
Following the joint announcement by Cisco, NetApp, and VMware on January 26, a number of my colleagues, including our senior Vice President for enterprise infrastructure, consumer, and telecom research (Vernon Turner), re-engaged in a what has become a wide ranging, and sometimes spirited, discussion about converged infrastructure. A number of product developments over the past couple years have been the catalyst for this ongoing discussion:
An underlying theme of all these announcement is the idea that companies need (or want) to deploy and manage assets in pre-defined "chunks" (e.g., a rack, an aisle, or an entire data center) of converged IT infrastructure rather than deploying and managing distinct products (e.g., servers, storage, or network switches). Thanks to technologies like virtualization, these "chunks" can then be allocated to support specific application sets. They can also be used much more efficiently.
Another critical theme in most of these announcements is that integrated management for automation and provisioning is a critical element converged infrastructure solutions. As my colleague Mary Johnson-Turner (Research Director for system management software) pointed out: "Without advanced, cross-tier management capabilities, you have a bunch of hardwired hardware that can't do much - definitely not dynamically."
Vernon posed a set of questions about this development to a number of us, and I though I'd share my initial answers. They focus on both industry/competitive issues and end customer issues. I'm looking forward to hearing comments from some of my colleagues and from all of you. Do you agree, disagree, or see things differently?
What does 'Converged Infrastructure' mean to you?
Converged infrastructure is the development of a data center (or closet) based on a set of standard elements (x86 processors for compute, PCI for the system bus, ethernet for inter-node connections, SAS for the storage interface, and virtual machines as the standard operating environment). The goal is no longer to deploy and manage each element individually, but to build the optimal (e.g., densest, greenest, simplest) data center.
We are at the beginning of this journey and it will take a decade to complete. To put it in context, we had convergence in the early 90s when it came to how individuals interacted with applications (PCs, ethernet, and TCP/IP). Only a few minor things came from that development (e.g., the Internet, IP telephony, digital cable). The one element not really touched was the systems in the data center. Now, its time has come, and the current discussions about deploying private clouds are one of the first developments.
Which IT vendors are likely to benefit from organizations move to a converged infrastructure?
An initial assessment would appear to conclude that the companies with a broad product and go-to-market portfolio (server, storage, network, management) would be in the most advantageous position, but I actually think the opposite. Most of these companies (e.g., HP, IBM, many of the Japanese IT systems companies) are most at risk of missing out because they will have to overcome the most internal inertia in terms of segregated product development and go-to-market structure. An unusual level of agility will be required.
Focused storage players like (e.g., EMC, NetApp, Compellent) would appear to be in the same boat, but the fact that storage companies are among the early movers in shifting to a converged platform gives them some interesting advantages. EMC, which would appear to be most like the large general IT players, is proving to be quite agile and opportunistic at the moment.
I think the some of the biggest winners in the long run are the next level of component suppliers (e.g., Intel, AMD, Broadcom, LSI, Seagate, Western Digital, Xyratex) though this sector will also be ripe for consolidation.
Which IT vendors are likely to encounter challenges due to converged infrastructure?
For me, the IT vendors with the most to potentially lose are the systems and storage management software suppliers. One of the key design points for converged infrastructure solutions will be to eliminate a majority of the basic configuration and administration tasks that IT staff have to do repetitively (and sometimes do rather poorly) with today's systems and storage management software products.
Of course, as Mary pointed out, converged infrastructure also requires new management solutions and will also present new management requirements (e.g., asset and resource management, performance monitoring, IT governance and risk management). The current storage and system management players have a strong position as incumbent partners, but they need to make sure that they don't spend too much time protecting old bastions as opposed to delivering new capabilities via internal development or acquisition.
What are the leading benefits CIOs should gain from implementing a converged infrastructure?
The leading benefit that many CIO's might think they will achieve is an upfront reduction in spending on hardware assets. I do believe that many will be able to get rid of much of the cable cluttering up their data centers (and this can be a surprisingly big savings), but in terms of hardware spend, the savings will be modest. The real savings here will come when they can significantly improve the utilization rates for all their IT assets in the data center.
The bigger payoff will be in management and administration costs. I touched on this issue and desire in my last blog post: Should You Want to Fire Your Storage Adminstrator?
What should be the leading question CIOs ask vendors who present a converged infrastructure solution?
What are you going to do (service, application development, financing) to help me navigate the transition to converged infrastructure in terms of technology, facilities, and operations?
Which takes us back to the Cisco, NetApp VMware announcement which represents one of the ways that these companies plan to answer that question. The meeting after that should be with your finance department, because a shift to a converged infrastructure is also likely to mean a complete rethinking of IT budgeting and cost allocation. Do any of you this finance restructuring issue as a major barrier, and how do you plan to address it?
Comments
Rick makes an important point when he highlights the connection between converged infrastructure and cloud. Whether it is Cisco/Vmware/EMC VBlocks, HP Converged Infrastructure, IBM Cloudburst or similar platforms, the goal of these products is to simplify and accelerate the virtualization of data centers to enable IT to better pool and share resources across multiple workloads and applications in order to improve cap ex, ope ex and business performance. In my view, cloud builds on virtualization but is distinct from virtualization in that it adds a vital services-centric approach to the way resources are allocated to workloads. Cloud requires a sophisticated view of services that spans the full stack of network, server, storage, hypervisor, management, middleware and applications -- in a way that allows them to be managed in a unfied manner using standardized, policy driven, automated. This is a major extension of responsibility for systems management tools and admin staff. I'm not too worried about the systems management staff and vendors running out of things to do. I think they are going to be plenty busy as management becomes the critical differentiator of hardware solutions as well as the interface for optimizing the way that services and resources are matched in real time.
End-users are adversed to managing more moving parts within their datacenter. Convergence allows IT organizations to end up with fewer elements to manage, which in turn leads to investment in fewer, "larger" pieces of hardware. The economies of scale of buying "larger" pieces of hardware should lead to getting better pricing, and improved capital efficency.
But convergence does not just benefit capital allocation. The ultimate end-goal is to improve operational cost efficiency as well. The net, net of the operational side, is that the driver for convergence is not a technology thing, but an economic function. It somewhat parallels the green argument - most companies want to be more green not because the corporation is a tree-huger, but rather, it is a cost saver.
I see convergence being about convenience and simplifying management. Convergence also provides standardization. The combination of simplication and standardization results in better allocation of operational budgets and more effective and efficient IT operations.