[Foss4g2009] Specs for a machine for the CCIP?

Mon Jan 19 20:09:02 EST 2009

We've been discussing some different options with regard to the showcase 
infrastructure.  We've come up with a slightly different view of things.

We were looking to have separate (but identical) machines for the 
databases, one for Oracle and ArcSDE (assuming we're hosting these) and 
one for PostgreSQL/PostGIS.  If others come to play we'll get one for 
them.  In this way the DBs are on an even footing, and the machines 
don't need to be all that state-of-the-art.  We can smaller machines for 
the application servers.  We may not need one machine per service, but 
we should really isolate the I/O as much as possible.  If we have 
GeoServer, MapServer and MapGuide running in different VMs on teh same 
box, it would be nice to give each their own drive so they don't thrash 
the machine.  How many of these servers depends, of course, on how many 
services we end up running.

I think it's a poor plan to have a single machine running everything, 
and some of the big boys aren't likely going to be happy with that.  By 
sourcing cheaper machines we can even the playing field and keep things 
comparable.

Mark Leslie
Geospatial Software Architect
LISAsoft

-------------------------------------------------------------
Ph: +61 2 8570 5000 Fax: +61 2 8570 5099 Mob: +61
Suite 112, Jones Bay Wharf 19-21 Pirrama Rd Pyrmont NSW 2009
-------------------------------------------------------------

LISAsoft is part of the A2end Group of Companies
http://www.ardec.com.au
http://www.lisasoft.com
http://www.terrapages.com

Cameron Shorter wrote:
> Ricardo,
> Based on your experience and track record in marrying Geospatial 
> Software and Virtual machines, your opinion holds a lot of weight.
> 
> So thank you for sharing your thoughts.
> 
> And yes, we would be delighted to have you play a significant role in 
> the CCIP and guide its development.
> 
> Ricardo Pinho wrote:
>> Hi Cameron and Frank,
>>
>> As you probably know, Virtualization is presently a very active area 
>> of business, with several players (big ones) trying
>> to get on the top, so it would be very easy to get a sponsor!
>> Please carefully  read the following pages:
>>  
>> http://www.virtualization.info/radar/
>>  
>> http://www.virtualization.info/2008/10/microsoft-already-took-23-of.html
>>  
>> For VM technology you can choose from a variety of solutions:
>> VMware, Sun, Microsoft VirtualPC & HyperV, Xen & VirtualIron, Linux KVM,
>> … an endless list!!!
>> http://en.wikipedia.org/wiki/Comparison_of_platform_virtual_machines
>>  
>>  
>> Well, I must agree with Frank, VMware are the original guys, the 
>> oldest on the market
>> and who offers today the best desktop and server performance, features 
>> and
>> stability (I suppose!). So they are a very good option, and that was 
>> mainly why
>> I have chosen VMware for the GISVM project! (no sponsor reason!!!)
>>
>> Never the less, it would be desirable to use a standard and compatible 
>> solution, so it
>> could be run on any technology. The closer thing I know to this is the 
>> work
>> done on: www.dmf.org
>>  
>> OVF:
>> http://www.dmtf.org/standards/published_documents/DSP0243_1.0.0.pdf
>> http://www.vmware.com/appliances/learn/ovf.html
>>  
>> and  
>> VMAN:
>> http://www.dmtf.org/initiatives/vman_initiative/  
>> But in practice, I compare this to the early years of OGC, just a big 
>> hope! ;-)
>>  
>> One more reason why vmware should be considered, is that VMWare is 
>> collaborating with the
>> Ubuntu distro, on their JeOS version! An advantage I am currently 
>> using on the
>> next GISVM release that I hope to introduce at ESLAP 2009 ( OpenSource 
>> in Portuguese
>> Public Administration Encounter - 28.Jan ).
>> http://mapas.igeo.pt/eslap2009/Eslap2009_Prog.html
>>  
>>  
>> Now, concerning the CCIP project. I must confess that, until know, I 
>> didn’t quite understood how you plan to execute that excellent idea!
>>  
>> But I would be very pleased to participate on that project and include 
>> all work done on the GISVM.
>> Only if you think it would be adequate, of course!
>>  
>> Cheers,
>> Ricardo Pinho
>>
>>
>>
>> ----- Mensagem original ----
>> De: Cameron Shorter <cameron.shorter at gmail.com>
>> Para: Frank Warmerdam <warmerdam at pobox.com>
>> Cc: foss4g2009 <foss4g2009 at lists.osgeo.org>
>> Enviadas: Segunda-feira, 19 de Janeiro de 2009 22:27:41
>> Assunto: Re: [Foss4g2009] Specs for a machine for the CCIP?
>>
>> The VM technology has not been specified yet. I've heard a number of 
>> people speak up for VMWare and we will try to get VMWare to load us a 
>> license for the conference in return for marketing.
>>
>> Re memory:
>> * I'm hoping that all virtual installs can be active for the duration 
>> of the conference. Is that a realistic expectation from based on 
>> hardware constraints?
>>
>> Re adding other servers:
>> I expect this to be an area of debate. Our requirements are:
>> 1. Demonstrate interoperability
>> 2. Allow side by side comparison of applications (some vendors might 
>> not like this). Specifically, make hardware specs equal so we can 
>> focus on the software.
>> 3. Provide a legacy testbed which can be passed on to future 
>> conferences with minimal re-configuration
>>
>> So maybe we will ask all vendors to build their image on a virtual 
>> server, but that they can bring their own server as well, or purchase 
>> their own server if they wish.
>>
>> Frank Warmerdam wrote:
>>  
>>> Cameron Shorter wrote:
>>>    
>>>> For background the CCIP is defined here: 
>>>> http://external.opengis.org/twiki_public/bin/view/ClimateChallenge2009/WebHome 
>>>> Basically: One server with multiple virtual machines
>>>>
>>>> As yet, we have not defined our hardware requirements, and 
>>>> everything is on the table as an option. (Especially since my 
>>>> experience with hardware is largely limited to installing linux on 
>>>> computers I've found in dumpsters).
>>>>
>>>> So please speak up with all the advise you can fire at us.
>>>>       
>>> Cameron,
>>>
>>> Ah, OK.  Is there a defined VM technology?  VMWare seems to be the most
>>> widely supported and well known.  Presumably Win32 (perhaps XP?) VM
>>> instances will be available as well as 32bit linux instances?
>>>
>>> Assuming that not all of the instances really need to be active at once,
>>> I would hope that the memory requirements will not be the sum of the
>>> memory requirements of the VMs.  Hopefully you would get at least 8GB
>>> of RAM which should be enough to have several active VMs with fairly
>>> beefy software running.
>>>
>>> Ultimately, I hope you will keep an option for some participants to 
>>> provide
>>> their own server that can be plugged into the network but I do like the
>>> convenience of a multi-VM main server as long as it is beefy enough.
>>>
>>> Best regards,
>>>     
>>
>>
>> -- Cameron Shorter
>> Geospatial Systems Architect
>> Tel: +61 (0)2 8570 5050
>> Mob: +61 (0)419 142 254
>>
>> Think Globally, Fix Locally
>> Geospatial Solutions enhanced with Open Standards and Open Source
>> http://www.lisasoft.com
>>
>> _______________________________________________
>> Foss4g2009 mailing list
>> Foss4g2009 at lists.osgeo.org
>> http://lists.osgeo.org/mailman/listinfo/foss4g2009
>>
>>
>>
>>       Veja quais são os assuntos do momento no Yahoo! +Buscados
>> http://br.maisbuscados.yahoo.com
>>   
> 
> 