Database Reference
In-Depth Information
21.5 Database Tuning
To maintain acceptable database performance, it is necessary to carry out periodic
performance tuning. As the database is being used, it will exhibit a natural tendency to
degraded performance. The current database system may not be performing acceptably,
based on user-defined criteria, due to any of the following:
Poor database design
Database growth
Changing application requirements (possibly including a
redefinition of what acceptable performance is)
Note however, that there might be occasions when database tuning efforts are not
fully effective. When components that are external to the database, yet vital to the entire
client-server application performance, fail to perform acceptably, database tuning might
not help without the corresponding tuning of these other application infrastructure pieces.
The main components external to the backend database are the backend operating system,
the network, and the client operating system. The major examples are the following:
Very weak clients (PCs)
Network saturation
Very weak, saturated, or poorly tuned operating system
21.5.1 Tuning Goals
There are different ways of determining the goals of a performance tuning effort. A DBA
should consider them all. Database systems can be sampled on various quantitative
measures; the most important of these are the following:
Throughput: This is the accomplished work per unit time, as
measured by transactions per second (tps); higher is better.
Response time: This is the time it takes for an application to
respond, as measured in milliseconds or seconds, lower is better.
Wait time: This is the elapsed time a program takes to run;
lower is better.
In any system, throughput and response time usually run counter to one another
as tuning goals. If response time is high (bad), throughput might be high (good).
If throughput is low (bad), response time might be low (good).
Common sense helps when sorting out these two conflicting measures. The more
users that are concurrently using a system within a certain amount of time, the more
likely it is that each user will experience longer delays than normal, but the number of
transactions going through the system will be greater. On the other hand, if you decrease
 
Search WWH ::




Custom Search