Cloud for All


Building and operating a private cloud is a complex undertaking. Most cloud platforms are designed to play well with thousands of physical servers. This is great for public cloud providers and extremely large enterprise organizations. However, smaller organizations that need a cloud built from tens of physical servers can find these platforms challenging. I’ve written about the possibility that some of these customers might get what they want without a cloud platform. But what if a cloud platform were easy to deploy? If you cloud deploy an OpenStack cloud in one day, would that help? This is one target for the Intel Cloud for All program.

I had the pleasure of attending the Intel Cloud Day 2016 event in San Francisco at the end of March. The program has a much wider purpose and has sub-programs for the different types of Intel partners. The overarching objective is to enable the deployment of tens of thousands of clouds. I read this as a pretty clear endorsement of the idea of the hybrid cloud. Some of these clouds will be public cloud, but the majority are going to be private clouds. At the Intel event, we had a variety of end-user organization stories. The one I was most impressed with was Quinton Anderson from the Commonwealth Bank of Australia (CBA). I expect banks to have a very sedate pace of change and a lot of risk aversion. The CBA has deployed a cloud with a full continuous integration and continuous deployment (CI/CD) for part of its IT. This allows the bank to make microchanges to applications every day. A developer’s changes can be in production in three hours, improving customers’ experience with the bank and empowering the bank’s developers. This is the real payoff for cloud methodologies: rapid innovation and developer enablement.

The ability to deploy a cloud in a day is great. But what about all the other days? We are seeing a lot of products coming to market that talk about how easy they are to deploy and how quickly you can start using them. I am a little skeptical about this trend. It is easy to build technical debt into a solution just to make it easy to deploy. Operating the system day-to-day is often far costlier than deployment. My concern is that decisions that are made to speed deployment will then complicate operations. To be clear, I haven’t heard that any of the easy-to-deploy products are hard to operate. In fact, quite the opposite: many of the infrastructure products that have come to market in the last few years are extremely easy to manage. I just want to raise the flag that “easy-to-deploy” isn’t the end game. Any infrastructure you deploy needs to deliver value over its lifetime.

While I have my cynical hat on, what about hardware? These clouds that can be deployed in a day need hardware to run on. The process to order, fulfill the order, and rack up the servers will take a bit more than a day. It’s likely to be weeks or months between deciding to deploy a cloud and having all the hardware in place. Does it really make a difference if the cloud can be deployed in one day or five? The major difference is the cost of the labor to deploy the cloud platform. Five cloud platform developers for five days, or five weeks, is quite a lot of money. One systems engineer for one day is cheap by comparison. One of the things I see in cloud providers is an expectation that people will be expensive. As such, they try to minimize the amount of human activity. This is the opposite of how enterprise organizations operate. Enterprises treat hardware and software as expensive and use a much greater number of manual processes. There are probably large savings to be had in enterprise IT by treating people as an expensive resource and automation as a way to reduce this cost.

I also like to examine motivations. Why does Intel want tens of thousands of clouds? This one is quite easy: to sell millions of CPUs. I’m pretty comfortable with this motivation. In the end, the CPUs will only be sold if there is a business benefit to the customer. Intel is taking the long view of enabling that business value. It’s nice that they are looking beyond this quarter’s profitability, but not surprising. Designing a CPU is not a short-term project: the payoff time for building a silicon fabrication plant is very long. Helping to develop the market for your product will help ensure the payoff happens.

Share this Article:

The following two tabs change content below.
Alastair Cooke
Alastair Cooke is an independent analyst and consultant working with virtualization and datacenter technologies. Alastair spent eight years delivering training for HP and VMware as well as providing implementation services for their technologies. Alastair is able to create a storied communication that helps partners and customers understand complex technologies. Alastair is known in the VMware community for contributions to the vBrownBag podcast and for the AutoLab, which automates the deployment of a nested vSphere training lab.
Alastair Cooke

Latest posts by Alastair Cooke (see all)

Related Posts:

Leave a Reply

Be the First to Comment!