Another trend affecting data centers – “Convergence”
Interesting trends are taking root around us and one of them is convergence. The term conjures up different thoughts depending on our background and experiences. Economists may say convergence is the parity of per capita income around the world. Convergence for telecom is the combination of voice, data and entertainment services. So what does it mean for data centers? In one of my recent informal webcast polls of technologists, one opinion was that convergence implied the union of telecom and IT. Reality is that data centers now are the hub and source for voice, video, data and application services.
So if we look at application workloads running in data centers, there are four infrastructure capacity variables – CPU, Memory, Storage and Network. One approach is to optimize on the utilization of one of these variables. If we decide to optimize on Storage, then it must be virtualized and/or provided as a service. Implementation would involve purchase of the best of breed storage hardware, and building highly skilled teams to manage, tweak and optimize performance of the storage resources. Similarly a COE(Center of Excellence) for servers (CPU and Memory) must be formed for servers and for networks. This implies that any project would involve multiple teams and project management would be a challenge, to put it lightly. This reminds me of my mainframe experience in relation to the distributed platform. We could get an entire application developed, tested and ready to go before getting a RACF id to even access the mainframe.
Enter converged infrastructure! Cisco’s Unified Computing System (UCS) is a pioneer, which combines the server resources with network and storage access setup. When UCS is pre-integrated with actual storage resources from EMC we have Vblock™. When it is pre-integrated with storage resources from NetApp we have FlexPod™. The advantage of converged infrastructures is that individual projects can now proceed much faster once the overall network and storage policies have been setup. Such offerings save IT departments the trouble of time-consuming infrastructure integration projects. Additionally, they simplify the troubleshooting efforts when problems arise. Without integrated infrastructures, when an application does not perform well, application owners shuttle between the COEs and are left to unravel the actual problem, because each team monitors their own domain and says there is no problem in their domain. Integrated infrastructures provide application owners a single point of contact for internal and external vendor support.
I am headed this afternoon to Las Vegas for CA World 2011, where we will showcase Cisco UCS. If you are at the show, please come by to see Cisco UCS in action. We also have speaking sessions dedicated to UCS at the show and in the Cisco booth theater. Additionally we will have talks in our theater by VCE – the company which brings us the Vblock™.