One of the questions found in the FAQ of Micromuse and the netcool users group,
is " how many events can a system handle?.
Although we admit that it IS a difficult question, Micromuse's default answer "contact a sales representitive" is rather blunt.
Admitting that it is nearly impossible to predict what kind of hardware is needed, it is obvious that following parameters will influence the overal performance:
- CPU-speed: 2GHz CPU probably better than 75MHz CPU
- CPU-data-width: 64-bit CPU probably better than 32-bit CPU
- CPU architecture: There MIGHT be a difference in performance between Intel, SUN and HPPA, if all other parameters could (!) remain identical
- Memory-size:Below a certain value, systems will start swapping (Bad Thing), above a certain value, no improvements should be expected.
- Memory-speed:The longer the CPU has to wait for its data, the worse the overal performance.
- The Operating System: On identical hardware platforms, some OS perform better than other.
- The version of the O.S.
It is *NOT* recommended to run Solaris-9 on a sun Sparc-Station-4, while Solaris-2.5.1 runs nicely though...
- The record size of the alarms.status database
- What and how many automations run, and at what interval
- How many events are injected into the server.
- How many desktops are connected to the object server.
With a given configuration, it is possible to increase the number of (inserted) events, and measure the cpu-load of the object-server process.
Thus, it is possible to detect what the maximum load is for that particular configuration.
In lab-situations, given several hardware configurations, it gives the opportunaty to determine wether a specific config is ether grocely under configured, or an overkill.
In customer-situations, you can prove the sistem can handle what you promissed.