logo

In the past, industry analysts have predicted that server virtualization would help customers reduce acquisition costs by 40% and administration costs by 20%.  And it probably has to date.  However, as processing densities continue to increase and allow greater numbers of applications to be consolidated on a single processing resource, two things happen. 1) Hundreds of physical servers will be hosting thousands of virtual machines 2) Availability and disaster recovery of the underlying infrastructure will be critical as more applications are vulnerable to a single hardware failure Both of these trends increase cost and complexity in the datacenter.  Don't get me wrong, I love virtual machine technology and the benefits that it brings to the datacenter.  However in this situation, standard servers as a platform for virtual machines increase I/O complexity and management costs. Now those servers require high availability, since more applications are vulnerable to a single hardware failure.  In order to migrate a virtual server automatically in the case of a failure, you need additional hardware resources and the network infrastructure setup correctly (cross-connect SANs/networks, permissively zoned SANs, etc.). For 100 physical servers, you now have 2400 components. According to virtual machine best practices, you should have 6 connections to the network (redundancy to multiple networks and failover) and 2 connections to storage (redundancy to shared storage). This translates into 2x the cost for network and storage infrastructure if you include the additional NICs/HBAs, Ethernet and fiberchannel cables and switch ports, as well as annual maintenance costs. In addition, you have to manage the I/O complexity and the different management domains for your physical and virtual servers. The cost alone for maintaining the infrastructure's reliability, availability and agility is enough to make my head hurt. Each of those 2400 physical parts can break, and each requires installation, testing, break/fix, and management of their independent lifecycles. What happens if something changes? Don't let TCO claims fool you, because TCO can be defined in many ways. Does TCO include all of your IT costs...both capital and operational? The only way to have a significant and meaningful impact on the total IT budget, and not just hardware capital costs, you need to reduce I/O complexity as well as server count. Solution? Virtualize the I/O Gartner predicts that virtual machines will embrace I/O virtualization as mainstream over the next three years. Why? 2400 components are reduced to less than 250, a 90% reduction. You do the math.