[Xen-users] Why CPU intensive domain gets 60ms time slice?

[Thawan Kooburat <kthawan@xxxxxxxxx>]

Hi,

My understanding is that the credit scheduler should let each domain
runs at most 30 ms at time.

However, when I set domain cap to 50% and pin it to a single CPU then
run CPU intensive job on that domain, I saw that the domain is running
60 ms at a time. The log show that r_time is 30000000, so credit
scheduler should wake up after 30 ms. However the domain continue to
run until 60 ms.

Is this normal or I miss configure something. I am using Xen
4.0.2-rc1-pre on Fedora 13

Below is the scheduler log from xentrace

CPU2  35923150108817 (+133384208)  switch_infprev    [ old_domid =
0x00007fff, runtime = 53493895 ]
CPU2  35923150109319 (+     502)  switch_infnext    [ new_domid =
0x00000001, time = 1189, r_time = 30000000 ]
CPU2  35923150109874 (+     555)  __enter_scheduler [
prev<domid:edomid> = 0x00007fff : 0x00000002, next<domid:edomid> =
0x00000001 : 0x00000000 ]
CPU2  35923150110264 (+     390)  running_to_runnable [ dom:vcpu = 0x7fff0002 ]
CPU2  35923150111082 (+     818)  runnable_to_running [ dom:vcpu = 0x00010000 ]
CPU2  35923224919179 (+74808097)  unknown (0x0000000000021002)  [
0x00010000 0x00000000 0x00000000 0x00000000 0x00000000 0x00000000
0x00000000 ]
CPU2  0 (+       0)  cpu_change        0x00000002
CPU2  35923299725334 (+74806155)  unknown (0x0000000000021002)  [
0x00010000 0x00000000 0x00000000 0x00000000 0x00000000 0x00000000
0x00000000 ]
CPU2  35923299737972 (+   12638)  switch_infprev    [ old_domid =
0x00000001, runtime = 60007693 ]
CPU2  35923299738504 (+     532)  switch_infnext    [ new_domid =
0x00007fff, time = 60007693, r_time = 4294967295 ]




-- 
Thawan Kooburat

Graduate Student
Department of Computer Science
UW-Madison

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