[Deploying Sakai] Deployment sizing question

Wed May 13 10:53:02 PDT 2009

Hi Todd,

Actually those are relatively old specs for our production  
infrastructure.  Here's our current production infrastructure at UM:

load balancer env:  HA pair Netscaler RS9800 load balancers
app server env:  5 Dell PowerEdge 1950 application servers, 16GB RAM  
running RHEL 5, apache 2.2.x, tomcat 5.5 w/ 6GB heap
(plus one additional PE1950 running outside the active cluster as  
search server)
db server env:  1 Sun Fire T2000 w/ 16GB RAM, running Solaris 10 and  
Oracle 10g (10.2.0.4)
file storage:  NetApp FAS3020 filer, multiple disk shelves for Oracle  
database and Sakai Resources file space, over NFS

This entire environment is replicated in a second data center (with  
one exception as we have a single Netscaler load balancer at the  
secondary site instead of a HA pair).  We use SnapMirror to sync the  
files between NetApp filers, and have an Oracle standby db at the  
second site for db replication & failover.

The diagram Anthony sent also reflects a significant number of  
external service/database dependencies.

We are planning to replace our Oracle db server hardware with Sun  
T5120's w/ 64GB RAM (expandable to 128GB) before Fall semester 2009.   
We're also increasing our file storage capacity on the NetApp filers  
before Fall.  We saw file storage usage increase by ~900GB during  
Winter 09.  We currently have about 5TB of file resources on the  
filers, most of which is from the past 2.0-2.5 years.

Oracle 11g is on our short list of infrastructure upgrades to address  
in the relatively near term.

Please feel free to contact us (CTools Operations Team) at ctops at umich.edu 
, we'd be happy to talk with you and anyone else about our  
infrastructure, our deployment considerations and decisions, etc.

Thanks,
Jeff

On May 13, 2009, at 12:49 PM, Anthony Whyte wrote:

> Todd--I recommend that you have a word with John Leasia at the  
> University of Michigan regarding deployment sizing given that in  
> rough terms, UF's user community is similar to that of Michigan.
>
> However, I should point out that UMich's deployment strategy (as  
> well as all others at present) involves supporting Sakai 2.x and  
> that at present no production sizing data exists for the  
> experimental work that comprises what some are calling "Sakai 3".  I  
> encourage you to check out the Sakai K2/Sling site at
>
> http://groups.google.com/group/sakai-kernel
>
> to learn more about the initiative.
>
> Cheers,
>
> Anthony Whyte
>
>
>
> Sakai CTOOLs 2.x infrastructure:
>
> Application server(s): 8 X Dell PowerEdge 2650, dual 2.4-3.2 GHz CPU  
> 32 bit, 4 GB RAM
> Total app servers: 8 (minimum 2 held in reserve)
> Db server(s): 2 X SunFire T2000, 16 GB RAM, standby mode  
> (replication server)
> File storage: 	File Storage: 1 TB file storage on AFS; database  
> Storage: 2 X NetApp FAS3020 filers, NFS. Bandwith allocation 1 Gbps;  
> tape backups, off-site storage. Standby database server and  
> corresponding NetApp filer at secondary data center location.
> Cluster/load balancing: 2 X NetScaler RS9800 Secure Application  
> Switch, High availability, 10/100/1000 Mbps copper, 1 GB RAM;  
> Radmind server/software provisioning, see http://www.radmind.org.
>
>
>
> <CTools Production.jpg>
>
>
>> From: Todd Williams <tcwill at ufl.edu>
>> Date: 13 May 2009 3:03:30 PM BDT
>> To: production at collab.sakaiproject.org
>> Subject: [Deploying Sakai] Deployment sizing question
>> Reply-To: tcwill at ufl.edu
>>
>> Hey folks,
>>
>> I'm a sys admin with the University of Florida and we're in the early
>> stages of planning the replacement of our Webct Vista system with  
>> Sakai
>> over the coming year.  I've been tasked with collecting deployment
>> information from other institutions that have implemented Sakai as  
>> their
>> primary/sole LMS.
>>
>> Specifically I'm interested in information that will help me develop
>> initial sizing recommendations in terms of hardware/software platform
>> for an institution of our size.  Just as a starting point UF  
>> frequently
>> has 1500 concurrent LMS user logins throughout the day.  We have  
>> ~45000
>> traditional students in addition to a fairly large number of
>> non-traditional distance education students.
>>
>> Coming from our current WebCT installation our preferred deployment
>> environment will be Linux and Oracle.  Via other projects my group  
>> has a
>> solid understanding of Apache/Tomcat as well.  So the underlying  
>> pieces
>> are familiar, but what isn't obvious is the sizing for our Sakai
>> deployment.  For WebCT we've been running with 6 application nodes  
>> and a
>> single, beefy Oracle 9i server.  After looking at the list of
>> deployments that the Sakai Foundation recommended(1), I see a  
>> number of
>> institutions of similar size using infrastructure similar to what we
>> want to do- this is encouraging.  However, I still have a handful of
>> questions, so here goes:
>>
>> What methods and metrics are people using to asses how many  
>> application
>> servers will be necessary?
>>
>> What network load balancers are being used to distribute load across
>> their application clusters?
>>
>> What methods are being used for node affinity within the load  
>> balancer?
>>  ie. stickytimes, cbr/cookie persistence, something else?
>>
>> Has anyone implemented Oracle RAC in support of their Sakai system?
>>
>> Has Oracle 11g been successfully implemented as the backend for  
>> Sakai?
>>
>> For what it's worth, given our desired implementation timeline it  
>> looks
>> like our pilot rollout and eventual full production rollout will be  
>> with
>> Sakai 3 (assuming the predicted release dates for RC and final  
>> Sakai 3
>> hold up).  I only mention that for consideration in case the move the
>> version 3 will affect any of the underlying components in the support
>> matrix.
>>
>> Thanks in advance for any information you can provide that will  
>> help in
>> my endeavor.  I'm excited about moving to Sakai, and I look forward  
>> to
>> working with the community.
>>
>> Cheers!
>> -- 
>> Todd Williams
>> tcwill at ufl.edu
>> CNS Open Systems
>> University of Florida
>>
> _______________________________________________
> production mailing list
> production at collab.sakaiproject.org
> http://collab.sakaiproject.org/mailman/listinfo/production
>
> TO UNSUBSCRIBE: send email to production-unsubscribe at collab.sakaiproject.org 
>  with a subject of "unsubscribe"

--

Jeff Cousineau
ITCS, University of Michigan
jeff.cousineau at umich.edu