Multi-server

[ShortyA]

Hi,
I am about to implement the following configuration :
Cache Server -2
Page Server - 2
Job Server -2
WCS, APS and FRS - 1 each.

Rather than implementing it on a multi-processor box I want to split it across a couple of servers. Does anyone have any recommendations as far as whether I should slam both of the server instances on one server or split them one on each ie both Cache servers on one box or one Cache server on each box ?

Many thanks - ShortyA

 

[synapsevampire]

That would depend on how you're using CE.

It appears that you're not considering failures in this scenario.

I haven't found any great documents on configuring CE, though the Using CE book has a few blurbs of interest, and there's a configuration guide that somes with CE.

The bottom line is the CD, not so surprisingly, suggests contacting one of their consultants.

I say imnplement however you like, but be prepared to monitor and adjust.

-k

 

[ShortyA]

Thanks for that. I think from a failure point of view that it may be worth splitting them across two. Is it possible to have two lots of each server component but have some disabled on the second server.

For example have two APS server components with one being passive in case of failure of the main one ?
ShortyA

 

[Turkbear]

Hi,
Crystal has a document ( developed for 8.5 but the concepts apply to 9 as well) called
'Implementing Crystal Enterprise for High Availablilty'
( search the knowledgebase at

http://support.crystaldecisions.com

for
a file called ce85_high_availability.pdf)

It describes various deployment scenarios for maximum availablity and performance.

While looking through the support site, also search for papers on
Disaster Recovery Planning (DRP)
and
Configuring Multiple File Repository Server Pairs.

All are very helpful in determining what is best for your situation.

 

[synapsevampire]

The High Availability is pretty vague (and old) too:

http://support.crystaldecisions.com/communityCS/TechnicalPapers/ce85_high_availability.pdf.asp

Though it does describe failover pretty well, which is desperately needed for CE

-k

 

[ShortyA]

Thanks for the info folks. I have downloaded the briefs and am having a look at them today. Once I have decided on the config I will post it up for info.
ShortyA

 

[RatSkin]

CE 9 handles the fallover better than previous versions.

The best part is the way the FRS now works. you can have two FRS servers now.

This is how i have setup a Two server Install.

each has all the CE services installed, then a cluster is setup accross the two machines.
APS database stored on a 3rd server
FRS are saved on to a Windows Clustered Server, then all the services are started, the APS decides which FRS is active, the other frs just sits and waits for a fallover then kicks into action. This is the same for the APS services.

once this has been setup and running you have a very reliable fault Fallover service.

 

[ShortyA]

Ratskin,
thanks for sharing your setup with me. Out of interest, how many CE licences do you have and what type are they ?
Thanks - ShortyA

 

[RatSkin]

That machine setup has got 10 named and 20 concurrent.

I have setup more with different configurations ranging from single processor license to 2 processors, 5 named to 260 named and 5 cons to 80 cons.
.

Rat

 

[nbk13nw]

In our current configuration we have the following:

2xWeb Servers - IIS / Web Connectors (Each looking at the two WCS servers)

2xWeb Component Servers - 2xWCS / Event Server / Cache Server

2xASP Servers - Clustered

2xJob/Page Servers - 4xJob Servers/Page Server/Input FRS / Output FRS

There is still a singlwe point of failure (CE 8.5) should the "primary" JSPS go down with CE looking for a single Output FRS.
15,000+ users and 2,500 concurrent licenses.