Strategic goal for zabbix development

[aotto1968]

Hi,
with this post I want to admit some ideas about zabbix improvements
to the community. the strategic goal is to replace BMC Patrol, an
allready known and widly used monitoring tool, with a zabbix open source solution. to achive this some modifications have to be done:
1. don't use a central data-store because
1.1 usage of this store increases with the number of hosts monitored.
1.2 the monitoring depends on a working network resource to the server
1.3 you add a single point of failure in your environment
2. add a local database to the zabbix agent (sqlite)
2.1 the agent only report values to the local database
2.2 no dependency to an external resource (network)
2.3 every node has the same database load
2.4 usually no correleation between data of different hosts
3. the zabbix server connects to the local zabbix agent
3.1 to retrieve data for someting like statistics
3.2 to send configuration changes
3.3 to provide a single interface to multiple agents
4. the zabbix client contact the server
4.1 to submit alarms, escalations or everything else need to transfer to the user
5. replace existing BMC Patrol installations
5.1 a step into an allready existing business

this are just the output of some ideas


mfg

Andreas Otto

[swaterhouse]

Here are my thoughts on your suggestions

#1 is now resolved by using distributed monitoring

#2 could be resolved by the agent simply logging things it cannot send to the server because of communication errors and sending all data when communication is restored.

One of the things I like about Zabbix is the agent footprint is VERY small (in disk space, ram and cpu) If you add db support (even sqllite) to that all three resources will go up dramatically.


#5 for Zabbix to replace systems like BMC there needs to be some stability refinements, some more features, but most importantly the documentation has to be massivly improved.