We've set up a report farm with two servers, both 64 bit with 4 CPUs each. One has 16Gig and the other 8Gig of memory. We're using Windows NLB and the load test software confirms that the NLB is working. When we run a number of concurrent reports, both servers get utilized, but they only work on a few at a time. The report server queue doesn't seem to be fully utilizing the hardware. From a prior post I've learned that the report server queue automatically runs 4 reports per CPU. This is not occuring for our setup. Has anyone else experienced the same? Are there any configurations that need to be set to open the queue up? The reports are heavy (300,000 records grouped and summed). Does this affect the queuing process?
For interactive requests the queue is managed by ASP.Net. You might want to confirm that its settings are allowing multiple requests. What about the catalog database? Have you confirmed that you are not seeing an IO problem on that?
http://blogs.msdn.com/jgalla/archive/2007/03/21/reportservertempdb-io-saturation.aspx
|||Thanks John.We tested our throughput with the catalog database and it’s not a problem.The MaxActiveReqForOneUser setting in the RSWebApplication Configuration File is currently set to 0, which allows unlimited requests for a single user.We are using only one user id for all report requests.It’s embedded in our web application.Our web application is using the webmethod WebReports.Render from the RS SOAP API to run reports.Is there some limitation on the SOAP API methods that may cause the queue to backup?Is there any limitation on the number of simultaneously processing reports for one user? (Not sure if there’s a limitation difference between Active Requests, which could be queued, and Processing reports)
Below is a sample of our WebReports.Render method.
WebReports.Render(/MyPiersReports/XXXXX/Criteria_Reports/Reg_02_RegionCommodity, XML, null, "", ,this._paramValues, null, null, "Unicode (UTF-8)", "text/xml", this._historyParams,null,null)
|||There are no limitations in the SOAP API which would serialize all requests from a single user.
No comments:
Post a Comment