You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@mesos.apache.org by Marc Roos <M....@f1-outsourcing.eu> on 2019/08/29 08:59:45 UTC

Converting vm to task (performance degraded)

I am testing converting a nameserver vm to a task on mesos. If I query 
just one domain (so the results comes from cache) for 30 seconds I can 
do around 450.000 queries on the vm, and only 17.000 on the task. 
When I look at top output on the host where task is running I see this 
task only using 17% cpu time (vm allocates 100% cpu). I have launched 
the task with cpus: 1

How/where/what should I check that causes this reduced performance?  I 
think some configuration is limiting because I can easily get 10k q/s on 
the vm and the task is only getting 1,8k q/s

Is there a configuration guide on how to change a hosts settings to 
optimize it for using with mesos?










Re: Converting vm to task (performance degraded)

Posted by Greg Mann <gr...@d2iq.com>.
Hi Marc,
Could you clarify the hardware configuration of the host? How many cores
does it have? Is this host the same one that you were running the VM on?
How many cores were allocated to the VM?

Regarding the 'top' output, if you're running 'top' with default settings I
would expect to see the nameserver process utilizing 100% of CPU, since
this would represent 1 core, so that does not match my expectations. Did
you supply any flags when running 'top'?

If this were something like a 6- or 8-core machine, and your previous VM
had been allocated all of the cores, then your numbers might make sense,
since you seem to be achieving approximately one sixth of your previous
throughput?

On Thu, Aug 29, 2019 at 5:59 AM Marc Roos <M....@f1-outsourcing.eu> wrote:

>
> I am testing converting a nameserver vm to a task on mesos. If I query
> just one domain (so the results comes from cache) for 30 seconds I can
> do around 450.000 queries on the vm, and only 17.000 on the task.
> When I look at top output on the host where task is running I see this
> task only using 17% cpu time (vm allocates 100% cpu). I have launched
> the task with cpus: 1
>
> How/where/what should I check that causes this reduced performance?  I
> think some configuration is limiting because I can easily get 10k q/s on
> the vm and the task is only getting 1,8k q/s
>
> Is there a configuration guide on how to change a hosts settings to
> optimize it for using with mesos?
>
>
>
>
>
>
>
>
>
>