The problem with many existing tools is that monitoring solutions that are saved with backup images or configuration scripts are difficult to update and test. You don’t really want to mess with your deployment process and verifying it’s working after a change is also a pain.

Our approach with Scout: a single crontab entry that loads a monitoring profile from our web interface at scoutapp.com. This profile can be changed at any time and won’t break a deployment. It’s super light-weight (just 1 line of code) and you can tailor monitoring without fear.

More info on our approach: http://bit.ly/PJW4o

I have been compiling a list of open source monitoring systems, tools and supporting what-nots here: http://www.heliosdev.org/open-source

Cheers.

I think what is different with “cloud monitoring” as an approach rather than a technology is that the cost model (how one is charged for usage) does not necessarily reflect the application resource model (i.e. cpu, memory) typical of traditional system management models. The management model can consist of multi meters some used in the calculation of the monthly billing (vendor rate plans) and others used for standard service level management.

What I really like about cloud computing is that there is a much stronger correlation of cost with resource usage and more than likely performance. Performance is not just some timing statistic used in service level management reporting but an actual cost that appears on a balance sheet. We now just need to start placing a value on activities (fine grain and not just a large service label) and calculating the real/virtual profits of such activity.

William

We are in violent agreement!

That is still just domain-specific application-level monitoring, to which the “cloud” is largely irrelevant. You can track anything by determining metrics and monitoring them, that doesn’t make monitoring the cloud anything new. It also doesn’t make them pointless, that monitoring is important.

It’s just a matter of nomenclature; the cloud is just another word for distributed resources that we’ve had in one way or another for many years. Now that it’s a service that you lease from another company over the Internet, it doesn’t deserve it’s own category, in my opinion. The issues involved in monitoring the cloud have always existed since we’ve had networks.

“Cloud Monitoring” looks a whole lot like, for example, running BMC Patrol and the BMC agent on your distributed servers in an intranet doing application-level performance monitoring (which I did 10 years ago at State Farm).

Calling it something new only helps marketing folks associate themselves with trends, and blogs to get hits. And here I am, helping validate that for John. *waves* (grin)

I really don’t mind, perhaps it’s useful to group a set of common use cases together as a “cloud,” but it’s not like it’s some kind of (cough) paradigm shift. Let’s just call a spade a spade.