operate
the desktop in a typical fashion and see what the paging rate looks like
over time. If you have sufficient memory the page out rate should be
very low. A high paging rate indicates memory starvation and the system
will start to thrash. In vmstat you will see processes blocked waiting
for I/O.

With all this said, I do think running Gnome or KDE with 1GB of memory
is a tad tight with a typical desktop mix of applications. But memory
is very cheap these days and adding memory can be a lot simpler and more
cost effective than trying to optimally tune up the virtual memory
system.

I run a Linux server with 1GB of RAM. It is interesting to watch the
memory utilization over time. The graphs are here:

http://www.c-m.us.com/~smoot/RRD

What you see is the buffer and cache utilization go up significantly
when the system does a lot of I/O. The spikes in the graphs corresponds
to when I run an rsync cron job to do nearline backups. The "free"
memory pool stabilizes at a low percentage over time.
--
Smoot Carl-Mitchell
Computer Systems and
Network Consultant
smoot@xxxxxxx
+1 480 922 7313
cell: +1 602 421 9005

--
ubuntu-users mailing list
ubuntu-users@xxxxxxxxxxxxxxxx
Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/ubuntu-users



Relevant Pages

  • Re: maximum .net process can use
    ... > No, it won't help you at all, unless you have a high paging rate. ... > | can use only up to 2G max of memory, but other items including the OS ... the system will OOM at allocation time when it's not able to ... > First thing to do is increase your page file size, and watch ...
    (microsoft.public.dotnet.languages.csharp)
  • Re: maximum .net process can use
    ... > No, it won't help you at all, unless you have a high paging rate. ... > | can use only up to 2G max of memory, but other items including the OS ... Your paging file is too small anyway, ... > First thing to do is increase your page file size, and watch ...
    (microsoft.public.dotnet.languages.csharp)
  • Re: maximum .net process can use
    ... The question YOU have to answer is - how large is this chunk of memory you ... are trying to allocate, note I say "chunk" of memory because I don't know ... |> No, it won't help you at all, unless you have a high paging rate. ... Your paging file is too small anyway, ...
    (microsoft.public.dotnet.languages.csharp)
  • Re: maximum .net process can use
    ... As I said before a 32 bit windows app. ... probably need more than 2GB of RAM or you have to accept the possible paging ... |> |I used taskmanager that came with window to look at the memory usage ... |> No, it won't help you at all, unless you have a high paging rate. ...
    (microsoft.public.dotnet.languages.csharp)
  • [RFC] page replacement requirements
    ... Submitting too much I/O at once can kill latency and even lead to deadlocks when bounce buffers are involved. ... Must be able to deal with multiple memory zones efficiently. ... When on completion of the write to their backing-store the reference bit is still unset a callback is invoked to place them so that they are immediate candidates for reclaim again. ... For traditional page replacement algorithms this is not a big issue since we just implement per zone page replacement; ...
    (Linux-Kernel)

Loading