Link to home
Create AccountLog in
Avatar of duncaninnz
duncaninnz

asked on

Interbase database running slow

I have a client that runs a 4 gig database on interbase 7.5. It runs slow. I am not the support for the application but support the server and network. I have suggested they should have a seperate interbase server for the application. They have about 25 users. They are running it on a single HP ML330 G3 server with two processors(about 3gig) 3gig ram two raid 1 arrays. The server is about 4-5 years old. The OS is SBS 2003 so they run Exchange on the same box. They also have a SQL Express database plus all the user docs and printing. The app provider are saying the server should be ok to run their app and there must be something else wrong. They are suggesting there is something wrong with the network at the moment. What I am looking for is an informed oppinion from a suitably qualified person that can point to any recommended standard or best practice to say that this approach is either right or wrong. I have another client running interbase and their app provider recommends a seperate server and doesnt support it on SBS as it doesnt run well with Exchange. I suspect this is an Interbase recomendation but cant find any reference to it. Also should they be running an interbase backup/restore on a regular basis.
Avatar of Ustin19
Ustin19

what is delta between oldest interesting and next transaction? It is possible that client app is designed incorrectly and transaction delta becomes too big at time.
You need do SWEEP DATABASE periodically and perform backup\restore loop
Avatar of duncaninnz

ASKER

Thanks I will check this out. Just a bit more info. I performed a test tonight with two users logged on and the server on a seperate switch ie no other activity on the network A transaction that basically calculates a quote based on a list of priced items and labour content took 1 min 50 seconds to complete. To print the results about 1 min 20 seconds. The same transaction run from two PCs at the same time took 50% longer. Neither result is acceptable to the client and does appear excessive for a fairly straitforward transaction. I realise that this is not an entirely scientific test.
ASKER CERTIFIED SOLUTION
Avatar of Ustin19
Ustin19

Link to home
membership
Create a free account to see this answer
Signing up is free and takes 30 seconds. No credit card required.
See answer
Avatar of Nick Upson
please do a backup and restore and then rerun the same test.
I think I may have put this question in the wrong zone. Its Borland Interbase not Firebird. Sorry.
the interbase/firebird zone is for both products as they come form the same base and are still very similiar, the request for output from gstat still applies, the path will just be different
Gstat attached
Publication1.jpg
SOLUTION
Link to home
membership
Create a free account to see this answer
Signing up is free and takes 30 seconds. No credit card required.
At the time of the gstat there were no active users.
there has to have been at least one active transaction
There was no one in the office but the users have a habit of leaving the app open and their screen turned on.
well that counts as a user, my guess is that someone has done that for several days, on holiday or something like that. You need to do so kind of timeout in the client program so that this can't happen.
I did suggest that the interbase service is stoped prior to the backup and restarted after. The app provider was non commital on this. Presumably this would kill any sessions left open by users. Are you saying though that this explains the 15k active transactions
yes, the 15k transactions are being held open by something, rather than stop the service and crash stop I'd take the database off-line which allows the database engine to manage the transactin shutdown (gfix -shut from memory)
the 15k may not be all the problem but you will never find the rest until that is sorted
Thanks for your help on this. I think I have enough to batter the app provider with this. They have been looking at it for 2 years so I would have thought they would have done something by now.
it can be client software problem (too big number of transactions which are not commited, or commited with retain option {.commitretaining method used}) - i think trouble is here, bcs 15k delta is TOO big for 25 extrimely active users
The app provider has now fixed their application and the outstanding transactions have massively reduced and speed issues have been resolved. Thanks for your help. Surprised it took the app provider two years for what you sussed in minutes.