Server virtualization is a good thing. It provides better utilization of resources, reduces server footprint and...
power and cooling in the data center and provides for more efficient management of the application environment. However, as with all things, too much of a good thing can be bad. The good news is that there are tools, technologies and strategies available to help right-size your server consolidation. In this tip, I will introduce you to some toolsets that will help you analyze your existing infrastructure and provide some tips for preventing over-consolidation on blade servers.
Tools of the trade
One of the best toolsets on the market is CiRBA's Data Center Intelligence software. It allows you to collect data over a period of time, across your entire server infrastructure looking at all aspects of server utilization, memory, network, disk, and processor usage. Once you have collected the data, a visual report is generated highlighting possible consolidation matches. This report can then have business rules layered on top to create what-if scenarios to better understand risk.
Another option is to get consulting services through IBM, HP or a VMware Authorized Consulting partner using VMware Capacity Planner. These services provide a full assessment of your infrastructure, much like the CiRBA toolset, but with a consultant running the tools and providing the assessment.
One tip on collecting data: No matter who does the assessment, I recommend data be collected for no less than two weeks, and a month would be better. If your business has specific "peak periods -- end-of-year financials, new product releases, tax season, etc. -- it is advisable to collect data during one or more of those peak periods to get an accurate representation of workload.
Virtualizing with blade servers
Since the advent of server virtualization for consolidation, system technologies have evolved to better accommodate and support a shared environment, such as larger memory sizes, multi-core multi-socket processors, etc.
One platform where changes have been made is in blade servers. Blade servers and virtualization go together beautifully. Blade servers provide more compute power per square foot of space than rack-mount servers and use less power per system than rack-mount servers. However, due to the fact that all power, cooling, external I/O -- network, SAN, NAS -- and cables are shared across all blades in a blade system, you have to be aware that by implementing server virtualization in this environment, you will be sharing already shared components. This is why it is so important to have an accurate characterization of your application environment, and consider using tools like the ones I mentioned earlier.
Although there may be issues consolidating in this already-consolidated environment, there are number of solutions. The first solution is to reduce the number of virtual servers running on each of the blades. Another option is to reconsider how you will implement your external networks, which can be implemented a number of ways.
One way is to use IP-based storage solutions -- such as iSCSI, SAN or NAS -- with high-performance accelerated network adapters. This works especially well if you have not already implemented a Fibre-Channel SAN or you are looking to consolidate the number of different fabrics in your IT environment. There are two 10 GB Ethernet adapters on the market from Chelsio and NetEffect, which can run all types of IP traffic through one adapter. This requires only one, dual-port high-bandwidth adapter per blade (rather than up to three), which connects into two (for redundancy) 10 GB switch modules in the back of the blade chassis, and allows plenty of headroom for all external networking. Another option is to upgrade your SAN infrastructure to interoperate with the newer 8 GB fibre channel adapters.This is a good option, especially if 8 GB fibre channel is in your future anyway.
Blades are not the only platform on which you can over-consolidate your servers. Rack servers are equally prone. However, because some of the bladed environment is already shared, it's important to understand the environment in which you are deploying consolidated servers, and to understand the solutions for mitigating bottlenecks.
About the Author: Anne Skamarock, Research Director at Focus Consulting, has been involved with computers and associated technology for nearly 30 years. For the past seven years, Anne has worked as a market analyst focusing on the convergence points around systems, storage, and software. Anne has published regular columns and tips and numerous business and technical white papers. She has recently finished her second book as co-author of Blades and Virtualization: Transforming Enterprise Computing While Cutting Costs.