Converged IT infrastructure and converged IT organizations are major transitions that complement a fundamental transition to cloud and virtualized computing. These trends, together with new IO-centric applications that converge transactional and analytic processes enabled by non-mechanical pervasive persistent storage, will revolutionize computing. The bottom line is reduced IT costs, improved time to solution, and vastly improved productivity from the application of IT.
Wikibon published a converged infrastructure projection, which shows that reference architectures will be the dominant deployment over the next five years. Reference architecture together with single SKU offerings (e.g. VCE Vblock, Oracle Exadata, etc.) are projected to be 65% of new sales by 2017.
Two converged infrastructure announcements from IBM and EMC underline these transitions. EMC announced VSPEX on April 12, 2012. IBM announced PureSystems on day earlier on April 11.
EMC VSPEX has focused on four initial solutions:
- Private cloud for Microsoft Hyper-V environments;
- Private cloud for VMware environments;
- End-user virtualization for VMware View environments (250 - 2,000 virtual desktops);
- End-user virtualization for Citrix XenDesktop environments (250 - 2,000 virtual desktops).
These VSPEX offerings are not as tightly integrated as the VCE Vblock; they provide more freedom of choice within a limited number of software and hardware vendors and less integration of the component pieces; there is no single SKU. The choice of servers is between Cisco servers (Dell, HP & IBM are in light type), EMC storage and Brocade storage networking. The "proven" infrastructure is extensive design, validation, documentation, and testing of these systems to minimize the cost of purchasing and operating of the overall deployment of an initial 14 IT solutions. These racks can be assembled and delivered with the logo of an EMC partner.
VSPEX in our view is a response to pressure from the channel and the industry to provide more choice and flexibility. Competitively, this has been the lead strategy of both NetApp and HP, while EMC's reference architecture activity has been largely un-marketed-- until VSPEX. While Vblock has been successful, the so-called reference architecture business is much larger and EMC clearly had an imperative to respond with a more cogent offering or risk losing broader market momentum. VSPEX is designed to fit that bill.
For its part, IBM introduced PureSystems with two key components:
- PureFlex, an integrated offering of IBM servers, storage, and networking;
- PureApplication, middleware that automates an application around a set of workflow processes and procedures designed by IBM, partners, or the user.
IBM's play is from it's server strength. Codenamed "clean slate" the PureSystems offering will essentially replace IBM's BladeServer and combines a new server and "scale in" networking architecture based on the BNT acquisition. As well, the solution integrates the IBM Storewize V7000 array. It is essentially a menu of options that is highly configurable, with templates that contain application and workload knowledge. If in fact IBM can demonstrate this application affinity, it will be the most general purpose application platform on the market.
Comparison between Approaches
IT infrastructure is designed to run applications. The predominant current model is roll your own (RYO); an IT department or reseller assembles a hardware solution to run an application or set of applications together with the necessary screwdrivers to maintain application delivery. The cost of maintenance of the environment for administrators and hardware suppliers is a very significant part of the total cost of deploying an application.
The rate of change of an application is painfully slow, and gets worse the higher the availability and security requirements, and the greater the level of integration of the applications. At the heart is the time required to test all the different parts of the ecosystem together in an environment where there is constant change of each component. Overall application change rates of once a year are very common.
Wikibon suggests five major criteria for assessing IT infrastructure;
- Cost of equipment and initial deployment;
- Both approaches may help in reducing the cost of initial equipment and deployment by eliminating configuration errors and over-provisioning, and by reducing the the pre-sales and post-sales effort required from vendors and their channels. In addition, the economies of scale will reduce the cost of infrastructure software, particularly management software. At the same time the components may not be the very best of breed, but “good enough”.
- Cost of lifetime vendor maintenance and operational administration, and de-commisioning;
- The degree of ability to reduce vendor maintenance and operational cost is as yet unknown for both vendors – it will depend on the quality of pre-testing of integrated hardware and software patches, and the quality of the management tools to monitor performance and availability.
- Number of changes that can be introduced to improve application value;
- An important aspect is integration between infrastructure change and application change from the application vendor. EMC has the ability to do this well with VMware environments.
- Integration into exisiting and ongoing applications, and integration into the processes and skill sets of the organization;
- EMC has chosen standalone environments for its initial offerings. IBM has the potential to integrate more tightly with PureApplication, but how much IBM service will be required to achieve this remains to be seen.
- Speed of initial deployment (where this is a constraint on delivering application value).
- Both EMC and IBM should improve on speed of initial deployment. This has the potential to save early delivery costs, and in some circumstances could lead to earlier realization of application value.
The strong announcements from EMC and IBM are good news for IT organizations. Cisco, Dell, HP, NetApp and others also have converged infrastructure offerings covering servers, storage and networking. Wikibon believes that this will introduce lower costs for IT departments.
As Wikibon described in a professional alert "Drawing Converged Infrastructure Boundaries" there are other boundaries that can be drawn round converged infrastructure offerings; for example, storage and backup software and processes can be combined with NetApp storage offerings.
One of the resistances to converged infrastructure adoption is IT organization, and the currently distributed management of servers, storage, networking, database and applications. More "converged" IT organizations will be needed to create effective "ownership" of converged infrastructure assets. If an existing siloed department takes ownership of converged infrastructure, it will profoundly determine the choice of technology.
Action Item: Converged Infrastructure deployment will become business as usual for IT organizations. The first challenge will be to create an IT organization that will facilitate the development and adoption of converged infrastructure (e.g., DevOps). The choice of department or the choice to create a new department to lead adoption should be in line with the business priorities of the major stakeholders. Thought should be given to the metrics that govern adoption; ability to make change (e.g., increase in the number of changes made) to increase application functionality is a key metrics that should be emphasized.