Re: Improving latency and ordering of the backfilling workload

[Loic Dachary <loic@dachary.org>]

[-- Attachment #1: Type: text/plain, Size: 4648 bytes --]



On 15/12/2014 18:20, Sage Weil wrote:
> On Mon, 15 Dec 2014, Loic Dachary wrote:
>> Hi Sage,
>>
>> On 15/12/2014 17:44, Sage Weil wrote:
>>> On Mon, 15 Dec 2014, Loic Dachary wrote:
>>>> Hi Sam,
>>>>
>>>> Here is what could be done (in the context of http://tracker.ceph.com/issues/9566
>>>> ), please let me know if that makes sense:
>>>>
>>>> * ordering:
>>>>
>>>>   * when dequeuing a pending local reservation, chose one that contains 
>>>> a PG that belongs to the busiest OSD (i.e. the OSD for which there are 
>>>> more PGs waiting for a local reservation than any other)
>>>
>>> I'm worried the reservation count won't be an accurate enough proxy for 
>>> the amount of work the remote OSD has to do.  
>>
>> Are you thinking about taking into account the number and size of 
>> objects in a given PGs ? The length of the local reservation queue 
>> accurately reflects the number of PGs that need work (because the length 
>> of the reservation queue is not bounded). But it does not reflect the 
>> content of the PGs at all, indeed.
> 
> Including that information could help, yeah, but the main thing is that 
> any estimate of "the busiest OSD" based on local information is going to 
> be weak if it's only based on info reservation requests.  

What other information would be relevant in addition to the number of PGs that need to backfill and their size (objects & bytes) ?

> Unless that 
> information is refreshed periodically by the requesting OSD (I think we 
> also discussed that a bit last week).

I tried to take that into account by proposing to calculate the priority when the reservation is dequeued from the waiting list instead of when it is added to the waiting list. When the local reservation is dequeued, it gets one of the osd_max_backfill slots in the AsyncReserver and will then get work to do : the delay between calculating the priority and actual backfilling is minimum. The delay actually is the latency between when the remote reservation is sent and when it comes back successfully. By adding the priority to the remote reservation request, we make the peer OSD aware of the local priority and compare it with the priority of the other OSDs asking for a remote reservation. The peer OSD will be grant us a remote reservation quickly if we are the OSD declaring to have most work to do.

I sense you have something else in mind in terms of algorithm and/or data sources. Hopefully this explanation will allow you to see what I'm missing and guide me ;-)

> 
>> It would be very easy to 
>>> piggyback some load information on the heartbeat messages which we should 
>>> already be exchanging with anyone we would backfill with.
>>>
>>> If we go down that path, there are a bunch of patches in the wip-read-hole 
>>> series that lay useful groundwork.  Getting that branch into shape 
>>> is the next big item after I finish the current batch of pull 
>>> requests.
>>
>> Would you mind telling me which of 
>> https://github.com/ceph/ceph/commits/wip-read-hole commits are relevant 
>> ? I assume 
>> https://github.com/ceph/ceph/commit/ee72f699e236371a5b8651cd900013a2bd2227fb 
>> is to some extent.
> 
> Yeah that's the one.  There's a later patch that give each PG a handy 
> reference to that struct for the acting set (for quick access), though in 
> this case not all backfill peers will be in acting.
> 
> Note that there is also a osd_peer_stat_t struct in MOSDPing that is 
> currently unused cruft.  We could replace/supplement that with whatever 
> information we thing would be helpful.
> 
> If we go down that path at least.. I think ahve reservers refresh their 
> reservation periodically with updated priorities would also work.
> 
> sage
> 
> 
>>
>> Cheers
>>
>>>>   * when sending a remote reservation request, set the priority to 
>>>> reflect the total number of pending PG (absolute workload) and the 
>>>> number local pending PG for the destination OSD (workload queued locally 
>>>> for the remote OSD)
>>>>   * on the receiving side, the priority of the remote reservation 
>>>> request makes sure the busiest OSD gets a remote reservation before the 
>>>> others
>>>>
>>>> * reducing latency:
>>>>   
>>>>   * if there are N pending remote reservations, reject a remote 
>>>> reservation request instead of queuing it so that the local reservation 
>>>> can be used instead of waiting.
>>>>
>>>> Cheers
>>>>
>>>> -- 
>>>> Lo?c Dachary, Artisan Logiciel Libre
>>>>
>>>>
>>>
>>
>> -- 
>> Lo?c Dachary, Artisan Logiciel Libre
>>
>>

-- 
Loïc Dachary, Artisan Logiciel Libre


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 198 bytes --]