Subscribe
About

Getting past square one

Therese van Wyk
By Therese van Wyk
Johannesburg, 06 Jun 2011

In the journey towards cloud computing, virtualisation is the first necessary step. It's a first step that can bring about huge savings and great flexibility in IT resources. But businesses often run into a problem once the servers and storage are done. Many organisations find it tremendously difficult to get all the benefits promised by virtualisation, or to continue towards cloud computing for the availability and agility it offers.

The reason? It's so easy to virtualise servers and storage, that the management maturity needed to go further is forgotten in the rush.

"Everybody is trying virtualisation," says Greg Hatfield, GM of cloud services at Dimension Data, "but few are getting beyond a 15% to 20% virtualisation level for a variety of reasons. Clients are balking even before getting to implementation. They realise they don't have the ability to look after a virtualised environment."

Virtualisation is one of several technologies that enable cloud computing.

Virtualisation vs cloud

"Server virtualisation is the ability to consolidate multiple physical servers with multiple virtual machines," says Richard Vester, executive head of hosted services at Vodacom Business Services.

"It's a consolidation technology that reduces the amount of physical servers you use to run different applications. Desktop virtualisation reduces the requirement for end-user applications on a desktop. Now, applications can be housed in a public cloud, allowing you to access data from anywhere.

“It's secure because you're not carrying data with you. When you run the entire desktop on a virtual environment, tied to the data centre, you immediately reduce the number of physical servers required to run those desktops."

Mike Schroeder, senior director, Oracle hardware marketing EMEA, concurs: "Virtualisation makes it possible to consolidate workloads on fewer numbers of physical servers and physical storage devices. It enables moving to a private, public or hybrid cloud.

“But very often virtualisation is presented as the end itself. One has to ask why you are virtualising. One of the main drivers for virtualisation is improving utilisation of physical servers. Typically, in industry-standard Windows environments, average utilisation of servers before virtualisation is around 10% to 15%. In Unix environments, it will be 20% to 25%."

But virtualisation is not a silver bullet for all IT, says Roelof Louw, team lead, business development at T-Systems SA.

"Virtualisation and cloud computing will not address your entire IT environment. At the end of the journey, there will still be systems that are not virtualised. You need to understand your business and IT landscape very well, understand the different cloud solutions that exist - and then match appropriate solutions to specific parts of the business."

Virtualisation is a technology, but cloud computing is a business model, continues Vodacom's Vester.

"Cloud computing is infrastructure-as-a-service, meaning the ability to consume resources in the cloud, as well as software-as-a-service, which is the applications you require on top of the infrastructure, as well as the platforms that house the applications, also called platforms-as-a-service."

There are many descriptions of cloud computing. At least one service provider prefers to stick to an official definition of it.

"We use a definition of cloud computing from the National Institute of Standards and Technology," says T-Systems' Louw. "The definition states that cloud computing needs to provide on-demand self-service, which needs to be measured because it is usage-based. Also, there needs to be flexibility in resource pooling with broad network access. Virtualisation provides the flexibility in resource pooling, where servers and storage become consolidated, shared resources."

Virtualisation is not the silver bullet for all IT.

Roelof Louw, T-Systems

There are a number of phases of virtualisation, and a number of different value propositions associated with those phases,says Robert Picton, technical architect consultant at Dimension Data.

"Most organisations are looking at that first phase of virtualisation, which is about server and storage consolidation, and saving on power and cooling of the data centre. Big savings are possible, depending on who you talk to.

"If you have a three-year server maintenance renewal cycle, and you are told you can replace 100 physical servers, by virtualising and buying 10 physical servers instead, it is a highly compelling value proposition," adds Hatfield.

"There are also operational savings on maintenance. Facilities will use less floor space, power and cooling. We have case studies where people have saved 60% in this phase of virtualisation."

Trouble ahead

Virtualising beyond servers and storage is where things get tough. The first reason for this could be that personnel lack the skills to operate complex virtualisation management tools.

Moreover, business procedures to prevent virtual machine sprawl, or to inform business managers about virtual IT resources operated on their behalf, may not be implemented properly. This is when other stakeholders start voicing their reservations.

"Everybody likes the idea of virtualisation, because consolidation saves you money and you have one server instead of a 100 or buy one PABX instead of five. So they start with a test and development environment, or a non-core application and everything looks good," says Hatfield.

"When they want to implement virtualisation beyond the 20% threshold, they start asking themselves, 'Do I have the ability to execute on my ITIL disciplines? Do I have the systems, people and processes to manage availability, to troubleshoot accurately, to provide my business stakeholders with SLAs?'.

“Invariably, the application owners who do not run the data centre start complaining. They say, 'I can't hug my server, the one my application runs on. I don't want to be thrown into the pot with 40 other applications. How can I be certain that if something goes wrong, we'll be able to isolate the problem and fix my application?'"

Nowadays we need a much more policy-driven approach. We need to manage our underlying infrastructure to meet service level agreements.

Christoph Dobroschke, VMware

Also, managing the six technology layers in a virtualised IT computing stack can be mind-bogglingly complex.

"At the bottom of the stack will be the storage devices and technologies such as primary and secondary tier disk or tape," explains Oracle's Schroeder. "The next layer will be your physical servers. The third layer contains the operating systems, hypervisors and virtualisation software. The fourth layer contains the databases, while the middleware that ties everything together is in the fifth layer. The applications are in the sixth layer."

Each layer could contain technology from several vendors. But available management tools only manage one or two layers properly and usually work best for the vendor's own technology. Vendor-agnostic management tools tend to lag behind and have less functionality and integration. Even experienced cloud computing players have to cobble together a bunch of tools to manage performance across the entire virtualised IT stack.

Finding a problematic needle in a virtualised stack can be far more difficult than its physical IT equivalent. An end-user service can be made up of a database, a Web server and possibly another application server.

Take the example of an ATM that makes a customer wait for a minute or two after the card has been inserted and PIN number entered, says Rory Green, data centre and virtualisation product sales specialist at Cisco.

"If you ask an IT person why it takes so long, he could say his database and applications are up and running, so he thinks everything is fine. But it's about how everything ties together and what the user experience is of the whole process."

If you think you are going to virtualise 80% of your data centre in six months to a year and everything will be peachy, you are naive.

Greg Hatfield, Dimension Data

Furthermore, the company may find that managing SLAs in the real computing world is bad enough, but can become unbearably uncertain in the virtualised world, which is even more difficult to manage. Delivering SLAs means choosing management tools to enable that. The choice of management tools can then influence the choice of technologies.

"A business can think virtualisation is great because the amount of equipment in the data centre has been reduced dramatically," explains Green. "But if you don't have management in your IT department's DNA, then lurking inside those few little servers in the corner could be a worse mess than you had before.

“At the business level, you may need to apply a security or quality-of-service policy. In the past, the policy was applied to a physical network. Now you have a virtual machine and a virtual network card connecting to a virtual switch, which makes it more difficult to troubleshoot."

A company could also find that a lot of its IT architecture must be overhauled before further steps towards cloud computing would be prudent. Its IT infrastructure might be incapable of bearing the extra loads required.

"If we look at networking infrastructure within the data centre, there have been a number of drivers that have forced a rethink of the way we've been doing things," adds Green.

"Storage requirements have been growing over the years. A secure and high-speed network to transport information from storage devices to servers is essential to take advantage of the workload mobility, redundancy and high availability features that virtualisation provides."

The road ahead

Virtualisation technology layers have many interdependencies. When management tools cannot do root-cause analysis across the technologies in one layer, as well as vertically across the layers in the IT stack, no one really knows what is going on. This lack of guarantee is not what business unit managers, dependent on applications, want to hear.

"Enabling virtualisation technology is your smallest and easiest step on the journey to cloud computing," says Louw. "After that, the real work starts. To really control, manage, secure and govern it, you need a whole lot of other solutions. Only then can you really unlock the benefits of the virtualisation platform and avoid problems.

“You need a mechanism to provision and deploy virtual machines, to track usage on them and charge business units, and to release them again so that capacity is redeployed once a project is finished."

Managing hardware in a virtualised environment is very different from managing a physical one, adds Christoph Dobroschke, senior product marketing manager EMEA, at VMware.

"We need to understand the change in managing hardware. It used to be vertical. There were specific management frameworks for servers, storage and networks. They were very rigid, though. Nowadays, we need a much more policy-driven approach. We need to manage our underlying infrastructure to meet service level agreements.

"As a consumer of IT, I'm not concerned about how hardware works," continues Dobroschke. "I want an e-mail or ERP service with a service level agreement that meets my needs. So IT management needs to adapt to this new strategy, where SLAs define policies, and the policies define the underlying hardware needs.

“The underlying hardware completes the cycle by providing a type of contract to the service or application that allows setting a new service level agreement. There needs to be a much more holistic approach."

Getting all the virtualisation problems sorted out for a business is no quick win. It needs strategy and planning to progress to further virtualisation levels and eventually, cloud computing.

"There is still a good business case in addressing all of those issues as long as you do it in an organised and planned fashion," states Hatfield. "If you think you are going to virtualise 80% of your data centre in six months to a year and everything will be peachy, you are naive. But if you are realistic, you put together a strategic roadmap for what you can safely virtualise now and over the next two stages.

"Then, in the next three years, the infrastructure environment will be evolved, as well as parts of the operating model. And it will be done in such a way that the business is not exposed to organisational risk.

“If you do it meticulously, you still get all the upside that everybody raves about in the press but it's not a quick win. It's a quick win when you virtualise 15% of your environment. If you want the real upside that virtualisation and cloud computing can offer, you have to invest in a strategy."

Getting beyond the quick wins of server and storage virtualisation is not for the faint of heart or lazy of governance. For those who have implemented virtualisation according to international cloud computing standards, paid attention to business requirements and related SLAs, the last step to cloud should just be a step. For others, it may mean a far longer journey to full virtualisation or cloud computing.

Share