I'm presently focusing on an n-tier system and fighting some database performance issues. An area we've been looking into may be the latency between your database server and also the application server. Within our test atmosphere the average ping occasions between your two boxes is around .2ms however around the clients site its more around 8.2 ms. Is the fact that somthing you should be concerned about?

For the average system exactly what do everyone think about a resonable latency and just how can you start testing/calculating the latency?

Karl

In a nutshell : no !

What you need to monitor may be the global performance of the queries (ie transport towards the DB + execution + transport to your server)

What you might do is make use of a performance counter to watch time your queries usually decide to try execute. You'll most likely call at your answers are within the nanosecond area.

There is no such factor as "Reasonable latency". You need to rather think about the "Reasonable latency for the project", which may vary a great deal based on what you are focusing on. People not have the same expectation for any real-time buying and selling platform as well as for a read only amateur website.

No. 8ms is virtually minor.

On the linux based server you can look at the result of latency yourself using the tc command.

For instance this command will prove to add 10ms delay to any or all packets going via eth0

tc qdisc add dev eth0 root netem delay 10ms

make use of this command to get rid of the delay

tc qdisc del dev eth0 root

More particulars available here: http://devresources.linux-foundation.org/shemminger/netem/example.html

All programs will differ, however i have certainly seen situations where 10ms latency has already established a substantial effect on the performance from the system.