This blog entry was first published on February 18th, 2015
I like to stay updated on what is published about cloud, and this week, an article in TechRepublic came to my attention: Private Clouds, a very public failure. My initial reaction was, really? Sure, Amazon Web Services will tell you the private cloud is doomed, but would you expect anything else? But beyond that statement, what are we really talking about? Before starting the discussion, let me point out I’m talking from my own experience. The only thing I can say is that, in the environments I work in, the private cloud is well established and growing fast. Let me highlight why.
Why do companies go to cloud?
Fundamentally there are two reasons for which companies go journey to the cloud. They need to reduce costs of operations, and they need more flexibility & responsiveness for some workloads, particularly the ones related to customer interaction and using mobile based interactions. How do they achieve those goals?
Small companies want to get rid of what many consider the “IT headache” all together. So they consume cloud services provided by a third party. In doing so they belive they’ll pay less and will get a better service. Frankly that remains to be seen. In the US, using third party cloud services often equates to AWS. It is less so outside the US for a variety of reasons having to do with security, compliance and the feeling of lack of control in a public cloud environment.
Larger companies are confronted with a different environment. They typically have a proper IT department that is well managed. However, the use of their infrastructure is far from optimal. Commonly, data center infrastructure is used between 8 and 15% of the time, depending on how you measure it. The time it takes to get new equipment results in over-provisioning. Virtualization has been used to improve the situation, and for most workload, cloud is only the next step after virtualization. It is the automation and provisioning elements that are key to improve the utilization levels. Utilizations of 30-40% are not uncommon in such environments. Now, if you do the math, this reduces the infrastructure (server) required by 50%, resulting not only in the need for less hardware, but also in energy and space reductions. Side benefits include better management of license keys, possibilities to charge individual departments and giving the end-user the opportunity to provision their infrastructure and applications without IT interaction. The next step in this process is DevOps, the automation of the hand-over between development and operations.
Which type of cloud, to achieve which purpose?
From my comments above, we can agree that companies are starting to use cloud technologies so they reduce the cost of operations, allowing them to work with smaller budgets while making funds available for innovation and transformation. However, the next question is: which type of cloud to use? This is where public versus private cloud comes to play. Interestingly enough, as far as I can see, it is not one or the other, but rather one and the other. Thus, here is where I differ from the article: where small companies look at ways to remove the IT burden within their own organization, larger enterprises look at streamlining it and position the right workload in the right place.
Key production workloads should go to private cloud. Some customers are adamant that the SAP environment can only move to a private cloud. The reason? Mainly security and making sure the functionality and data remains under their full control. You may feel this to be old fashion. Well then, many companies are old-fashioned. Development and test environments may go to a multi-tenant environment managed by a third party, but definitely not a public cloud. Here again, security concerns play an important role. We will come back to that.
While we are on the topic of what types of cloud should be used, you will probably tell me SalesForce and WorkDay are demonstrating production workloads that can be located in public cloud, and you are right. May I just point out to you that these are two very specific areas, CRM and HR. You do not find the same success in the ERP, Finance or PLM space. Why? Companies use the latter to differentiate themselves from their competitors or consider the information to be too competitive. Interesting, isn’t it?
This is why the world will be geared largely toward hybrid cloud. Companies will use a variety of environments to address their requirements from a cost and agility perspective. This actually makes one technology critically important: that of the cloud broker. You do not want your users to have to understand the complexity of your deployment. You want to shield that from them. Using a single portal, an integrated service catalog and orchestration software to link the user to the appropriate environment is what they expect. That is the cloud environment companies will build.
Why are private clouds perceived to fail?
A recent Gartner study on the topic is very interesting. First, the question is not why they fail, but rather what the problems of implementation are. The first failure reason is “failure to change the operational model”. If you have followed me for a while, you will know this is a point I did highlight regularly. Cloud is all about service delivery, and many IT departments are still technology focused. It is not the same thing. But actually this does not just apply to the private cloud, but to all usage of cloud. Most public cloud consumption is done outside the IT department where users have become acquainted to consuming services on a day to day basis from their mobile phones and tablets.
IT often aims to understand and dictate the technologies used. How often have I heard customers telling me they will only use one of our cloud services if it uses product XYZ to secure or manage the environment. If you consume a service, you should assess whether the service addresses your requirements, NOT how those are addressed. Move from a technology to a services mindset.
I agree that this is difficult for an IT department. It is definitely a barrier to the use of cloud, but why is it considered a barrier to private cloud and not to public cloud, frankly that remains a mystery to me.
The second failure reason is “too little too late”. Cloud is indeed a big bang. You have to change your hardware/software environment, your operations & management, your governance and your key processes. That’s quite a transformation. And what’s more- you need to do it all together. Many accounts only address one or a few of these. The result is a half-way house that does not operate correctly.
Those two already represent more than 50% of the reasons. If you add to that the not invented here syndrome that makes large IT department create their own cloud technology stack rather than using a proven one, and resistance to change you have most of the problems the Gartner study highlights.
How to approach it then?
Start small and surround yourself with people that did it before. Choose a proven software stack. Don’t re-invent the wheel. Do a proof of concept to understand the technology and its implications to your organization. A good place to start is development & test. Review your operations & processes. A good way to do that is to embrace the IT4IT reference architecture as it was built with cloud in mind. Choose a small team within your organization to lead this proof of concept by identifying individuals that are open for change and eager to learn. They will become the core of your transformation team at a later stage. B
And OpenStack in all of that?
OpenStack has now be around for around 4 years. It is still evolving and yes, it’s going in many directions. However, in the same way enterprises do not use “plain vanilla” Linux, but rather a Linux distribution, I would suggest you to choose an OpenStack distribution. To achieve all the functionality you need today, you may have to complement OpenStack with some proprietary components, particularly in the areas of orchestration, service catalog, policy management and measurement/billing. In your applications, use the OpenStack APIs as much as possible, they will facilitate your migration when you move from one version of OpenStack to the next one. And as OpenStack matures, so will the surrounding functionality. Make sure your orchestration software supports the TOSCA standard. It will enable the interoperable description of application and infrastructure cloud services, the relationships between parts of the service, and the operational behavior of these services (e.g., deploy, patch, shutdown). Again, building on standards will facilitate the migration.
Consistently I keep hearing about the failure of OpenStack. Sure, 20 years ago I heard the same about Linux. And look at where we are today.
The very idea behind a “private cloud” is broken- or is it really?
Frankly I don’t understand the conclusion statement of the article. Enterprises are looking at increasing the efficiency of their IT environment. Business people want it cheaper and IT looks at being able to respond faster. In that process they use technology, starting with virtualization, adding automation, standardization and self-provisioning. Whether you call that a “private cloud”, a “software defined datacenter” or give it another name. Fundamentally, that doesn’t matter. What counts is that the company has an efficient environment that allows it to respond to ever changing demand of their market and their customers. The environment best suited to address that in our mind is a hybrid cloud. And that one includes private cloud technology.