The Silicon Valley Cloud Computing Group kicked off a series of Infrastructure as a Service (IaaS) meetings last night with a presentation about CloudStack. CloudStack was recently relicensed by Citrix under the Apache Software License and is a public and private cloud provisioning and management stack, akin to OpenStack. One of its main benefits is it is both hypervisor agnostic (XenServer, VMware, Oracle VM, KVM, bare metal) and storage agnostic (local disk, iSCSI, Fiber Channel, NFS, Swift).
However, one of the main takeaways for me was the distinction between the usage of cloud compute resources based on the underlying stack. VMware vCenter/vCloud was portrayed as being best suited for traditional enterprise apps & client-server computing where the environment may have hundreds of hosts and in which applications (VMs) assume reliability thanks to, for example, vCenter’s fault tolerance features (e.g., vMotion). In contrast, CloudStack and other similar products/projects are best suited for big data, massive scale, next generation apps which scale out to environments with thousands of hosts running applications that assume failure such that loss of multiple resources doesn’t interfere with the operation of the overall application.
It is an interesting distinction and I wonder how accurate it is. If it is accurate, it is another point to consider for enterprises deciding on the technology to use to build out their private cloud infrastructure. If you have an observation either way or want to share how your enterprise has built out virtualization and/or a private cloud, add a comment to this blog entry to continue the discussion.
