Recommendation for scaling

[Former Member]

Dear SDN,

I am implementing a portal for 10,000 users

The amount of concurrent users is unknown, but lets assume 1000

The portal harware is SUN v440 4Xcpu and 16GB ram

The DB Hw is a sun V240 4Xcpu 8GB ram

Content is 2 IAC iviews for ESS benefits

How many Dispatchers and servers would you use?

Would the Database (oracle) go on the larger machine than the portal server?

Thanks

Jeremy Baker

[Former Member]

Thats funny because SUN made the same recommendation for my client, I want to know why?

Thanks

Jeremy Baker

[Former Member]

Basically that is what that server was bought for so the database was loaded there. : )

[Former Member]

Actally in 64 bit solaris 9 you can set the heapsize up to 3GB.

Thanks for all your input it has been quite helpful

PS. Why would you run the DB on the larger server?

Jeremy

[Former Member]

I dont know if the same rule of thumb applies, it probably does .... but on EP6 SP2, the guideline was one portal server node per 2 cpu (scale-in). Bearing in mind even on 64 bit solaris there may be a max jvm size (1.5Gb I think??) so your portal server looks a bit heavy on RAM, or light on CPU's.

[Former Member]

We run our Database on the larger server.

--Keith

[Former Member]

No, I mean do you think that the db should go on the larger machine? Or should the J2EE be on the larger server?

Thanks

[Former Member]

Jeremy,

When I mentioned switching the servers, that was meant only if you clustered the Portal J2EE Server with another server.

Thanks,

Keith

[Former Member]

I could switch the servers that the components are on,

However I thought that the DB uses less resources than the J2EE

Thanks

[Former Member]

Jeremy,

We recenly rolled out ESS via the Portal to a large number of users (50,000 Named Users). I would suggest that you look at adding an additional Portal Server and make this a cluster. This not only buys you the ability to split the load between the two servers, but it give you fault tollerance should one of the servers fail.

I would also consider loading the Oracle database on the larger server, especially if you cluster the portal server.

Hope that helps.

Keith