Conquering NUMA land

[Daniel Waddington]

Hi Norman,

I agree with Udo's comment about needing to be more flexible about 
resource partitioning across v-cores.  Of course you might use 
configuration scripts to help configure v-cores initially, but it is 
important to be able to dynamically adjust partitions at run-time 
according to workload needs.  You should strive to separate the concerns 
of enabling NUMA and the policies behind NUMA.

I think that you can only evaluate the success of your NUMA capabilities 
with real applications.  Hopefully we can help you there.

Daniel

On 03/18/2013 04:40 AM, Udo Steinberg wrote:
> On Mon, 18 Mar 2013 11:49:40 +0100 Norman Feske (NF) wrote:
>
> [Details snipped]
>
> NF> What do you think? Would the vcore idea be worthwhile to explore? Those
> NF> of you experienced in the field of manycore NUMA systems, do you see
> NF> additional pitfalls? Or even better, does anyone has alternative ideas
> NF> to explore? Also, I am very interested in ways to validate work in this
> NF> domain. How can we measure our success?
>
> There are also use cases where you don't want to partition. One example is a
> multi-core VM, where each virtual CPU could run on a different physical core
> and yet all of those virtual CPUs share the same memory.
>
> Rather than going for an extreme design point, where virtually nothing is
> shared (e.g., Barrelfish), I think it would be better to provide an
> interface where the user has precise control over what is shared and what
> isn't.
>
> I'd go for concurrent invocation of services first. Then you'll know what
> data structures you have contention on. And then you can decide whether you
> want that data replicated (read-mostly) or shared (frequently written).
>
> IMHO, dealing with replicas, distributed protocols, consensus and all that
> is a lot harder than implementing a few locks or atomic ops on pieces of
> shared memory. Especially now that we have HLE and TSX coming really soon.
>
> Cheers,
> Udo
>
>
> ------------------------------------------------------------------------------
> Everyone hates slow websites. So do we.
> Make your web apps faster with AppDynamics
> Download AppDynamics Lite for free today:
> http://p.sf.net/sfu/appdyn_d2d_mar
>
>
> _______________________________________________
> Genode-main mailing list
> Genode-main at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/genode-main

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.genode.org/pipermail/users/attachments/20130318/77d7bca9/attachment.html>