Country missing? Please select your nearest region...
IBM’s Matt Hogstrom explains his company’s vision for software-defined everything, its approach for 'smarter IT' which has a focus on data and is delivered through the Cloud
30 August 2013 by Penny Jones - DatacenterDynamics
DCD FOCUS: Everyone seems to have a software-defined play these days. IBM’s is the ‘software-defined everything’. What does this mean?
Matt Hogstrom: Everyone is talking about software defined now, in lots of contexts – the software-defined data center (SDDC), software-defined network (SDN) and software-defined storage (SDS). These are the three legs of the stool. Another term out there is software-defined infrastructure (SDI). This refers to the collective of compute, storage and network and the intelligence for managing the infrastructure.
When IBM talks about software defined, we use the term software-defined environment (SDE). It is a more global term for us. It is about how we view the infrastructure as the critical, perhaps even foundational, basis. What we find more important – or interesting – is the ability to capture information about workloads and the way information is processed, so we can set levels or objectives from a workload perspective, and manage these according to SLAs.
A good example can be seen when writing policies. I may want to specify that I have an expectation of about 700 IOPS for this data that will be stored. I may also have a policy that that data needs to be encrypted when it is placed on disc. I might also want a checkpoint and for this and for it to be backed up on a 12-hourly basis.
SDE can specify policies on this business level – the human level – then let the infrastructure make the appropriate decisions to be able to achieve those goals.
In the past, someone would say I need to be backed up every 12 hours, and someone would go out and look at the environment profiles, then carry out lots of manual processes before it could be automated. We are now taking those institutional paradigms and putting them into a recognizable context covering infrastructure from a variety of vendors.
So despite being delivered over the Cloud, this is more about best practice rather than technology?
When we look at a breadth of our customers, we see that they end up building very similar systems. Governance is not that unique to an institution. What we want to capture is these best practices. To take them as patterns, then form policies based around these. Virtualization and API (application performance interface)–driven deployment has already allowed the industry to get to the place where infrastructure is provided almost on demand. We can automate almost every aspect of the data center, right form the network provisioning to the storage set up and provisioning.
You talk about a common language required for SDE to work?
From an IBM perspective, everyone has to have a common view of the infrastructure for the SDE. There is a lot of churn in the market, and a lot of companies are wrestling with software defined. At IBM we talk about the SDE consumer and SDE providers. An SDE consumer is someone who has a workload they want to deploy and they want to specify policies and non-functional requirements. An SDE provider is a person building an infrastructure to serve these consumers, that could be internal IT, which is moving IT to the Cloud, or someone like Rackspace or Amazon, providing cloud services public or private. The important part is, both need to be able to communicate.
Are we talking then about open standards?
This is why IBM is investing heavily in Open Stack. If you look at what hat community is doing, Open Stack is to infrastructure what JCP (Java Community Process - set up for the development of Java standards) was to programming models. It’s the place where vendors come to build a consistent view and see how their products would plug into this API-driven behavioural model of an environment. It is a good framework for getting some consistency in thinking and the approach to doing software-defined everything.
We believe the community is critical to aligning people’s thinking around the API ecosystem through the Cloud.
We are also rallying on Open Stack APIs so we can avoid lock-in and use the community for building our adaptor.
How does this relate to IBM’s own product sets?
We are looking at all products in IBM in relation to SDE. We want to see how they can snap into this framework. The idea is to deliver, via APIs, behavioural models so that customers can build a cloud that looks and feels the same as any cloud from Open Stack. But vendors can innovate and deliver additional capability within that framework.
We have been working on this vision for a number of years. We had patents back in 2009 around virtual systems and we think we are really ahead of the curve. These are logic based deployments. We think where we could really accelerate this, however, not by having patents and products but by getting the community to embrace the way of expressing these patents.
The policies are not out there yet, and we have a sort of a working code we want to take back to the Open Stack community around that. The real value is for providers to be able to deliver on service objectives. The language we use is just a template but we need a common way of describing it. At the end of the day, no one wants to be locked in. They want to know they will still have value if they change portions of their stack in the future.
We have a number of development environments we use to incubate collections of IBM products, and like all companies we like to put our research and innovations into a roadmap. But much of what is happening in this space is cultural.
A lot of customers who get into this space already have legacy environments and applications they have been using for years, if not decades. How you bring these worlds together is the important part. We have to make sure as well as incubating new technologies. We are also drawing a line where folks can reach their aspirations as well as meet their immediate needs and pressures.
We expect to find, however, that as subject matter experts dig deep into one leg of the stool they will start to get appreciation for another leg, and eventually we will see people become more fluent in all legs of the stool. This is an interesting operational challenge for customers of IBM.
I used to be in the Systems Software Group, in a group looking at application integration middleware, where Websphere was born. I moved to a systems technology group to build systems software which built on the experience of SDE from the consumer side. Now I am looking at it from the provider side. Even in IBM, skills are being adjusted along with subject matter to cross pollinate. We all have to deliver a different perspective and become more efficient.