[Openstack-operators] [scientific-wg] Maximise resource utilisation (was Scientific Working Group IRC meeting, Weds @0700 UTC in #openstack-meeting)

Álvaro López García aloga at ifca.unican.es
Fri Jun 10 11:14:24 UTC 2016


On 09 Jun 2016 (16:17), Blair Bethwaite wrote:
> Hi Tim,
> 
> (...)
> 
> From our discussion in the meeting it sounds as though Blazar could be
> trivially altered to support preemptible instances that could backfill
> Blazar leases. I think you'd have set of specific low-SLA flavors tied to
> the base aggregate that Blazar uses on hosts with no active reservations.
> Blazar would need to be modified to find and delete instances of those
> types when starting a lease. No doubt this is glossing over a lot, but in
> the whole host mode that Chameleon is using it seems quite plausible.
> Wouldn't it be nice if Chameleon nodes were open for elastic compute
> workloads when not reserved?!

This could be achieved without modifying Blazar (I guess), but with a
combination of Blazar + opie (preemptible instances). 

If I understood this correctly, Blazar modifies the aggregates for some
hosts so that they are removed from the "normal" nova scheduling. Then a
user submits a request with a special hint specifying the reservation,
so the request is force submitted to those aggregates. Therefore, the
opie scheduler can fill with preemptible instances the Blazar reserved
hosts and later, whenever a request is be scheduled to a particular
host, it will transparently terminate those preemptible instances
leaving room for the incoming request.

Cheers,
-- 
Álvaro López García                              aloga at ifca.unican.es
Instituto de Física de Cantabria         http://alvarolopez.github.io
Ed. Juan Jordá, Campus UC                      tel: (+34) 942 200 969
Avda. de los Castros s/n                            skype: aloga.csic
39005 Santander (SPAIN)



More information about the OpenStack-operators mailing list