Cloud Based Network Monitoring Systems
John Willis has a blog post asking how to handle NMS in the cloud.
At my current gig, we have one piece of hardware that isn’t in the cloud — that runs our our wiki and our RCS. I chose to put our Zenoss NMS station in the cloud. It was a toss up between running it on an Ec2 or our ‘intranet’ system. I did this mainly so that I could have a separate machine and location for monitoring our intranet server.
It was a weird feeling to give up the traditional firewall/dmz/private network design. In the long run, it’s freeing. EC2 is a known quantity and I don’t have to maintain anything below the software stack. I have backups and procedures to restore if the instance ever fails.
At my last gig, which was a dotcom leftover, we were investing more and more into cloud computing — again, primarily EC2 and S3. We already had a well established NMS environment (nagios+cacti). We used that along with lots of wiki documentation to monitor instances.
Last I heard, they were constantly running between 10-20 EC2’s and about to do a major product launch. They were also using Alertsite for auditing performance.
I see the need for outsourced/cloud monitoring solution. There are plenty of outsourced monitoring services now, but I don’t think they are very cost effective. Alertsite costs just about as much per month per machine monitored as running an EC2 instance. It’s more if you want the bells and whistles.
There are two things I need:
- Traditional monitoring of things like network usage, disk space, processes, etc. This wouldn’t necessarily be limited to instances either. It should be capable of monitoring my existing data center as well.
- Monitoring of my more abstract services, like S3, hosted email and managed DNS. I’m imagining something like Friendfeed but for network services.
It’s tricky to get out of the firewall monitoring going. There’s definitely trust issues, along with the fact a lot of the critical systems are behind NAT.
Having a cloud monitoring service would be awesome though. It would make it easy to enforce SLA’s.
