logo

This blog entry is about the need to simplify data centers, not to mask or automate the stuff that’s there already. And it’s about Converged Infrastructure: what it is and why you should know about it. And finally, it’s about why some of today’s infrastructure management products – which simply manipulate in-place complexity – are dangerous Band-Aids masking your real operation and cost issues. The two philosophies There are two major philosophies/architectures for how to manage production infrastructure. The first is to control in-place assets with code or run-book processes. The other abstracts the topology, then re-assembles it in real-time. One’s a Band-Aid. The other is a cure for complexity. There is an analogy in the software world: OS virtualization vs. run-book server provisioning.  While automated run-book approaches work, you’re still limited by physical software. In contrast, virtualization’s abstraction has revolutionized and simplified the software world. Here’s a bit more detail:

• In-place automation technology masks underlying topologies and complexity by manipulating the in-place infrastructure assets.  While this approach has had merit (it was also pursued for years by Cassatt and Scalent), it is ultimately constrained by in-place physical topology, and doesn’t reduce or simplify physical infrastructure or capital investments. This approach has yet to be proven in the production data center.
• Converged Infrastructure (CI) technology (also known as Unified Computing).  This approach first abstracts/virtualizes I/O, networking, and storage connectivity so it can be re-configured at will. In doing so, it reduces and simplifies underlying physical infrastructure (the need for NICs, HBAs, cables, switches, etc.), reducing both capital and operational costs yet still increasing agility/flexibility.  (More elaboration on these two approaches is available on the Fountainhead Blog)

Here’s an example of implementing each approach: With in-place automation, the management engine is constrained by physical data center topology.  If a physical server needs to be re-purposed but does not physically contain the proper quantity of NICs or  HBAs, or is not physically connected to required switches or IP load balancers, it cannot be used . Similarly, in the case of Disaster Recovery, if the recovery environment topology is not a close physical match to the production environment, then it cannot be used.  In contrast, Converged Infrastructure technology permits IT management to infinitely reconfigure physical servers, I/O, switches and IP load balancing on-the-fly. This allows individual servers – or indeed entire environments – to be configured to exactly to match specific topology. This also allows 100% sharable pools of servers, and 100% sharable failover (Disaster Recovery) environments. How vendor approaches are playing out – It’s all about CI The reasonable question to ask is: Which vendors are choosing to offer which approach? As it turns out, nearly all vendors have elected to go with CI.  Cisco is most notable with their UCS, followed by HP with their BladeCenter Matrix, IBM with CloudBurst and open fabric manager, and finally (humbly) Egenera with PAN Manager.  All use forms of CI to abstract I/O, network and switching. Interestingly, Dell has chosen to break with the pack by acquiring in-place automation technology for their Infrastructure Manager product, based on in-place automation…. although they continue to OEM PAN Manager from Egenera. Why has CI/unified computing essentially become the de-facto industry direction?  Because of its perfect complementarity with virtualization, and its ability to solve the data center infrastructure complexity problem.  Analogous to how hypervisors virtualize and simplify the data center software domain, CI effectively virtualizes and simplifies the data center I/O and network infrastructure domain. When used in concert, OS virtualization and CI abstract both software and infrastructure, letting IT operations configure and deploy entire server profiles at will, and with minimum time/cost/components. Because of the broad CI following by most vendors, we should expect to see a degree of standards and, potentially, even interoperability.  And for these reasons, IT shops that go with CI will also experience a lower degree of technology risk simply because of its breadth of adoption. Furthermore, we’ll begin to see accelerated maturity of CI products enabling even more IT management functions. Where Egenera’s placing its bets For the past 10 years, Egenera has backed the Converged Infrastructure approach. Initially implemented as a bundled HW/SW combination on our BladeFrame, it’s now also available in a software-only form on Dell blades (and stay tuned for other platforms). We believe that Converged Infrastructure and use of fabric-based networking is the future of the data center – and the way of simplicity.   And we believe that those who take another path will wake up one day to find themselves woefully behind the industry.