Re: [PATCH RFC ticketlock] Auto-queued ticketlock

From: Paul E. McKenney
Date: Tue Jun 11 2013 - 14:05:56 EST


On Tue, Jun 11, 2013 at 10:53:06AM -0700, Davidlohr Bueso wrote:
> On Mon, 2013-06-10 at 17:51 -0700, Linus Torvalds wrote:
> > On Mon, Jun 10, 2013 at 5:44 PM, Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> > >
> > > OK, I haven't found a issue here yet, but youss are beiing trickssy! We
> > > don't like trickssy, and we must find precccciouss!!!
> >
> > .. and I personally have my usual reservations. I absolutely hate
> > papering over scalability issues, and historically whenever people
> > have ever thought that we want complex spinlocks, the problem has
> > always been that the locking sucks.
> >
> > So reinforced by previous events, I really feel that code that needs
> > this kind of spinlock is broken and needs to be fixed, rather than
> > actually introduce tricky spinlocks.
> >
> > So in order to merge something like this, I want (a) numbers for real
> > loads and (b) explanations for why the spinlock users cannot be fixed.
>
> I hate to be the bearer of bad news but I got some pretty bad aim7
> performance numbers with this patch on an 8-socket (80 core) 256 Gb
> memory DL980 box against a vanilla 3.10-rc4 kernel:

Looks pretty ugly, sorry that it doesn't help in many of your situations.

Any info on what bottlenecks you are encountering?

Thanx, Paul

> * shared workload:
> 10-100 users is in the noise area.
> 100-2000 users: -13% throughput.
>
> * high_systime workload:
> 10-700 users is in the noise area.
> 700-2000 users: -55% throughput.
>
> * disk:
> 10-100 users -57% throughput.
> 100-1000 users: -25% throughput
> 1000-2000 users: +8% throughput (this patch only benefits when we have a
> lot of concurrency).
>
> * custom:
> 10-100 users: -33% throughput.
> 100-2000 users: -46% throughput.
>
> * alltests:
> 10-1000 users is in the noise area.
> 1000-2000 users: -10% throughput.
>
> One notable exception is the short workload where we actually see
> positive numbers:
> 10-100 users: +40% throughput.
> 100-2000 users: +69% throughput.
>
> Thanks,
> Davidlohr
>

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/