IT Management and Cloud Blog

« | Home | »

Will Network Management Be the First Causality of the Cloud?

By John | March 19, 2008

A while back, I was having one of my regular lunches with the world famous BSM blogger Doug McClure, and I was going on about cloud this and cloud that. I mentioned at the time that I wanted to install one of the open source monitoring tools on EC2 and start figuring out what metrics make sense and what ones don’t. I also said that I might install OpenNMS as well, and Doug’s response to installing OpenNMS on a cloud was why. Since then, I hadn’t given much thought to the aforementioned question, but, today, I am meeting with an old brick and mortar IT customer that wants me to talk about cloud technology. This customer has a heavy investment in NMS. So, now I have about an hour to re-think this. I think that I agree with Doug, but I need to put some more thought into it. All suggestions welcome.

Does network management have a play in a cloud?

Topics: amazon, aws, cloud computing, ec2, s3 | 10 Comments »

10 Responses to “Will Network Management Be the First Causality of the Cloud?”

  1. Jack @ Tech Teapot Says:
    March 19th, 2008 at 9:09 am

    Personally I don’t think so. Networks provide services, NMS test the services to make sure they’re working. However I choose to deliver the services I still need to ensure that the service is being delivered. Unless you’re of the opinion that the cloud will be so reliable as to make service oriented monitoring to be irrelevant?

  2. The Cult of Gary » Blog Archive » Cloud Based Network Monitoring Systems Says:
    March 19th, 2008 at 9:50 am

    [...] Willis has a blog post asking how to handle NMS in the [...]

  3. Matthew Small Says:
    March 19th, 2008 at 11:35 am

    We use monitoring daemons to collect process and hardware specific metrics from EC2 instances. How else are you going to make decisions to scale? “I see that my application servers are loaded down, let’s launch a new one.” “I see that my MySQL threads are steadily increasing over time, time to scale up to a larger DB instance.” Of course you can automate various tasks based on those monitoring results, such as scaling, running scripts, or relaunching instances…so maybe network managers are on the way out, but you will always need network management.

  4. John Says:
    March 19th, 2008 at 12:26 pm

    Jack,

    No… trust me I don’t trust the cloud. I should have been more specific. What I mean by NMS is the purely Network management metrics. The secret sauce metrics that you find in products like HPOV, Netview, and OpenNMS. I totally believe you will still need OS monitoring metrics and also transaction, and application monitoring. Although HPOV, Netview, OpenNMS, … all have overlap with the pure play OS monitors when I say NMS I really truly mean network metrics in the context of this question. Things like interface/switch router, packets, … kind of stuff, are those kind of things necessary.
    Maybe a better question is will you need products like Netview HPOV,or OpenNMS ins the cloud and my guess is no.

  5. John Says:
    March 19th, 2008 at 12:29 pm

    Matt,

    Totally agree. we will definitly needs to look at all the traditional OS monitoring metrics (processes, CPU, memory, file/io, disk space (maybe). Also transactional and application monitoring IMHO, is hear to stay. Like I said above , I am just thinking the pure network metrics are more likely not needed anymore. I love you gus feedback and keep it coming if you dis-agree with me.

    Thanks

  6. Matthew Small Says:
    March 19th, 2008 at 4:49 pm

    I think that it depends on your application. Some indicators may work well for some applications, where others don’t. I think it also entirely depends on the user. Some users are going to just want to know what’s going on. Urchins for data.

    Here is a sample alert template from RS.

    high network tx activity
    if interface/if_octets-eth0.tx > ’50000000′ for 30 min then escalate to ‘critical’.

    Escalations are the actions that are defined to be taken.

  7. Jack @ Tech Teapot Says:
    March 20th, 2008 at 5:44 am

    @John – the bits and bytes network metrics will become a bit irrelevant simply because you won’t have a network upon which your services are running — the cloud provider has that. So one would hope that they will do the network metrics but of course they are aggregating their infrastructure over a large number of customers so I suspect that the net effect will be a reduction in demand for low level network metrics. But service level monitoring will still be very important. Service monitoring will clearly be seperated from the low level network metrics.

  8. John Says:
    March 20th, 2008 at 5:59 am

    Jack,

    Great feedback. That was basically my thoughts as well. However, it looks like Matthew has found some useful metrics for their RigthScale product. Monitoring in the cloud is going to be a very interesting topic in the upcoming years.

    Thanks
    John

  9. David Says:
    March 20th, 2008 at 8:00 am

    You may have agreements in place for service levels where round-trip transaction monitors should still apply. OpenNMS has began developing monitors, such as the MailTransportMonitor, that monitor the “end-to-end” availability and the performance of a service… in this case the of sending messages through the network (cloud). This particular service monitor tests the availability and performance without actually attempting to monitor/measure the details of the underlying network technology. A job best served by element managers that the cloud provider will certainly have in place. Continuing to collect metrics from local resources on each side of the cloud makes sense as well as receiving messages from the cloud provider’s monitoring systems for use in correlating the end-to-end service outages is something to be considered.

  10. John Says:
    March 20th, 2008 at 8:06 am

    David,

    Some good points. This is what I was looking for is a starting dialog on this topic.

    BTW, I was hoping Tarus would have come over here screaming and yelling at me – Wasup :)

    John

Comments