Submitted by: rudi f (rudibr78)

Votes: 1

This is a rather weird behaviour and I probably wouldve not create a ticket but I saw it happening on 3 distinct and unrelated win 2003 servers running FB 2.1.3 and 2.1.4

We had a large surge of users as of last week and the server response got extremely slow.
After a lot of debuging and optmizing queries/indexes, it became clear all of the slow down was being caused by slow connection speeds.

We did a small benchmark tool to test for connection speed and concurrent connections. The tool consists on connecting and disconnecting N times to the database. By executing/instantiating the tool multiple times we could simulate concurrent connections.

The connection times averages were pretty abysmal :
1 connection : avg 70 ms connect time
2 connections : avg 120
3 connections : avg 180
4 connections : avg 250

Having to deal with the issue ASAP we got a new more robust win2008 64 bit server. We were glad to see the problem was gone. Our benchmark tool had 6 concurrent connections at 15-30 ms (same as 1 connection).

Out of curiosity tho, I decided to run the tool on my local win7 23bit 1gb dev machine.
I got the same fast results as the win2008 server with lots of concurrent connections.

I then ran the benchmark on 2 other random win2003 machines (some of our client's machines).
We are not the admins of those setups and they may vary, but the progressively bad connection results appeared on those machines too.

I understand this was something of an issue with FB 1.0 and fixed in 1.5 , it doesnt sound to be exactly the same thing but its happening anyway.

On our problematic server we upgraded, we had 60 users using temporary connections (permanent connections/pooling unfourtunately is not an option for us) , when the connections reached about 25-30 simultaneously, the benchmark tool was responding up to 2.5 - 3 seconds to estabilish a new connection.