Messages in this thread | | | Date | Wed, 11 Jan 2012 09:11:23 -0500 | From | "John Stoffel" <> | Subject | Re: [PATCH -mm] make swapin readahead skip over holes |
| |
>>>>> "KOSAKI" == KOSAKI Motohiro <kosaki.motohiro@gmail.com> writes:
>> Also, I doubt current swap_cluster default is best value on nowadays. KOSAKI> I meant, current average hdd spec is, KOSAKI> - average seek time: 8.5ms KOSAKI> - sequential access performance: about 60MB/sec
KOSAKI> so, we can eat free lunch up to 7MB ~= 60(MB/sec) * 1000 / 8.5(ms).
What if the disk is busy doing other writeout or readin during this time? You can't assume you have the full disk bandwidth available, esp when you hit a swap storm like this.
John
| |