www-infrastructure-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "seba.wagner@gmail.com" <seba.wag...@gmail.com>
Subject Re: A server for Openmeetings
Date Tue, 24 Sep 2013 09:46:10 GMT
Hi Tony,

this sounds all very reasonable.

The bottleneck in our calculations are often bandwidth requirements.
Memory/disk usage does pretty much scale linear with number of users, but
bandwidth does more like exponential with the number of concurrent sessions.

However I would not be too concerned about the bandwidth, this server is
basically for functional verification, not for load tests.
On an average daily peak there might be maybe 10-20, on a quartly year peak
there might be up to 50 people online at the same time.

One of the goals would be for instance to run automated Selenium tests
against the server and have a daily builder scripts that installes nightly
snapshots of the application.

So this should be sufficient for our needs.

My suggestion would be that we use the LDAP connector of OpenMeetings to
integration with the Apache Ldap, so that only PMC can login. This is
something we can set up once we have the server, however we might need some
pointers from Infra to get the right urls/configs in place.

Thanks!
Sebastian



2013/9/24 Tony Stevenson <tony@pc-tony.com>

> Ultimately we will need a spec for the VM yes.  However it is infra who
> will approve the spec in the first instance.  The only real restrictions
> are:
>
> 1 - available resources on the physical host
> 2 - subject to fair use of the resources
> 3 - we oblige any local restrictions (providers etc)
>
> So in my head I woudl say something like :
>
> 2-4 vcpus
> 8 GB RAM max (needs some level of justification)
> 100GB disk
> and reasonable use of bandwidth
>
> I know these might be inappropriate - but these are nothing more than
> starting points for the conversation to start.
>
> Thoughts?
>
>
> On 24/09/13 08:49, seba.wagner@gmail.com wrote:
>
>> I think they are waiting for a spec.
>> However I am still a bit confused who needs to approve that spec?
>> Do we need the ASF board to approve that spec?
>>
>> Thanks
>> Sebastian
>>
>>
>> 2013/9/24 Maxim Solodovnik <solomax666@gmail.com>
>>
>>  Dear Tony,
>>>
>>> Are the any updates on this topic?
>>>
>>>
>>> On Fri, Jun 7, 2013 at 1:23 PM, Tony Stevenson <tony@pc-tony.com> wrote:
>>>
>>>  We need to get a vmware host specced up, agreed, purchased, delivered,
>>>> racked, installed, configured, a VM created etc,
>>>> Once we know this is what we want to do we will need to allow ~1 month
>>>> turn-around time.  Maybe 6 weeks.
>>>>
>>>> Primarily the delay is in the order, racking, and installing stages.
>>>>  But
>>>> also bear in mind our new EU colo has very few services in it.
>>>>
>>>>
>>>>
>>>>
>>>> On 7 Jun 2013, at 07:19, Alexei Fedotov <alexei.fedotov@gmail.com>
>>>>
>>> wrote:
>>>
>>>>
>>>>  Artyom,
>>>>>
>>>>> Tony said,
>>>>>
>>>>>> actually someone just reminded me that our new EU colo provider
>>>>>>
>>>>> has enough bandwidth to cope with your request - however it is not
>>>>>
>>>> QoSed
>>>
>>>> so it cannot be guaranteed - but giving the amount they made available
>>>>> to us I dont think you would make a significant dent in that.
>>>>>
>>>>> Tony, what would be our next step to set up the server at EU colo
>>>>>
>>>> provider?
>>>>
>>>>>
>>>>> Thanks!
>>>>>
>>>>>
>>>>> --
>>>>> With best regards / с наилучшими пожеланиями,
>>>>> Alexei Fedotov / Алексей Федотов,
>>>>> http://dataved.ru/
>>>>> +7 916 562 8095
>>>>>
>>>>>
>>>>> On Mon, Apr 22, 2013 at 10:19 PM, Alexei Fedotov
>>>>> <alexei.fedotov@gmail.com> wrote:
>>>>>
>>>>>> Tony, as I have said, the requirements are not as important as the
>>>>>> fact we have the server running. For those who feel that bandwidth
is
>>>>>> not enough or is not reliable enough there is always an option to
>>>>>> switch to other test servers (or install their own server). We may
add
>>>>>> a warning message at the welcome screen.
>>>>>>
>>>>>> As for your previous question, my AWS bills for two test servers
are
>>>>>> ~$30 per month though micro-instance is kinda slow, and hetzner bills
>>>>>> are approx. ~$30 per server. For 100 Mbit/s connected server located
>>>>>> in Russia (with higher hosting costs) I pay 250$ / month.
>>>>>> --
>>>>>> With best regards / с наилучшими пожеланиями,
>>>>>> Alexei Fedotov / Алексей Федотов,
>>>>>> http://dataved.ru/
>>>>>> +7 916 562 8095
>>>>>>
>>>>>>
>>>>>> On Mon, Apr 22, 2013 at 10:08 PM, Tony Stevenson <pctony@apache.org>
>>>>>>
>>>>> wrote:
>>>>
>>>>> Tony Stevenson wrote on Mon, Apr 22, 2013 at 06:46:55PM +0100:
>>>>>>>
>>>>>>>> Alexei Fedotov wrote on Mon, Apr 22, 2013 at 09:32:12PM +0400:
>>>>>>>>
>>>>>>>>> Tony, I'm sorry. This was discussed for sure. Things
sometimes
>>>>>>>>>
>>>>>>>> change over
>>>>
>>>>>  time, that's why I have asked once again.
>>>>>>>>>
>>>>>>>>> We have have several test servers and chat channels hosted
around.
>>>>>>>>>
>>>>>>>> My
>>>
>>>>  intention was to have more
>>>>>>>>> development resources hosted @apache and branded by apache.
>>>>>>>>>
>>>>>>>>
>>>>>>>> Generally that is a good thing.  However your requirements
are
>>>>>>>> significant, and this is not something we could in all good
faith
>>>>>>>>
>>>>>>> commit
>>>>
>>>>>  too. Sorry.
>>>>>>>>
>>>>>>>
>>>>>>> Alexi, actually someone just reminded me that our new EU colo
>>>>>>>
>>>>>> provider
>>>
>>>>  has enough bandwidth to cope with your request - however it is not
>>>>>>>
>>>>>> QoSed
>>>>
>>>>> so it cannot be guaranteed - but giving the amount they made
>>>>>>>
>>>>>> available
>>>
>>>>  to us I dont think you would make a significant dent in that.
>>>>>>>
>>>>>>> The only issue now is the lack of hardware in this location to
'make
>>>>>>>
>>>>>> it
>>>
>>>>  so'. We could look at purchasing a new box for the site, but we
>>>>>>>
>>>>>> wouldnt
>>>
>>>>  be able to guarantee you the 25% of the server you wanted, and it
>>>>>>>
>>>>>> would
>>>
>>>>  be a virtual machine that we'd give you.
>>>>>>>
>>>>>>> The colo would be in Europe, specifically in Germany.  Would
this be
>>>>>>>
>>>>>> an
>>>
>>>>  issue for you?
>>>>>>>
>>>>>>> Please let us know, and I *might* be able to carry this forward.
>>>>>>>
>>>>>>>
>>>>>>>> Please do not mis-read my reply - I was trying to be open
and
>>>>>>>>
>>>>>>> honest.
>>>
>>>>  While over time we might well be able to offer you all the hardware
>>>>>>>>
>>>>>>> you
>>>>
>>>>>  need, it is so very extremely unlikely we could meet your bandwidth
>>>>>>>> needs. All our bandwidth is donated to us for free, and as
such it
>>>>>>>> doesnt come in huge quantities; and where we do have it we
have to
>>>>>>>> prioritise core services over that of development instances
(i.e.
>>>>>>>>
>>>>>>> svn
>>>
>>>>  and http over that of pmc specific needs).  It's not that we dont
>>>>>>>>
>>>>>>> want
>>>
>>>>  to help, but we only have finite resources, and we need to be
>>>>>>>>
>>>>>>> careful
>>>
>>>>  not to over commit them.  Bandwith is one such commodity where we
>>>>>>>>
>>>>>>> need
>>>
>>>>  to be careful.
>>>>>>>>
>>>>>>>> One option that springs to mind is the use of AWS and friends,
while
>>>>>>>> they will not guarantee you the resources you want I suspect
you
>>>>>>>>
>>>>>>> might
>>>
>>>>  get somewhere close most of the time. Though again, if you need
>>>>>>>>
>>>>>>> these
>>>
>>>> as
>>>>
>>>>>  a minimum resource at all times we will have to look elsewhere.
>>>>>>>>
>>>>>>>> There might be some money free in the budget to allow you
to use
>>>>>>>>
>>>>>>> such
>>>
>>>> a
>>>>
>>>>>  service. Before we throw up n VMs in an attempt to guess if it will
>>>>>>>>
>>>>>>> work
>>>>
>>>>>  for you, can you perhaps go back to your PMC and discuss the idea of
>>>>>>>>
>>>>>>> VMs
>>>>
>>>>>  with the likes of AWS, and when you find a vendor please do a little
>>>>>>>> research into the options, then let us know what you want
to do.
>>>>>>>>
>>>>>>>   Please
>>>>
>>>>>  note that the budget is not endless, and if you want to saturate
>>>>>>>>
>>>>>>> 20Mb/s
>>>>
>>>>>  or more, then I guess the bill from AWS will be significant and you
>>>>>>>> might not get a lot of AWS time as a reuslt - so please look
>>>>>>>>
>>>>>>> carefully
>>>
>>>>  and come back to us with options and we can take those forward for
>>>>>>>>
>>>>>>> you.
>>>>
>>>>>
>>>>>>>> I'm pushing this back to you to do the research as we are
kinda
>>>>>>>>
>>>>>>> snowed
>>>
>>>>  under somewhat and if you can do this part we might just be able to
>>>>>>>>
>>>>>>> say
>>>>
>>>>>  "go for it" and let you have at it.  I stress 'might' :)
>>>>>>>>
>>>>>>>> --
>>>>>>>> Cheers,
>>>>>>>> Tony
>>>>>>>>
>>>>>>>> ------------------------------**----
>>>>>>>> Tony Stevenson
>>>>>>>>
>>>>>>>> tony@pc-tony.com
>>>>>>>> pctony@apache.org
>>>>>>>>
>>>>>>>> http://www.pc-tony.com
>>>>>>>>
>>>>>>>> GPG - 1024D/51047D66
>>>>>>>> ------------------------------**----
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>> --
>>>>>>> Cheers,
>>>>>>> Tony
>>>>>>>
>>>>>>> ------------------------------**----
>>>>>>> Tony Stevenson
>>>>>>>
>>>>>>> tony@pc-tony.com
>>>>>>> pctony@apache.org
>>>>>>>
>>>>>>> http://www.pc-tony.com
>>>>>>>
>>>>>>> GPG - 1024D/51047D66
>>>>>>> ------------------------------**----
>>>>>>>
>>>>>>>
>>>>>>>
>>>>
>>>> Cheers,
>>>> Tony
>>>>
>>>> ------------------------------**----
>>>> Tony Stevenson
>>>>
>>>> tony@pc-tony.com
>>>> pctony@apache.org
>>>>
>>>> http://www.pc-tony.com
>>>>
>>>> GPG - 1024D/51047D66
>>>> ------------------------------**----
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>
>>>
>>> --
>>> WBR
>>> Maxim aka solomax
>>>
>>>
>>
>>
>>
>
> --
>
>
>
> Cheers,
> Tony
>
> ------------------------------**----
> Tony Stevenson
>
> tony@pc-tony.com
> pctony@apache.org
>
> http://www.pc-tony.com
> GPG - 1024D/51047D66
> ------------------------------**----
>



-- 
Sebastian Wagner
https://twitter.com/#!/dead_lock
http://www.webbase-design.de
http://www.wagner-sebastian.com
seba.wagner@gmail.com

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message