| 1 |
On 03/01/2014 23:52, Stefan G. Weichinger wrote: |
| 2 |
> Am 03.01.2014 15:07, schrieb Alan McKinnon: |
| 3 |
> |
| 4 |
>> I'd like to see the results of any benchmarks you do with BFQ on VMs |
| 5 |
> |
| 6 |
> If we run BFQ on the host (as it runs SSDs and HDDs and we want it to be |
| 7 |
> snappy) this would maybe mean that in the VMs we need Noop ? |
| 8 |
> |
| 9 |
> In the IBM-pdf "Best practices for KVM" they tell me to |
| 10 |
> |
| 11 |
> "use the Deadline I/O scheduler on the KVM host and guest operating |
| 12 |
> systems for I/O-bound workloads on enterprise storage systems." |
| 13 |
> |
| 14 |
> But they don't mention BFQ at all. |
| 15 |
|
| 16 |
|
| 17 |
that's expected, BFQ isn't in mainline |
| 18 |
|
| 19 |
|
| 20 |
> |
| 21 |
> I can test here on my systems as soon as I find the time and the brains |
| 22 |
> ... or on an upcoming server I will setup within the next weeks for a |
| 23 |
> customer. |
| 24 |
> |
| 25 |
> Stefan |
| 26 |
> |
| 27 |
> |
| 28 |
> |
| 29 |
|
| 30 |
|
| 31 |
-- |
| 32 |
Alan McKinnon |
| 33 |
alan.mckinnon@×××××.com |
Archives