Server consolidation is one of the primary drivers of companies looking to adopt virtualization. The advantages to consolidating more workloads on fewer servers are obvious (reduced hardware cost, easier management and better resource use), but there are still many difficult questions and potential pitfalls when you're developing a server consolidation plan. Let's take a look at some of the most common mistakes related to developing a plan for server consolidation.
Focusing too much on consolidation ratios
We measure server consolidation in terms of a consolidation ratio -- the number of virtual machines per physical server. Consolidation ratios will vary depending on the capacity of physical servers and the types of workloads you're looking to virtualize. The important point to remember is that while a higher ratio may be possible, it may not be best for the business. Simply pushing for the highest possible consolidation ratios will leave servers overtaxed or unable to restart workloads from failed servers in the event of a disaster. Ultimately, identifying your ideal server consolidation ratio is a balancing act.
When it comes to consolidation, you can have too much of a good thing. Increasing your consolidation levels to the point where you do not have enough spare capacity to handle a server failure can leave your data center crippled while you scramble to provision a new server or make repairs. Over-consolidation can limit your ability to migrate virtual machines (VMs) and balance workloads. Over-consolidation can also hurt application performance, which can slow as VMs compete for limited resources.
Getting complacent with your consolidation levels
It's common for organizations to roll out virtualization in a tiered approach, starting with test and development systems before moving onto production and mission-critical workloads. However, if you take the first step and virtualize a handful of servers but never re-evaluate your server consolidation plan, you're missing out. One of the simplest ways to increase your consolidation level is to replace older servers with new, higher-capacity servers that can handle more workloads. If you're not ready or don't have the budget to completely replace a server, focus on upgrading memory, which will give you the biggest bang for your buck. If you simply don't have any money to spend, re-evaluating your VMs' resource allocation levels and setting limits can free up enough resources to add additional VMs.
Forgetting to address VM sprawl
Jumping into virtualization without a strategy for controlling VM sprawl can quickly lead to problems. The reason so many people fall victim to VM sprawl is that it takes months or years to develop -- quietly building until it manifests in a big problem. Fortunately, you can prevent sprawl before it happens by making VM lifecycle management a part of your initial server consolidation plan.
Failing to consider failover and migration needs
There are two basic approaches to designing a server consolidation plan: leaving spare capacity on each server or having spare servers standing by when a failover or migration is needed. Both approaches will work, but it really comes down to what your failover and migration needs are. Leaving spare capacity on each server creates the flexibility to live-migrate workloads in order to balance resource use or update a server without bringing a new system online. However, maximizing consolidation and keeping spare servers for failover needs can often mean better resource utilization. You also need to consider your workload when developing your server consolidation plan. If you have a very demanding application, it may make more sense to have a spare server standing by rather than leave a large percentage of a running server go unused.