Server Spec for Crystal Reports Server 1

[Sacheveral]

We are hoping to install a new Crystal Reports Server XI installation (20 concurrent users) in the near future, with both Crystal and Xcelsius content, and on the principle that the hardware is a lot cheaper than the licenses, we would like to get the best server up to the point where it stops adding any value.

Firstly if we were to go for 8 dual core processors, how would this compare to 4 quad core processors in terms of performance?

Secondly if we got 64 GB of RAM would the application be able to use it all? If not what is the max it can handle?

Lastly any other hardware specs that you would recommend for top end perfomance?

Thanks,

Joe

 

[MJRBIM]

To be frank, I haven't seen any BOBJ sizing-guide documents that show the comparable performance on multi-core servers...if they have them, they don't share.

The sizing-guide used in the "Design and Deploy" course is based on older single-processor servers.

You might want to try contacting BOBJ support directly on this request.

 

[Sacheveral]

Tried that, they were not very helpful. All they seem able to quote is the minimum spec! I got the distinct impression that they didn't really know, and I have asked several of their engineers about server specs.

Aside from the specific Crystal application, do you have any general rules of thumb over processors?

Thanks,

Joe

 

[MJRBIM]

Some of the Processor/RAM specs are limited by the OS you are going to run under.

Windows Server 2003 R2 "Standard Edition" has a limit of 4 GB or RAM. So don't buy any more than that if that is the OS you are going to be running.

Windows Server 2003 R2 "Enterprise Edition" supports up to 64 GB or RAM.

http://www.microsoft.com/technet/windowsserver/evaluate/features/compare.mspx

Since you are only running 20 concurrent users in CR SERVER for RPT and Xcelisus, I would suggest that you at most need four (4) processors - and ensure you have a Fast RAID compliant disk array on that server.

 

[Sacheveral]

Thanks for that. Although we only have 20 concurrent users, we also pre-generate about 15,000 report instances a day, and email about 5,000 of these, so therefore need a fair bit of resource.

Thanks,

Joe

 

[MJRBIM]

15,000 report instances a day (approx 625 per hour)

OK, well be really careful about how you set-up your multiple CMS services, and your scheudled instances.

Based on the baseline single-processor spec, BOBJ says each service will handle a MAX of 150 simultaneous requests per Service/ per CPU.

That means that if you have a "Peak-Load" of scheduled instances (eg. 1,000 instances at 7:00 AM), you would need 7 CMS services running on a minimum of 7 CPUs (7 x 150 = 1,050 MAX. simultaneous requests).

Your other option is to split them into smaller blocks at different times.

EXAMPLE : If you were running four (4)CMS services on a dedicated (4) CPU server you could (4 x 150 = 600 MAX. simultaneous requests). So you would schedule 500 instances at 7:00 AM, and the other 500 instances at 7:15 AM).

If you overwhelm the CMS services with too many simultaneous requests, you can get "ghost" instances where a RPT-instance is picked-up by two CMS services at the same-time, and they BOTH create an Instance in HISTORY plus a copy of the Recurring Job.

 

[Sacheveral]

Hmmm, interesting. A lot of things we haven't previously considered there. Our current setup is as follows:

1 server with 4 single core hyperthreaded processors (so 8 virtual cores). We use Crystal Enterprise 10, so hopefully roughly equivalent to Crystal Reports Server XI. We have our reports scheduled in 3 blocks:
2000 at 8pm
5000 at 12am
8000 at 8am

They are all scheduled for exactly the same time in their batch, although are triggered by a file event that occurs just after that time, generated by the SQL Server to say the data is ready.

However we only have 1 of each Crystal service set up and have never experimented with more. Also the Report Job Server has the "Maximum Jobs Allowed" set to 12, which makes the server work hard, but without collapsing. The reports just queue up and run in random order amongst the batch scheduled at that time. The number of hours that the server is generating reports solidly per day is roughly 15 hours, so that is about 1000 reports per hour, or 1 report every 3.6 seconds. Its been running like this for a couple of years now, although the number of reports has been steadily increasing, hence our desire to upgrade the hardware.

What would be the potential benefits of creating more instances of some of the services? Also how would you suggest we quantify these changes and what would you set "Maximum Jobs Allowed" to? I am happy to change the scheduling if you can suggest something more optimal.

Many thanks for your help,

Joe

 

[MJRBIM]

Why are you stepping-down from CE-10 to Crystal Reports Server XI...?

Usually as a support-paid CE-10 client, you would migrate to Business Objects Enterprise XI (R2) - rather than the LITE version in CR Server.

Under BOE-XI (R2), these numbers below are "ballparks" from the sizing-guide when services are running ALONE on a single-processor, so you will need to finesse them for your hardware and reporting loads.

CMS = 1 CPU can handle 150 simultaneous user requests
CACHE = 1 CPU can handle 200 simultaneous viewing sessions
PAGE = 1 CPU can handle 25 simultaneous viewing sessions
RAS = 1 CPU can handle 25 simultaneous processing threads
CR JOB = 1 CPU can handle 5 concurrent report jobs

 

[Sacheveral]

Once again thank you for your helpful response, however I am still not sure how many instances of each service I should have. Would one per CPU be a reasonable solution, and what dictates the number of instances of a service, i.e. why not just have for example one report job server service running on a server with 4 CPUs, therefore using your sizings above giving you 20 concurrent report jobs? What would make you want to have separate instances of the service running on the same server?

In answer to the question of why effectively downgrade, it is because the licensing of Business Objects is such that even with a support contract, to get the same functionality as our current CE10 setup we would have to invest significantly in further licenses, and as we do not wish to use WebI this seems unnecessary, although it is possible we could pursue the BO Pro route in due course.

Thanks,

Joe