The chart above, courtesy of Cloud Spectator, illustrates quantitatively the points made in my last post (part 10: bare metal cloud) regarding the reason that physical servers continue to be so pervasive within data centers. As you can see from this – and as you would expect given the nature of hypervisor-based cloud servers – database reads and writes perform significantly better on bare metal cloud servers.
The performance advantages of non-hypervisor-based servers within the data center in no way negates the advantages that hypervisor-based servers, whether implemented with VMware vSphere, Microsoft Hyper-V, Citrix XenServer, Redhat KVM/QEMU, or other hypervisors, bring to bear regarding automation and some forms of agility and adaptability. But what has been occurring in many real data centers is an increased tendency for the acceleration of virtualization penetration to decline over time. This has nothing to do with the merits of virtualization – which are profound and many – but instead have to do with the inertia of some types of workloads coupled with the performance and scalability needed by some applications.
Obviously non-hypervisor-based servers (bare metal cloud servers) aren’t the next big thing – instead these point to the needs of data center staff that virtualization isn’t yet fulfilling. More on that in my next post in this series.
This is part 11 of an on-going series. Part 1: virtualization isn’t the next big thing (NBT) because it was the last big thing (LBT); part 2: data center IP traffic growth; part 3: data center IP traffic sources; part 4: cloud workloads; part 5: large data centers and administrator to server ratios; part 6: strategy overview – adapt or be crushed; part 7: automation, agility, adaptability; part 8, automation vs agility and adaptability; part 9: virtualizing everything; part 10: bare metal cloud.