Foglight for DB's: Agent or Agentless? I can't decide?!

As a consultant with Quest Software, I am frequently architecting monitoring solutions for customers.

Part of this role is to build a solution which is both functional for the customer but also practical to deploy based on the constraints of the environment.

One of the most common questions I recieve is regarding the database monitoring capability: Should I go agentless or agent based?

 

My answer is "Yes". After a commonly confused look on the face across the table, I then explain...

To give you great flexibility, Foglight offers two integrated methods of monitoring databases, both of which are typically deployed to accomplish the full customer requirements:

  • Remote monitoring (Often using db calls direct from the Fogligth server out to the monitored databases)
  • Agent based monitoring (With agents sitting right on the database server for "Transactional workload" level monitoring)

 

In almost all database monitoring requirements, customers want to perform monitoring/alerting but also they would like to have an ability to do detailed historical diagnostics and analytics. These are two fairly distinct needs and not well accomplished by a single collection technology. Therefore, Quest maintains two collectors feeding a single integrated console giving you the flexiblity to deploy specific to the goals and constraints of your environment.

 

That said, here is a general recomendation I often make to determine what to deploy where: "Deploy agentelss monitoring to all databases and Agent based where you need it (typically critical production environments as well as Q/A testing environments)."

 

Here are some details behind that recomendation and why I make it:

 

Remote (Agentless) Monitoring:

  • This can and should be deployed to all databases in your enviornment.
  • Practical deployment using a web based, wizard driven deployment
  • Only requires a login/schema to use in the database (I often recomend creating a "Quest" user so monitoring is separated from your applications.
  • Collection rate is configurable but often set from 30 seconds to 5 minutes depending upon collection.
  • Great for detecting a problem and alerting you. It points the finger in the right direction and gives you the diagnostics to know what is occuring.
  • The trade-off of remote monitoring is granular detail due to a low sampling rate. Therefore when very granular detail is needed, the Agent is added...

 

Agent Based Monitoring (Also called "PA" or "Performance Analysis")

 

  • Due to a need to deploy an agent, this is often deployed to critical Production and all Q/A testing environments.
  • In smaller environments (30 and under DB's) the agent is often deployed to all databases as well.
  • Overhead is low due to a memory based sampling (it does not "trace" or "query" your database).
  • Requires local installation of agents. The only drawback here is if you have hundreds of databases, agent installs require more maintenance than remote collections.
  • Colleciton rate is very high (many samples per second) for excelent detail.
  • Overhead is low due to a memory based sampling (it does not "trace" or "query" your database).
  • Historical analysis is unmatched and offers full transactional level detail.

 

 

So long story short: Foglight gives you flexiblity when monitoring your database. Where other solutions try to use a single collection method (complete with limitations of that method), Foglight offers flexiblity, practicality, and full functionality for your DB monitoring requirements.

For further info, please check out www.quest.com/foglight

Anonymous