Evaluating Clouds

When evaluating clouds there are three major criteria that are used first before we get into the nitty gritty of how the system works. The criteria is not always in writing but it is in the back of everyone’s mind and in many cases are nebulous to define. Herein, I will try to look at those criteria in order to aid others in making the same decisions I have had to make lately while evaluating clouds.

There are literally 100s of cloud providers promising you all sorts of business, process, and computing improvements. However, those improvements are meaningless if you cannot get through the initial conversation with out saying, “I think not”, and moving on to the next cloud provider. In many cases, the real criteria for picking a cloud is not even written down, it is in the back of the evaluators head, percolating to come forth. The initial conversation with any cloud provider should not violate what you have written down about your needs, such as those concerning security, data protection, management requirements, or even how data will move to the cloud.

Our requirements were fairly simple we are moving to the cloud because our data center is moving. Once the data center moves, we may or may not ‘stay’ in a cloud based on cost, functionality, ease of management, and problems encountered during the move. In essence, we want things to just work with very little effort on our part and if all that goes well, then we may stay there. We really could care less which hypervisor is in use, but at the same time, we are concerned about getting our data back again.

Let’s look at four types of vendors and how some conversations have panned out with respect to those basic requirements, yes we have more, we need to protect our data, have a good SLA, etc.

  • Borrowed Resources – I have a friend who had some virtualization resources he would let me borrow. Cost being an issue and the short term nature of the move, this is the way I went when we last moved the data center. Unfortunately, their current resource usage would not allow us to put the workloads into their virtual environment without compromising performance. In essence, they were at capacity. This was not the way to go, our performance requirements would be exceeded. A no-go for us.
  • Inexpensive vCloud – We run a 100% virtual and cloud environment using VMware vCloud Suite and we were looking at one of the more inexpensive clouds out there. Upon calling, we encountered our first problem, cost. They did not meet their own response time goals, which was to keep to time schedules the vendor set for return calls. This is one of those nebulous response time requirements as they are set by the vendor not us. If the vendor says they will call you back within 24 hours, well I expect them to call back within 24 hours.  There was no easy to use order form, there was a requirement to be call a sales associate. This sales associate could not answer our SLA questions. I would assume they did not want our business, as they did not respond with a call back. So I must also assume they have no valid SLA, have horrendous response time for problems that may occur, and would not be a good candidate going forward.
  • Expensive vCloud – We looked at an expensive vCloud option. Our first call to them took the form that was disastrous. They wanted us to provide all sorts of documents related to our business so that they could design a proper cloud for us and when pressed for a 4 VM cloud with certain memory, CPU, and network requirements, explained they would like to become our partner and as such they need more information.  In a very real sense, this threw up several red flags. The first was the way they wanted to know intimate details of our business and the second was the term partner, in a very real sense it sounded they wanted me to sign over rights to our business. This first conversation was invasive to say the least. When they were pressed for per VM pricing their prices were quoted at roughly $800.00 per VM which exceeded our budget by a mile.

We revisited this option later in our evaluating clouds process and because we knew someone who worked for the company got a more reasoned answer for pricing and functionality.  The new price was NOT outside the bounds of our budget, so they once more were in the running. Unfortunately, we still worried about a well defined SLA and getting our workloads into the cloud.

  • Fully Automated Cloud – We looked at a fully automated cloud as well, this cloud automated not only ordering but had an easy to use portal. Actually, this cloud ended up being the most affordable, with the most options for security, data protection, and a well defined SLA including how to seek credits for their failure(s). They promise workloads in 20 minutes or less and it usually took less. There was no need to talk to anyone except to confirm the order which happened just as quickly. How did I come across this option? I saw them at a trade show and a friend used them for his business and I got a look at their user interface with some real world usage.

So what are these nebulous criteria we have for Evaluating Clouds?

  • Easy to understand conversations: In general, those evaluating clouds who have gotten to the point of actually making phone calls are looking for final details about SLAs, pricing, etc. They are not looking for the cloud provider to be their ‘business partner’, they are not looking for the cloud provider to call them back at some nebulous time in the future.
  • Response Time: When a potential cloud tenant calls to get information and the sales associate does not know it, and a time for callback occurs, we expect those times to be met. Response time of sales impacts response time of support. Yes, it is understood that they are different groups, but if one cannot meet their dictated response time, how can potential tenants expect others to meet their response time within the same company.
  • It is all about cost. Generally, the first conversation potential cloud tenants have is about cost. Be able to answer that question without knowing the details of what is going to be run. If a tenant has the desire to run a LAMP stack, the only other real details that may be necessary is the number of visitors per day and the type of data served up (video vs text for example). Or better yet, have a pricing calculator on the site.
  • Cloud providers should know whom they are talking to as well. Listen to the potential cloud tenant and if necessary do a wee bit of web research to determine who the person is. LinkedIn and other social media sites will help here. Do not talk down to those who have the background and knowledge in the cloud space.

Some potential cloud tenants require a lot of hand holding, others require none at all. For those who have the knowledge and a good IT staff a fully automated cloud may be the way to go, while those who have no knowledge a provider who does it all for you may be the best way to go.  In our ongoing search in picking a cloud provider and evaluating clouds, we wanted more control over our own environment.

Which way did we go? Well we have not gone there yet, but are moving in that direction. But I will say, we like automation.

Posted in SDDC & Hybrid CloudTagged , , , , ,