Skip to main content

We are running Maximizer 7.5 Enterprise Pervasive C/S on a Windows 2000 Terminal Server.

Our configuration has a dedicated file server that acts as our Pervasive SQL Server - then a second server as the application server.

We have some custom forms for data entry that require Maximizer ODBC so the network traffic is quite high.

Any suggestions on performance increases possible by hosting the Pervasive SQL Server on the Terminal Server as well? This would eliminate our network traffic, but increase the load on the machine. We only have 10 users currently, so load is not being tapped at the moment.

Any discussion appreciated.

Paul
Original Post

Replies sorted oldest to newest

I'm assuming that you have the Pervasive client connector running on the Terminal Server? Running the SQL Server on the TS may not give you enough of a performance boost to negate the extra load on the TS, but this would depend on what assumtions you made when sizing your TS originally. If the biggest bottle neck is the LAN, then implementing gigabit for the servers and switches might make more sense than reconfiguring the Pervasive SQL Server configurations.

Performance issues are very subjective due to the fact that it all depends on the pricise configurations in place, from network infrastucture to server configurations and what else the servers are doing. Have run performance monitors on your servers to try and determine where the bottlenecks are?

-Amer
We are going to try the Gigabit connection as we have Gigabit cards in both servers. The reason for my suspecting the LAN traffic is simply our observations that that ODBC connections for Maximizer are pretty heavy and any latency issues are generally cause for trouble.

Do you have some specific suggestions for performance measures witn Windows Performance Monitor? I'm not sure which are the best indicators to determine network bandwidth, latency, or other general network issues.

Thanks,
Paul
The simplest tool to measure latency is, believe it or not, the ping command. If the return time is <1ms, then latency isn't your main problem. User perfmon to measure throughput, bandwidth, and failed packet stats for the network interface(s) and the TCP, IP, and UDP protocols. Also, measure CPU, RAM, pagefile, and disk stats to see if they are adversely affecting the server's performance. However, even if any, or all, of these counters seem to be returning high values, that might be perfectly normal for the specific configuration you have and may be an indicator of the system reaching some sort of load limit. If you are convinced it is a network load issue, then you may also want to consider implementing load balancing via NIC teaming, if you have multiple NIC's in your servers.

Also, after gathering your performance stats, look into tweaking the settings in your ODBC configuration, i.e. connection pooling etc., and running the performance monitors again to see if there were any changes.

It all depends on how much time you want to put into this, which I guess depends on how poor the current performance is.

HTH,
Amer

Add Reply

Post
LEGAL INFO
CONTACT US
Copyright 2007-2018 Advoco Solutions Ltd. All Rights Reserved.
×
×
×
×
Link copied to your clipboard.
×