Nov 19th, 2009 |
Curt FranklinInterop Conference: The Impact of Cloud Computing on Network Management
A nice break for caffeine replenishment, and we’re into the next set of sessions. This one is already filling up, and the reason is fairly obvious: everyone wants to know just how cloud computing is going to hit their network. I’ve seen models and heard theories but I think that I, along with everyone in the room, would like to hear some facts to go along with all the theories.
Jim Metzler starts off saying that network performance monitoring is hard now, and will be an order of magnitude harder with cloud computing. This is especially true with public clouds, and that’s made more difficult because there are so many different definitions of cloud computing.
Jim says that, in the trio of “better, faster, cheaper” that cloud computing offers two out of three: it can be faster and cheaper, but it’s probably not better in any traditionally-measurable sense.
Now Steve Shalita is talking about the challenges of measuring network performance. He’s first running through the definitions and configurations of physical clouds (stratus, cirrus, etc.) and comparing those to IT clouds. He draws some good analogies boiling down to the idea that business dependence on clouds is growing along with their complexity. He defines clouds by the abstraction of services from infrastructure, with on-demand and at-scale delivery thrown in for good measure.
Looking at cloud transactions and the systems involved — a critical step in beginning to understand how to measure performance. There’s no question that it’s an extension of the problem of measuring performance and impact of applications in a virtualized environment — it’s just as important to understand what is being virtualized, and which segments of the network are being hit by each. Steve is talking about the necessity for visibility into the cloud — this may be the most significant issue in understanding what’s really going on with your application.
Another great session, and the room continues to fill, but there are more things at Interop for a roving blogger. More in a bit…
We continued the discussion with stimulating feedback from the audience. Although I couldn’t mention it at the time, CA was putting the final touches (including transferring the cash!) on our acquisition of NetQoS, a leading network performance and service delivery solutions provider. The acquisition was announced a few hours later. This validated many of Steve’s points, and results in an offering for the first time consisting of best-in-class flow, packet analysis, SNMP management tied in with model-based root cause analysis and end-user transaction monitoring for the network, server (especially virtual server), database and application layers. Infrastucture that’s application aware and management that starts with the end user, but goes down to the packet - all to assure service delivery.
When coupled with automation for heterogeneous virtual and physical platforms (VMware, hyper-V, XenServer, Solaris Zones, IBM LPARs), the holistic management needed to unlock the potential of the cloud, including transitioning from internal private clouds to public cloud services like Amazon Web Services, is within reach. Exciting times indeed.