What is the viability of taking a mix and match approach to cloud hardware and software components?
A recent question came up about the viability of implementing a cloud architecture using a mix of Cisco routers, NetApp block storage, VBlock hardware and the cloud.com stack. The short answer is that while none of these systems are officially vendor certified to work together, that does not mean that they will not integrate perfectly well. Due to the dearth of cloud standards and lack of cross vendor partnerships, the only real way to find out is to test the systems in the data center. The bigger question for the enterprise is how much an organization is willing and able to take on a DIY cloud architecture integration project rather than relying on certifications and reference architectures with the notion that it will reduce (or at the very least, shift) business and technology risk back to the vendors. Since enterprise cloud products are so new, I would argue that purchasing vendor certified cloud infrastructure products reduces enterprise business risk only slightly if at all. The vendors are only willing to certify their systems to the limits of their own internal testing and no further.
In an ideal world, all cloud components in the infrastructure, hardware and software stack, would be completely interchangeable, totally transparent to the user and application agnostic. That is certainly the intended future goal, but sorry to disappoint you folks, we are not quite there yet. As I mentioned in an earlier blog on the topic of hardware, for truly high performance cloud environments, hardware does in fact matter at cloud scales. Systems that seem to work together might not optimally configured, so decisions need to be made about relative cost and how much improving efficiency is worth the investment in resources. If the intention is to roll out a large commercial cloud service, then yes, that investment is clearly worthwhile. If the intent is to build ten racks of internal dev/test, then more likely the answer will be no.
If the IT skills are missing in the enterprise and/or there is a desire is to pursue a vender certified approach to building an enterprise cloud there are several options. There are hardware oriented stacks such as VBlock from VCE and FlexPod from NetApp. Unsurprisingly, given that one of the major investors is VMware, VCE certainly encourages customers to use VMware as the hypervisor and cloud management tools of choice, but there is nothing preventing a company from exploring other options. For folks that are comfortable with open source solutions or looking for alternatives to VMware, CloudBlock from Cloudscaling, or Cloud Private Edition from Rackspace are both Openstack implementations that come with a recommended list of tested hardware and a set of management tools. HP VirtualSystem and Dell vStart also are bundles that are a mix of storage hardware and management tools worth looking at particularly for companies that already have strong HP or Dell relationships. All of these solutions are going to come with higher upfront costs, but do have the potential for being cheaper to deploy and manage than a roll-your-own option.
In the end, since often the most common reason for deploying a cloud in the enterprise is for efficiency and cost savings, each company needs to evaluate their capacity for making the trade off between reducing upfront costs and paying for the privilege of shifting risk to the vendor.