IBM: Virtualisation is the real deal

Systems programming has a new, hip word — virtualisation. It has captured the hearts of IT professionals in small and large companies with its promise of freedom from the tyranny of hardware.

Think your company has invested in the wrong database technology? Virtualise it. Not happy with the hardware? Virtualise it. Virtualisation offers the promise of running more or less anything, more or less anywhere.

IBM can lay claim to have invented the concept through the virtualisation, which 10 years ago had almost vanished from use. Now it plays a key role in the company's plans for the future. EMC may have stolen the virtual crown with its purchase of the hugely popular VMware, but Big Blue wants to take it back.

ZDNet UK talked to the man charged with guiding this offensive, IBM's vice-president for virtualisation, Richard Lechner, to find out where IBM is going with virtualisation, what benefits it offers customers and if virtualisation is so important to IBM, why did it not buy VMware when they had the chance?

Q: How did you come to head up IBM's virtualisation push?
A: 20 years ago, I was a coder doing device microcode for banking terminals, ATMs and so on. Then I did OS/2 development, worked on NDS development and was then a salesperson for the software group. Then I was director of strategy for the software business worldwide, where we had our initial engagements with Red Hat and SuSE and then I helped define our strategy around e-business and enterprise Java. I was in mainframes for five years, looking at taking advantage of our virtualisation capabilities on the mainframe and leveraging Linux on the mainframe. I spent the past two years in storage where storage virtualisation was a key element in our strategy.

Why is virtualisation important to IBM, and what's your top-level strategy?
Virtualisation is a key enabler for our strategy of on-demand. There is a lot of interest in virtualisation but if you were to ask 10 people what it means you would probably get 11 different answers. You could define virtualisation as the abstraction of physical resources to better integrate across your infrastructure and to get better use of those resources.

We assume our clients have a heterogeneous environment so we intend to differentiate our hardware through the virtualisation capabilities built into the hardware based on our mainframe heritage. So we provide virtualisation across our systems, assuming heterogeneous resources so we help our clients work across diverse infrastructures.

We cover about 80 percent of what's out there, and about 100 percent of what's out there in terms of storage. On the server side we manage our own servers, as you would expect. We integrate very well with VMware, and we can support certain levels of HP-UX and Solaris as well as the key Linux distributions.

So if a client has a mixed environment of Intel servers, RISC servers, Unix, Linux, mainframe, storage devices, networking and the storage fabric, Cisco and Brocade and so on, and we can help them with a very large percentage of that.

I do think that kind of holistic approach is unique in the industry today.

How do you explain to a layman, someone working in a financial brokerage say, how virtualisation can help him?
Let me say first that it is a mistake to characterise virtualisation as something only for big companies. More than half, in fact more than 60 percent, of our virtualisation clients are what we characterise as small and medium businesses — less than 1,000 employees. So virtualisation is a technology that is being embraced by a pretty broad sector.

I categorise it in three ways. The first is that there is a significant reduction in TCO in a number of ways. All of our clients are worried about growth per server. Their server farms show unbridled growth. This very large growth is against the background of a very lightly utilised asset. Then even though the cost of hardware is dropping, the costs of software, management and people are not. So from a TCO perspective we can offer the client a chance to get substantially better use out of the assets they already have, reduce the growth rate in servers and storage. In the storage domain in particular, we can help them to eliminate a lot of the redundancy — in having to store multiple copies of the data.

Also, a lot of our clients are worried about this whole area of resiliency. They are being asked by the board to implement this and they have these complex and fragile infrastructures that are hard to make resilient and are expensive to make resilient. By simplifying the infrastructure, it makes it easier to mirror it, and the virtualisation capability also allows clients to have more flexibility in terms of what they deploy at headquarters and what they deploy at a secondary site.

Instead of running both sites "hot, hot", then many clients are more than willing to employ secondary technology if they are only using the secondary site for failover. You cannot do that in the physical world, especially in storage. Storage vendors mandate that you have the same, exact technology at the secondary site that you have at the primary site.

The second major benefit is shared access through a virtualised, integrated infrastructure. There are a lot of benefits through SOA and SOI and this whole integration thing, with the sharing of data and the elimination of redundant data to allow applications to more freely co-operate with each other, which allows your business processes to be more flexible and to interact with each other. Virtualisation allows that to occur.

We have many clients who are afraid to touch their infrastructures because they have allowed them to become so fragile and interwoven. If you can...

For more, click here...

...simplify it and virtualise it, that gives them more flexibility there.

The third motivator is freedom of choice. Just as Java and Linux allowed applications to become un-tethered from specific platforms. It used to be that if a client chose a particular application that choice then mandated a certain operating system, a certain database, a certain hardware platform, certain storage devices. So there had to be an investment in those tools. Java and Linux helped break some of those bindings and then it allowed us to leverage virtualisation to eliminate physical devices and consolidate some of those applications.

In the storage world, virtualisation is doing much the same thing. Now storage devices can be mixed and matched for the right job. So inexpensive hardware can be used when that is all that's required. You can have one administrative interface for your server and storage devices and train your users on that. This is as opposed to saying, "Well, I've got to buy EMC because all my users are trained in EMC, all my scripts are for EMC, all my replication services are EMC's, etc."

Virtualisation allows a lot more flexibility and forces the vendors to compete on features, functions, quality of service, support, etc — which is how it should be.

You mention virtualisation and most people are likely to immediately think of VMware. How do you, as IBM, get their attention?
First, we are happy to talk about VMware. We are one of their largest distributors and we have a lot of clients using it. Often we will talk to them about the benefits of doing virtualisation across your systems.

There are very few clients who only have Intel processors and systems. Most clients have a mixture of at least two architectures. So being able to do partitioning across all the platforms is important. Then you combine that with the ability to make multiple copies of a single instance and then consolidate that in one place. So, for example, take 10 copies of an Exchange server and stick that on a single server. And more and more clients are seeing the advantages of taking different topologies and putting them into a virtualised server.

Now these are capabilities that we have been offering for years on the mainframe and can offer on different platforms and they are things that IBM can do that other vendors cannot.

What is important is to be able to host multiple environments — AIX, Linux, Windows, etc — and to have multiple applications and have those applications talk to each other, without rewriting; to be able to talk IP within the server without having to go outside into the physical network because if you can do that you can get some substantial security benefits and some substantial savings.

So our mid-range servers — the iSeries and zSeries — have the ability to talk IP over memory and we have lots and lots of clients who are doing just exactly that. They are collapsing their entire infrastructure into a single system. They are not just taking those 10 mail servers, those 10 file and print servers and putting them into a larger Intel server, but taking file and print servers and Web application servers, database and transaction servers and hosting all of those inside of a single system. And then, in addition to the performance benefits you get the greater utilisation because all of those servers can mix and match between partitions in a more effective way. And to make the systems more resilient you take one server and replicate it, as opposed to fifty servers.

What are the next steps in virtualisation?
Up to now, people have thought about virtualisation as taking lots of little things and linking them into one big thing and my point is, they are always thinking about similar things — lots of similar things. Not everybody wants to do that, not everybody wants to take everything and put it inside a single system. So the next challenge is how can you take a lot of little things and make them look like one big thing.

How can you make your distributed, heterogeneous servers and storage resources behave as if they are a mainframe? There is great value in that because it can have the same management capabilities, the same workload balancing, the end-to-end performance optimisation, the same service level agreements, the same security capabilities and so on.

Why didn't you buy VMware when you had the chance, before they were snapped up by EMC?
I won't answer that. But VMware is a good partner of ours; the way we work with them is twofold.

Firstly, as a technology distributor we make our Intel servers the best hardware platform for VMware.

Secondly, we can integrate a VMware partition into an Intel server — whether it be from us or someone else — and make it a key element within this distributed infrastructure.

There are other interesting technologies within the Intel space that support partitioning, Xen for instance, and we will support them as well.

We can see the innovative possibilities of virtualisation, but where does the innovation come from?
The innovation comes from the client. How can we help an automobile maker or an electronics manufacturer become more innovative? We can help them by simplifying and reducing the costs of their infrastructure, so they are able to put their time and their energy into new application development, new processes as opposed to managing the assets that they already have. We can help them by giving them this access under shared infrastructure. You cannot underestimate the value of being able to take the information and re-integrate the enterprise so that all of the information is available from all of the knowledge workers and is available to outside parties, etc.

How important is Linux and open source to your virtualisation strategy?
Forty percent of our mainframe customers are running Linux on the zSeries. The greatest penetration we every anticipated was 25 percent. Clients are finding new and innovative ways to deploy that technology just from living in close proximity to their business and picking up ideas.