High Availability: Virtualization Storage & Resources

By February 28, 2013High Availability, Storage

virtual storageThere are lots of pieces that fit into the puzzle of achieving high availability. Check out last week’s post on virtualization configuration.

Shared Storage Access

When building a highly available virtualized infrastructure – no matter the hypervisor – you are going to need shared storage.  This is your storage area network (SAN). If you want to take advantage of technologies to move guests, or VMs, between hosts – either manually or automated – then this is a no brainer. But, how are you going to connect everything to the SAN? If you recall our post about environmental factors, we discussed multiple power supplies and power outlets and UPS, etc.  Take those same ideas and bring them to your shared storage.

The SAN is the commodity here (not utility power) – and you do not want single points of failure. You will want multiple Internet Small Computer System Interface (iSCSI) or Fibre Channel ports. These ports should be on multiple host bus adaptors (HBAs) or multiple NICs.  In fact, best practices dictate breaking your iSCSI paths into isolated/separate iSCSI network segments so multipath I/O (MPIO) is truly redundant and functional.

Virtual Hosts

If you are building a virtual environment to be highly available, you need to think “N+1”.  N is the amount of resources you need and best practice dictates that you want +1 more of that.  Think about it: if you are building a highly available environment, part of that is the flexibility to be able to perform maintenance.

If you are fully “utilized” on your hosts, you cannot move workloads from host A to B in order to perform maintenance on A. You want N+1 so you can migrate workloads, free up A to do your maintenance, and then migrate workloads back.  Keeping N+1 in mind, remember that if you have one host in your cluster with twice the number of cores and 3x as much RAM as the other hosts, if that node dies, will the others be able to sustain it?  You have to always consider the worst case failure scenario.

As we delve deeper into the considerations that highly available environments require, we are going to get more technical. However, that doesn’t mean that these practices and behaviors aren’t important. If you don’t have someone on your payroll who knows what we’re talking about, call us today!