This is in response to a few queries on how UO pooling is performing in a live environment.
Firstly a few points to remember 1. This is the use of pooling in UO.NET in an ASP.NET environment. 2. It is important you understand how iis and asp.net work b by understand I mean the use of application pools, and how they thread. 3. Connection pooling basically runs the connection to the database as a thread under the application pools worker process. Each of these connections is for an ip and account and takes a licence. Therefore if you have multiple worker processes either because you have multiple sites not sharing application pools or are in a web garden environment then expect multiple licences being taken up. 4. UO.NET connection pooling uses performance counters which means it needs access to the windows registry. You need to either update the access on the registry keys required for performance counters or change the user that the application pool runs as to one with access ! b Google on basp.net performance counters registryb to find out more on this. 5. There seems to have been a bug in the earlier releases of UODOTNET.DLL b I tested with versions 2.1.0.7137 and 2.1.1.7211. My pooling seemed to suddenly stop working b I think it may have been something to do with using named common but I never got round to testing that theory. Version 2.1.2.7232 dated 29 Oct 2007 fixed the problem. I would recommend you use the latest uodotnet.dll ! 6. Pooling means that each request for u2 does not need to open a new socket and start a new udt process. So the biggest gains in using pooling is If you have lots of small requests as you would have in web environment. If you open a UO connection and spend a minute performing actions then close it, pooling would not give much of a gain. Now some stats b These are on a Windows server 2003 Standard edition server with 1 dual Xeon, and unidata 7.x on RHEL ES version 4 on a dual twin proc Xeon. Both machines hosted in a hosting centre in Manchester UK and running about 12 live websites written in asp.net and connecting to ud on the backend (example sites are http://findaholiday.timesonline.co.uk/ , http://holidaysearch.telegraph.co.uk/europe.xhtml , http://express.adprs.net/ , http://travel.theholidayplace.co.uk/cuba.xhtml , http://jet2holidays.adprs.net/spain.xhtml ) 1. Firstly the results of a live website that is not under an especially high load. This is quite a good test as the asp.net side of this site does very little, It basically places an xhtml snippet, generated by the u2 server, into a xhtml page, that is inturn shown as an iframe inside a customers page (examples of this are the advertiser links boxes on the http://www.telegraph.co.uk/travel/ site). These are some stats from a ping service that accesses a particular page every 5 minutes. The site used connection pooling up until the 6th March, I then changed it to not use pooling for 5 days. It then uses pooling again from the 11th. As you can see there is a marked difference. Date Avge Time Total No Hits 1 511 ms 190,097 2 480 ms 233,044 3 540 ms 254,799 4 568 ms 217,077 5 526 ms 258,157 6 1,312 ms 277,823 7 1,677 ms 185,731 8 1,143 ms 160,699 9 1,044 ms 200,688 10 1,169 ms 263,025 11 582 ms 2. Now some performance hits b this was done using Microsoftbs Web Application Stress tool (WAS) on a PC connected to the internet with a normal domestic 8Meg ADSL line (this accounts for the high TTLB). Two test results, both with 25 threads lasting 1 minute. Both tests where run multiple times with very similar resultsb& a. Without pooling Hit Count: 184 Requests per Second: 6.13 Time to first byte (in milliseconds) -------------------------------------------------------------------------------- Average: 2884.90 Min: 258.68 25th Percentile: 1942.94 50th Percentile: 2313.63 75th Percentile: 4264.76 Max: 8468.82 Time to last byte (in milliseconds) -------------------------------------------------------------------------------- Average: 3635.09 Min: 354.75 25th Percentile: 2106.91 50th Percentile: 3320.31 75th Percentile: 5279.63 Max: 13653.23 b. With pooling Number of hits: 559 Requests per Second: 18.62 Time to first byte (in milliseconds) -------------------------------------------------------------------------------- Average: 595.02 Min: 166.76 25th Percentile: 272.33 50th Percentile: 316.37 75th Percentile: 606.62 Max: 9698.60 Time to last byte (in milliseconds) -------------------------------------------------------------------------------- Average: 1111.46 Min: 268.82 25th Percentile: 336.82 50th Percentile: 423.89 75th Percentile: 1956.06 Max: 21289.83 Conclusion - pooling in our setup is about 3 times faster. HTH Symeon. ------- u2-users mailing list u2-users@listserver.u2ug.org To unsubscribe please visit http://listserver.u2ug.org/
