|
Prev: Need help: wio rease from 1% to more than 50% after increasing SGA size
Next: ROWID, ROWNUM and views
From: strepxe on 27 Jun 2008 02:15 Hi, We have an application that uses ODBC to connect our client application to an Oracle database. Currently we run at approximately 10-100 concurrent users with no major problem. A request has come in asking if the application would be able to support approximately 1000 concurrent users. My feeling on this is that the number of concurrent users could adversely affect performance and some pooling technology/ architectural change would be needed. What I lack is any guidelines on the limitations of ODBC in this regard. Can anyone point me in the right direction? Ger
From: gym dot scuba dot kennedy at gmail on 27 Jun 2008 07:10
<strepxe(a)yahoo.co.uk> wrote in message news:c107326d-54d2-4bf7-abe2-e8d49b0fbbae(a)k37g2000hsf.googlegroups.com... > Hi, > > We have an application that uses ODBC to connect our client > application to an Oracle database. Currently we run at approximately > 10-100 concurrent users with no major problem. A request has come in > asking if the application would be able to support approximately 1000 > concurrent users. My feeling on this is that the number of concurrent > users could adversely affect performance and some pooling technology/ > architectural change would be needed. What I lack is any guidelines on > the limitations of ODBC in this regard. Can anyone point me in the > right direction? > > Ger The answer is really more complex. We don't know how well written your application is or is not. Using ODBC doesn't mean it is well written or not. (it could be either) If you do not use bind variables then there is little chance your appliaction will scale. In short insufficient information. Basically, with 1,000 users you will bottleneck somewhere, how large is that bottleneck and what is it. If you don't use bind variables buying more CPU (pr faster CPU) or memory won't do bupkis to solve the problem. If you are IO bound then adding more CPU won't help. (It may actually hurt; each query will get in line for the disk faster.) Jim |