Open Stack

hmmm....thanks a lot...i'll try reading more papers for the time being on
load balancing and benchmarking


On Thu, May 2, 2013 at 9:41 PM, Clay Gerrard <clay.gerrard at gmail.com> wrote:

> Hrmm.... Sounds more like an estimator.
>
> There's lots of things that can effect different types of performance
> besides number of nodes - ring params, storage tiering, hardware, drives,
> load balancing, ssl termination, network.  Depends on your use case and
> what you're trying to optimize for.  Unfortunately a lot of learned lessons
> and findings here are still gut feel.  We were lucky to have some deployers
> & operators at the OpenStack design summit that willing to talk about their
> setups, plus lots of folks working on benchmarking Swift - so we can start
> to get real world data and some numbers behind all this.
>
> It's possible that the combination of open discussion around deployment
> topologies and benchmarking might one day lead to broader confidence in
> predicting how "adding servers to a swift cluster" might
> "improve performance [..] so that the user can known in advance at least to
> an approximate value what this improvement in performance might be before
> actually *using* any swift resources" [emphasis added]
>
> An example of another project that sorta does what you're describing here
> might be really helpful since I'm still not quite sure I understand fully
> what you're describing - but I'm pretty sure nothing like this exists for
> Swift.
>
> OTOH, If you ask some experienced swift deployers and operators (many of
> whom hang out on this list or #openstack-swift on Freenode) you're likely
> to find someone willing to chat about the *specific details* of your setup.
>  We might even be able to get some back of the napkin math behind it...
>
> GL,
>
> -Clay
>
>
> On Thu, May 2, 2013 at 2:49 AM, gagan bopanna <bops.91 at gmail.com> wrote:
>
>> @clay,
>> thanks for the reply....before actually adding servers to a swift cluster
>> to improve performance i want to try and simulate the same so that the user
>> can know in advance atleast to an approximate value what this improvement
>> in performance might be before actually using any swift resources...i am
>> still not sure if this is possible but i hope u got the general idea
>>
>>
>> On Wed, May 1, 2013 at 8:32 PM, Clay Gerrard <clay.gerrard at gmail.com>wrote:
>>
>>> simulator... for the api?  Or do you mean like a load simulator?  Is
>>> there another simulator for maybe a different system that would
>>> be comparable to what you are trying to create for swift?
>>>
>>> "what kind of information would be useful before you actually start
>>> using swift to store data"
>>>
>>> Are you going store user generated content and serve it back on the web?
>>>
>>> Yes
>>>
>>> Use swift ;)
>>>
>>> -Clay
>>>
>>>
>>> On Wed, May 1, 2013 at 7:25 AM, gagan bopanna <bops.91 at gmail.com> wrote:
>>>
>>>> Hey guys,
>>>> I have been trying to create a simulator for swift...so if there is
>>>> anyone who has been using swift for a long time can you please suggest what
>>>> kind of information would be useful from the simulation...that is what kind
>>>> of information would be useful before you actually start using swift to
>>>> store data.
>>>> -thanks in advance
>>>>
>>>> _______________________________________________
>>>> OpenStack-dev mailing list
>>>> OpenStack-dev at lists.openstack.org
>>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>>
>>>>
>>>
>>> _______________________________________________
>>> OpenStack-dev mailing list
>>> OpenStack-dev at lists.openstack.org
>>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>>
>>>
>>
>> _______________________________________________
>> OpenStack-dev mailing list
>> OpenStack-dev at lists.openstack.org
>> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>>
>>
>
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstack.org/pipermail/openstack-dev/attachments/20130503/ef6fe850/attachment.html>