Re: [PATCH v2] mm: scale kswapd watermarks in proportion to memory

2016-02-25 Thread Johannes Weiner
Hi Joonsoo, On Thu, Feb 25, 2016 at 09:37:44AM +0900, Joonsoo Kim wrote: > On Mon, Feb 22, 2016 at 03:33:22PM -0800, Johannes Weiner wrote: > > In machines with 140G of memory and enterprise flash storage, we have > > seen read and write bursts routinely exceed the kswapd watermarks and > > cause

Re: [PATCH v2] mm: scale kswapd watermarks in proportion to memory

2016-02-24 Thread Joonsoo Kim
Hello, Johannes. Just nitpick below. On Mon, Feb 22, 2016 at 03:33:22PM -0800, Johannes Weiner wrote: > In machines with 140G of memory and enterprise flash storage, we have > seen read and write bursts routinely exceed the kswapd watermarks and > cause thundering herds in direct reclaim. Unfortu

Re: [PATCH v2] mm: scale kswapd watermarks in proportion to memory

2016-02-23 Thread David Rientjes
On Mon, 22 Feb 2016, Johannes Weiner wrote: > In machines with 140G of memory and enterprise flash storage, we have > seen read and write bursts routinely exceed the kswapd watermarks and > cause thundering herds in direct reclaim. Unfortunately, the only way > to tune kswapd aggressiveness is thr

Re: [PATCH v2] mm: scale kswapd watermarks in proportion to memory

2016-02-23 Thread Johannes Weiner
On Mon, Feb 22, 2016 at 06:23:19PM -0800, David Rientjes wrote: > On Mon, 22 Feb 2016, Johannes Weiner wrote: > > > In machines with 140G of memory and enterprise flash storage, we have > > seen read and write bursts routinely exceed the kswapd watermarks and > > cause thundering herds in direct r

Re: [PATCH v2] mm: scale kswapd watermarks in proportion to memory

2016-02-22 Thread David Rientjes
On Mon, 22 Feb 2016, Johannes Weiner wrote: > In machines with 140G of memory and enterprise flash storage, we have > seen read and write bursts routinely exceed the kswapd watermarks and > cause thundering herds in direct reclaim. Unfortunately, the only way > to tune kswapd aggressiveness is thr

Re: [PATCH v2] mm: scale kswapd watermarks in proportion to memory

2016-02-22 Thread Johannes Weiner
On Mon, Feb 22, 2016 at 08:36:33PM -0500, Rik van Riel wrote: > On Mon, 2016-02-22 at 15:33 -0800, Johannes Weiner wrote: > > > Beyond 1G of memory, this will produce bigger watermark steps than > > Is that supposed to be beyond 16GB? The old formula formula is min = sqrt(kb * 16) low = mi

Re: [PATCH v2] mm: scale kswapd watermarks in proportion to memory

2016-02-22 Thread Rik van Riel
On Mon, 2016-02-22 at 15:33 -0800, Johannes Weiner wrote: > Beyond 1G of memory, this will produce bigger watermark steps than Is that supposed to be beyond 16GB? > the > current formula in default settings. Ensure that the new formula > never > chooses steps smaller than that, i.e. 25% of the

[PATCH v2] mm: scale kswapd watermarks in proportion to memory

2016-02-22 Thread Johannes Weiner
In machines with 140G of memory and enterprise flash storage, we have seen read and write bursts routinely exceed the kswapd watermarks and cause thundering herds in direct reclaim. Unfortunately, the only way to tune kswapd aggressiveness is through adjusting min_free_kbytes - the system's emergen