Xen devel - Jul 2010 - summary : xen performance tools

Hi all,

Here''s a summary of available performance tools on xen (please let me
know if I missed any. I''ll put this somewhere on xen wiki later):

1. xenoprof: http://xenoprof.sourceforge.net/
      - Patch oprofile which then must be run on one or more guests.
      - Uses __HYPERVISOR_xenoprof_op hypercall. 
      - src in xen/arch/x86/oprofile

2. xenperf: 
     - Reports software ctrs in xen. Must compile xen with perfc=y
        (#ifdef PERF_COUNTERS). 
     - Source: tools/misc/xenperf.c
     - Reports global counters for the system. Counters must be reset
       manually before each run.

3. xenmon: 
    - Source: tools/xenmon/*
    - run xenmon.py to see gotten/blocked/waited cpu usage, one cpu at a time.
    - Uses raw trace buffer data .
    - Not all options work.
    - Needs work for more than 1 vcpu.

4. xentop:
    - works well.
    - src in tools/xenstat/xentop

5. xenlockprof: 
    - src in tools/misc
    - compile xen lock_profile=y (ifdef LOCK_PROFILE)

6. xentrace:
    - good performance or debugging tool. 
    - based on trace hooks in hypervisor code.


Would be nice to create subdir perf in tools and put xenperf, xenmon, 
and xenlockprof in it so folks can quickly find perf tools. Also, more tools 
can be added and old/un-maintained removed easily then.

I am trying to generic''ise xen_sysctl_perfc_op so that it can be used
to add more tools via union {} easily. Please let me know
if that would be a problem. Unfortunately, oprof hypercall can''t be
generic''ised and adding a new generic perf hcall will require dom0
recompile.

thanks,
Mukesh


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Mon, 12 Jul 2010 10:48:36 +0800
<ben_ian@trend.com.tw> wrote:
> Hi Mukesh,
> I had a different result on dom0 memory allocation report by xentop
> and top. The difference is usually around 7XXMB, and xentop is always
> reported fewer.
> 
> ------------------------------------------
> For example,  (xentop vs top => 1177508k vs 1960880k)
> 
> xentop - 12:24:17   Xen 4.0.0
> 5 domains: 1 running, 4 blocked, 0 paused, 0 crashed, 0 dying, 0
> shutdown Mem: 41879780k total, 41845196k used, 34584k free    CPUs: 8
> @ 2533MHz NAME  STATE   CPU(sec) CPU(%)     MEM(k) MEM(%)  MAXMEM(k)
> MAXMEM(%) VCPUS NETS NETTX(k) NETRX(k) VBDS VBD_OO   VBD_RD   VBD_WR
> VBD_RSECT  VBD_WSECT SSID centOS --b---         27    1.6
> 10489824   25.0   10489856      25.0     1    0        0        0
> 1 0        0        0          0          0    0 centOS4
> --b---          8    1.5    8392672   20.0    8392704      20.0
> 1    1        0        0    1 0        0        0          0
> 0    0 Domain-0 -----r        184   11.5    1177508    2.8   no
> limit       n/a     8    0        0        0    0 0        0
> 0          0          0    0 win2003 --b---          8    0.2
> 10489824   25.0   10489856      25.0     1    1        0        0
> 1 0        0        0          0          0    0 winxp
> --b---          7    0.2   10489824   25.0   10489856      25.0
> 1    1        0        0    1 0        0        0          0
> 0    0
> 
>  ---	 
> top - 12:24:48 up 16 min,  5 users,  load average: 0.52, 0.38, 0.22
> Tasks: 226 total,   3 running, 223 sleeping,   0 stopped,   0 zombie
> Cpu(s):  0.5%us,  0.8%sy,  0.0%ni, 96.4%id,  2.2%wa,  0.0%hi,
> 0.0%si,  0.1%st Mem:   1960880k total,  1922052k used,    38828k
> free,     6528k buffers Swap: 51199992k total,        0k used,
> 51199992k free,   272924k cached
> 
> ---------------------------------------------------------------------------
> 
> 
> 
> I collect this result on Fedora12+kernel 2.6.32+xen4.0.0, and this
> will cause a kernel crash on my system when I try to enable
> ballooning and balloon one more 768MB dom-U under the upper
example''s
> status(dom0_men=1024)
> 
> Do you have any idea what might cause this result??
> PS: Under kernel 2.6.18+ xen3.4.2, the two values are the same.
> 
>  
> Thanks very much!
> Regards,
> Ben Ian
> 
Hi Ben,

Sorry, I''ve not dug into the details of the tools to see
what''s going
on. If I happen to figure it out, I''ll let you know.

thanks
Mukesh


_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel