Messages in this thread | | | Date | Thu, 06 Aug 2009 16:46:57 +0300 | From | Avi Kivity <> | Subject | Re: [RFC] respect the referenced bit of KVM guest pages? |
| |
On 08/06/2009 04:06 PM, Wu Fengguang wrote: > On Thu, Aug 06, 2009 at 07:44:01PM +0800, Avi Kivity wrote: > >> On 08/06/2009 01:59 PM, Wu Fengguang wrote: >> > > scheme KEEP_MOST: > > >>> How about, for every N pages that you scan, evict at least 1 page, >>> regardless of young bit status? That limits overscanning to a N:1 >>> ratio. With N=250 we'll spend at most 25 usec in order to locate one >>> page to evict. >>> > > scheme DROP_CONTINUOUS: > > >>> This is a quick hack to materialize the idea. It remembers roughly >>> the last 32*SWAP_CLUSTER_MAX=1024 active (mapped) pages scanned, >>> and if _all of them_ are referenced, then the referenced bit is >>> probably meaningless and should not be taken seriously. >>> > >
Or one scheme, with N=parameter.
>> I don't think we should ignore the referenced bit. There could still be >> a large batch of unreferenced pages later on that we should >> preferentially swap. If we swap at least 1 page for every 250 scanned, >> after 4K swaps we will have traversed 1M pages, enough to find them. >> > > I guess both schemes have unacceptable flaws. > > For JVM/BIGMEM workload, most pages would be found referenced _all the time_. > So the KEEP_MOST scheme could increase reclaim overheads by N=250 times; > while the DROP_CONTINUOUS scheme is effectively zero cost. >
Maybe 250 is an exaggeration. But even with 250, the cost is still pretty low compared to the cpu cost of evicting a page (with IPIs and tlb flushes).
-- error compiling committee.c: too many arguments to function
| |