2017-07-03 17:28+0800, Yang Zhang:
The background is that we(Alibaba Cloud) do get more and more complaints
from our customers in both KVM and Xen compare to bare-mental.After
investigations, the root cause is known to us: big cost in message passing
workload(David show it in KVM forum 2015)
A typical message workload like below:
vcpu 0 vcpu 1
1. send ipi 2. doing hlt
3. go into idle 4. receive ipi and wake up from hlt
5. write APIC time twice 6. write APIC time twice to
to stop sched timer reprogram sched timer
One write is enough to disable/re-enable the APIC timer -- why does
Linux use two?
7. doing hlt 8. handle task and send ipi to
9. same to 4. 10. same to 3
One transaction will introduce about 12 vmexits(2 hlt and 10 msr write). The
cost of such vmexits will degrades performance severely.
Yeah, sounds like too much ... I understood that there are
IPI from 1 to 2
4 * APIC timer
IPI from 2 to 1
which adds to 6 MSR writes -- what are the other 4?
already provide idle=poll to mitigate the trend. But it only eliminates the
IPI and hlt vmexit. It has nothing to do with start/stop sched timer. A
compromise would be to turn off NOHZ kernel, but it is not the default
config for new distributions. Same for halt-poll in KVM, it only solve the
cost from schedule in/out in host and can not help such workload much.
The purpose of this patch we want to improve current idle=poll mechanism to
Please aim to allow MWAIT instead of idle=poll -- MWAIT doesn't slow
down the sibling hyperthread. MWAIT solves the IPI problem, but doesn't
get rid of the timer one.
use dynamic polling and do poll before touch sched timer. It should not be a
virtualization specific feature but seems bare mental have low cost to
access the MSR. So i want to only enable it in VM. Though the idea below the
patch may not so perfect to fit all conditions, it looks no worse than now.
It adds code to hot-paths (interrupt handlers) while trying to optimize
an idle-path, which is suspicious.
How about we keep current implementation and i integrate the patch to
para-virtualize part as Paolo suggested? We can continue discuss it and i
will continue to refine it if anyone has better suggestions?
I think there is a nicer solution to avoid the expensive timer rewrite:
Linux uses one-shot APIC timers and getting the timer interrupt is about
as expensive as programming the timer, so the guest can keep the timer
armed, but not re-arm it after the expiration if the CPU is idle.
This should also mitigate the problem with short idle periods, but the
optimized window is anywhere between 0 to 1ms.
Do you see disadvantages of this combined with MWAIT?