By Koika (Own work) [CC BY-SA 1.0 (http://creativecommons.org/licenses/by-sa/1.0)], via Wikimedia CommonsBefore we go into virtual datacenter problems, let's talk about town planning for a bit.
I like the analogy of a town to describe a datacenter, with the town citizens as the users or owners of servers. And there are apartments which are shared housing for multiple families.
Why is town planning important? Because without governance, anyone can build a house, and so houses get built like crazy. This is a case of sprawl, or unorganized development.
While immediate needs are surely met, the problem with unorganized development is that it leads to larger maintenance efforts and usually leads to chaotic environments.
Also, houses need power and water - resources that are typically not available in abundance and have to be used with discretion.
There's also varied plans between lots, so understanding each lot and its plans is important to providing the right amount of resources and ensure that the town's needs are met.
And there's seasonal variations, like holidays when an influx of people and activity are expected. At these times, the town planning council must ensure that there's adequate additional resources - but at the same time this would not be a permanent need.
'Nuff said. I rest my case here. You can see how much the same rules applies to virtual datacenters, VMs, owners and tenants, line of business plans, capacity planning, etc.
So governance (like town planning) is a real need. How much of this is possible, and where do you start?
First of all, do you have a visual representation of your virtual datacenter, spanning multiple virtualization domains, if needed? if you have VMware and KVM, then a tool that gives you a VMware vSphere overview alone is not enough.
With the visual representation are you sure that all your elements are obtained - crucial things like orphaned VM disks (disks whose parent VMs are not present) and unused VM snapshots, or even low-utilization VMs should not be missed? Are you aware of the heavy resource users and categorization / grouping of these VMs?
This Wikipedia and Wikimedia Commons image is from the user Chris 73 and is freely available at //commons.wikimedia.org/wiki/File:WaterSavingToiletJapan.jpg under the creative commons cc-by-sa 3.0 license.Are you able to look into what's going on inside a VM guest OS? This might give a good symptom of what's causing high resource usage - like that apartment tenant who let the water running and went to work, and ended up depriving everyone else of water till evening.
Note that this is "shared resource" world - so it is important that you ensure that there's adequate capacity to meet needs when multiple VM's resource usage peaks together. The answer is not always a DRS load-balancing type movement - which is a reaction, not a 'planned' step. There are several circumstances where a VM online movement failed or took long amount of time causing application disruptions. Finally, if there's not enough capacity, simultaneous movement of large VMs (32GB memory, lets say) through a shared network switch can cause a high data flow causing a potential outage on the switch.
You see so far that these demands are very different from traditional monitoring.
You need to use a virtualization monitoring tool in order to monitor VMs and the hypervisor hosts better. The traditional agent - based or network polling type approaches are still needed, but change is needed in the monitoring approaches.
For instance, if an application on a VM is runing slow, with the traditional approaches we would monitor where the app runs and this still needs to be done for checking logs, etc, but in a virtualized datacenter, this would not be enough. One would need to monitor the hypervisor host and other VMs on the same host where this VM / app is running and find out where the latency is creeping in from.
Do note that while the monitoring approach is enhanced, you need the other parts like operations management, event correlation, trouble ticketing. In this context the operations model containing the CIs, services and their relations becomes even more important.
So, let's talk about the solution for this.
HPE's Cloud Optimizer is a great tool to monitor your virtual environments as well as private and public clouds.
Monitor multiple virtualization technologies in a standardized and user-friendly manner.
Use guest OS drilldown to look into what's happening in each VM guest OS.
Visualize your whole virtual environment easily and view performance / usage trends for each entity in a cool drag-drop interface.
Integrate with HPE Operation Bridge and Cloud Suites, as well as with HPE OneView.
Plan capacity needs, perform "what-if" simulations, predict future usage.
Provide placement suggestions, usage measurement for billing, and right-sizing with optimization guidance.
Plan infrastructure needs from business metrics via the business metric analyzer.
Cloud Optimizer is available for free and full-function evaluation for 60 days, and permanently free for managing a small environment of less than 25 instances (hosts + VMs). T
Try it out and give us your feedback on how it helps you manage your "town" (datacenter) better.
Here's a great overview video of Cloud Optimizer's capabilities and functions.
You can also see demonstrations and find out more features of HPE Cloud Optimizer and related solutions in our sessions and at booths during HPE Discover Las Vegas, June 7 - 9. Click on the image below orhereto register!
Ramkumar Devanathan (twitter: @rdevanathan) is Product Manager for HPE Cloud Optimizer (formerly vPV). He was previously a member of the IOM-Customer Assist Team (CAT) providing technical assistance to HP Software pre-sales and support teams with Operations Management products including vPV, SHO, VISPI. He has experience of more than 14 years in this product line, working in various roles ranging from developer to product architect.