Re: [patch] mm, slab: faster active and free stats

[Joonsoo Kim]

On Wed, Nov 09, 2016 at 04:38:08PM -0800, David Rientjes wrote:
> On Tue, 8 Nov 2016, Andrew Morton wrote:
> 
> > > Reading /proc/slabinfo or monitoring slabtop(1) can become very expensive
> > > if there are many slab caches and if there are very lengthy per-node
> > > partial and/or free lists.
> > > 
> > > Commit 07a63c41fa1f ("mm/slab: improve performance of gathering slabinfo
> > > stats") addressed the per-node full lists which showed a significant
> > > improvement when no objects were freed.  This patch has the same
> > > motivation and optimizes the remainder of the usecases where there are
> > > very lengthy partial and free lists.
> > > 
> > > This patch maintains per-node active_slabs (full and partial) and
> > > free_slabs rather than iterating the lists at runtime when reading
> > > /proc/slabinfo.
> > 
> > Are there any nice numbers you can share?
> > 
> 
> Yes, please add this to the description:
> 
> 
> When allocating 100GB of slab from a test cache where every slab page is
> on the partial list, reading /proc/slabinfo (includes all other slab
> caches on the system) takes ~247ms on average with 48 samples.
> 
> As a result of this patch, the same read takes ~0.856ms on average.

Hello, David.

Maintaining acitve/free_slab counters looks so complex. And, I think
that we don't need to maintain these counters for faster slabinfo.
Key point is to remove iterating n->slabs_partial list.

We can calculate active slab/object by following equation as you did in
this patch.

active_slab(n) = n->num_slab - the number of free_slab
active_object(n) = n->num_slab * cachep->num - n->free_objects

To get the number of free_slab, we need to iterate n->slabs_free list
but I guess it would be small enough.

If you don't like to iterate n->slabs_free list in slabinfo, just
maintaining the number of slabs_free would be enough.

Thanks.