Cisco this month unveiled what some company executives dubbed one of the biggest announcements for the network equipment maker – ever. With the announcement of its Unified Computing, Cisco officially enters the computing space, although not quite as previous speculation in the industry had outlined. Instead of just producing its own blade server, Cisco unveiled a system that integrated the compute infrastructure, with networking, with storage, with virtualisation.
In this Part 2 of our next generation data centre architecture series, Andre Smit, Cisco’s managing director of data centre sales in the Asia Pacific, and Pete Nicholls, business development manager for data centres in the region, explains the thinking behind the strategic move and the technology powering the new architecture.
Read Next generation data centre architecture: Juniper’s single switching fabric here.
Andre Smit – on the strategy behind Unified Computing
Most innovation today happens in a single domain. You have new innovation in the network space, in the storage space, in the server space or in the virtualization space, but there is no single solution that pulls all these components together in an integrated way. We are announcing Unified Computing system today to address these challenges.
Unified Computing provides a cohesive system that unites computing, networking, storage, and virtualization into one system and simplifies the management layer on top of it.
Unified Computing offers considerable business benefits to our customers, the biggest of which, considering the current economic climate, is reduced total cost of ownership for our customers. Today, most of our customers spend a very high percentage of their IT budgets just keeping the lights on inside the data centre environment. 35% of the IT budget today is spent in the data centre just to maintain the status quo for that they have. Very limited budget is left for innovation and changes in the data centre of our customers.
We provide better agility for our customers to better support the changing needs of their business today. This also improves energy efficiency for our customers.
…on the addressable market for Unified Computing…
Worldwide the market for data centre is roughly US$85 billion. This consists of data centre hardware, software, services and networking. The market that we are going after, that is addressed by the Unified Computing system, is roughly US$20 billion of that, so roughly less than a quarter of the market that also, once again, applies to all these elements of networking, software, services, and hardware. So we are not going after the entire market, but different segments within this market. The segment that we are looking at is any customers that is looking at, and have started to implement virtualisation on their server platform, as well as customers that have extended, is extending, or wants to extend the virtualisation all the way down to the desktop platform that they are running today.
So it is not just about the data centre, it’s an extension of virtualisation all the way through their network infrastructure that this addresses.
So why Cisco – the newest player in this space today? First of all, we are uniquely position. The network of which we have 30+ understanding of, a real fundamental understanding of down to the smallest detail, touches every single component in the data centre. It touches the servers – the compute side, the virtualisation side, as well as the storage element. We believe the network is the platform for customers to grow and deliver all these business benefits in the data centre environment.
Three years ago, Cisco recognised the challenges that customers were facing, the limitations in capability, the management complexity, etc. We started three years ago by assembling a team, drawing expertise from the industry into Cisco to understand all the different silos. The team has taken a brand new approach in developing a system that breaks down the different silos between the compute, networking, virtualisation, management and applications that are on top of the system.
We started delivering components of this vision last year. So the launch of Unified Computing system today is a combination of all the solutions we have launched over the past year. Some of you might ask, is Cisco for the long run. Are we just dabbling here and then get out. Cisco has a long history of entering new markets, markets in transition, changing the market and staying in that market. Every market we entered, we have a concerted vision to be the number one or number two player in that market. This is no exception to us.
Pete Nicholls – on the technology and capabilities of Unified Computing
This is really an evolution that maps into our data centre 3.0 strategy. We have multiple waves of innovation that leads to the Unified Computing strategy.
The Unified Computing system really employs all these technologies that forms the central nervous system of what we are then adding with Unified Computing to create a new class of standards based x86 computing using the network as a platform.
If we take a look at what is available today and what has become best in class so far in servers. The past twenty years have really seen an evolution in size rather than thinking. Servers are as they were years ago, they still have configuration around CPU, memory, file settings, firmware, network settings, everything is hardware dependent, and everything adds complexity. Blade servers are the latest iteration, through rack mount, rack optimised and in the last few years, the blade. Blade servers have taken off in several areas for good reasons. They do introduce a more efficient form factor. Basically, they are taking traditional servers, turning them on their sides and inserting them into a chassis where the blade server makes use of shared components on a common back pane. However, every chassis still needs embedded switches, Ethernet switches, fiberchannel switches, management modules, plus virtualisation and we end up with hundreds of thin devices to manage.
In Cisco, we don’t have a legacy in the compute space, so we were able, three years ago, take a fresh approach and design a new system. So assembling that multi-discipline team around computer, storage, networking, virtualisation and application specialist, and we asked, ‘wouldn’t it be better if we pre-integrated all these elements for our customers?’ We can pre-test them, tuned them and the place to really begin is management because our customers tell us that management complexity is the top inhibitor to realising the scaling, the flexibility and total cost of ownership benefits of virtualisation.
As a first step, rather than have a suite with many management applications, we embed management in each component and into the fabric of the system itself. This allows the entire system to operate as one via the GUI interface and the command line interface that Cisco makes available, but most important through our rich API that is available so third party tools and third party systems management can also be use.
Then we use the unified fabric of all traffic types, so this negates the need for multiple legacy networks. So we strip away the extra cabling and this results in a significant reduction in the complex, expensive, power hungry cabling systems and their associated layers of Ethernet and fibrechannel switches and this further simplifies management. It allows for wire once and walk away performance. As an example, Cisco is consolidation our U.S. data centres into a facility in Richardson in Texas, and we expect to save a million dollars in cabling cost with that feature.
Then we optimised the platform for virtualisation and this is a major target for us, to solve the problems that arrive with the introduction of virtualisation. So we introduced further innovation here by overcoming the existing memory constraints available in server platforms today. We optimised the throughput per virtual machine in terms of networking and we added virtual machine-aware networking. We are introducing a new term here, we created an environment of stateless computing. Stateless means that each compute node has full resources, enabling unprecedented scale and flexibility as a platform for virtualisation.
The benefit for customers is that they can place more virtual machines on a given server estate. This enables customers to save costs by optimising their resources. Because each compute node is stateless, it can be auto-discover within the system and it can be provisioned by use of what we are calling, service profile – this is an obstruction where we separate the configuration from the hardware, so the server, the network and storage can create a service platform that can then apply as a personality to any server within the system. It gives the server a definition of all those configuration settings the service needs, so it can then deduce up either it can support a virtual machine environment, or Oracle, or Microsoft SQL, or the software platforms that are support on the system today.
Finally, we strip away the remaining redundant components, where all the additional Ethernet, fibrechannel switches, the multiple redundant management modules that are no longer needed, and that are replaced by an integrated fabric. The end result is that, compared to legacy systems, we have less than half of the support infrastructure needed to support the same level of application.
…on the system’s specifications…
So this is quite unique. Cisco Unified Computing system is a single system, unified x86 compute, unified network fabric, manageable scalable virtualisation, and consistent access to storage. It scales up to 40 server chassis and that enables support for 320 servers. The important point is that those 320 servers are all managed as a single unit.
Stateless compute provides the location, hardware and provisioning freedom. The platform is optimised around thousands of virtual machines and certainly we manage a virtual machine in this environment exactly as you managed a physical machine.
First and foremost, the Unified Computing platform is an architecture, so that is what we are announcing, an architecture that lends itself to evolving form factors. So today, a lot of the discussions are around blades, but we need to be careful not to miss the message that this architecture lends itself to compute in other form factors.
The key point of this, to simplify it for our customers, is that it is configured and shipped as a single system. So there may be integration work with the customers’ existing management tools that is required, but a lot of the pre-work has been done to remove that load of IT.
We are learning in two-way engagements with customers and finding that on average, we expect to be able to deliver 20% savings in capital expenses, and up to 30% savings in operating expenses.
Our blade enclosure for our servers is much simpler than the complex integrated designs from other vendors today. So the chassis itself is 53% open, there’s a lot of fresh air flow directly from front to back, we don’t need ducking to bend air and avoid hot spots. The power supplies are over 90% efficient.
So really the architecture that we are launching – was this a reactive move? This was three years in development and involved bringing expertise into the company. This was about reacting not to the market, but reacting to the challenges that our customers had, challenges of virtualisation and the complexity that brings to an organisation. It also delivers on further proof points on the Data Centre 3.0 vision that we laid out some 18 months ago.
Data centre 1.0 was really about the closed computing environment that was then changed to what we called data centre 2.0, the change to client/server and distributed computing, very open but also not necessary very efficient and quite difficult for IT to maintain quality and control. What we are moving to now is a much more efficient, highly integrated and virtualised environment, and virtualisation is driving that change. So we are building up here our platform of data centre networking, and this is all about data centre cost containment through standardisation, standardisation in the LAN, in the WAN, and also in the Layer 4-7 services.
On top of that, we have a unified fabric, which appeared from Cisco in the middle of last year, which is optimising data centre technologies through consolidation at the server access layer. Now with Unified Computing, we have integration and this really delivers on the three components of consolidation, virtualisation and optimisation. It delivers location freedom, hardware freedom, and in the service delivery platform for cloud computing, provisioning freedom and the flexibility that brings.
So this is a new paradigm for computing, we see it as something dramatically different from what is available today.
Andrew Smit – on potential market competition
I have to say one thing about this solution. This solution is not a blade for blade solution. This is a new market we are entering here. We are creating a brand new market. This is a market in transition that we are entering here. This is not comparable to any solution that is out there in the market today. Until there is competition, there isn’t – this is a new segment.
Pete Nicholls – on the difference of Unified Computing and mainframes
That’s interesting. I’ve presented this topic a few times now and that is the not the first time that I’ve heard that comment – that this does look like an x86 mainframe and I guess in many respects, it is. But bear in mind it is an open system, it is not a closed system, it is allowing scalable compute, it’s providing a less complex environment, so rather than using a management overlay to try and hide the complexity of what is there, we actually simplified what is there. So we can get levels of availability, scalability that is expected in some instances in the mainframe environment. But with the comparison, there is one area where we digress, is that this is many vendors working with us to provide x86 standards-based computing as a platform for virtualisation in a very open, collaborative fashion.