Listening to Paul Maritz describe VMware’s vision for automated Data Center Operations at VMworld 2011 in Las Vegas last month (a review of which is in this post), caused a very important issue to surface in follow up conversation with several attendees at the show. The issue is the degree to which the current organization of Data Center Operations would prevent an enterprise from being able to attain the benefits of automated IT Operations.
The most common articulation of the issue took the following form. “If this is going to work, then the automated actions need to extend into the physical infrastructure that supports the virtualization layer. This is not under our (the virtualization team’s) control today, and the current organization would prevent any of our products from taking automated actions in those layers.”
To bring up a simple case, in most organizations today, the storage team is in a completely different management silo than the virtualization team. Yes they talk to each other when new services need to be provisions or when their are problems to be resolved (but then only reluctantly). But this organization structure is not designed to have the virtualization team work hand in glove with the storage team so as to facilitate smooth and automated IT Operations.
The same is true for the other hardware layers that underlie the virtualization stack. In many organizations server hardware, the local network the WAN, the SAN, the Storage Arrays, and the teams that support the key operating systems in the environment are all in silos (in some cases in fiefdoms) where the whole design of the organization seems focused upon preventing effective teamwork across silos, and between the virtualization team and those silos. The same is most often true for the teams that support the key operating systems and middleware platforms that run in the virtual machines. In fact in most organizations, the virtualization team is just one more silo along with the rest of the silos in the organization as shown below.
Siloed Data Center Operations
Virtualized Data Center Operations
Of course there is a better way. This will likely get met with heavy resistance from the people who are in the silos (or fiefdoms as the case may be), but the answer is that the teams that engineer and support the infrastructure of virtualization need to be part of the virtualization team. The same is true for the teams that support the key systems software that runs in the guests of the virtualization platform.
Virtualized Data Center Operations
Notice the two big changes which will have a profound impact upon how IT Operations runs. The first is Virtual Operations is instantiated as a function that has within it what the virtualization team does today AND the necessary Systems Engineering and Support to engineer and operate the supporting WAN, LAN, SAN, Storage and key guest operating systems.
The second big change is that if virtualization is going to succeed then the notion of how applications are supported needs to change. The bottom line is that an Applications Operations team is needed that is responsible for the operations and performance of all applications in production. These teams need to know enough about how the applications work in production to be able to be responsible for their response time profile to the owners of the applications. These teams will not replace the teams that build and test custom applications, nor will they replace the teams that qualify, modify and upgrade line of business purchased applications like SAP and Oracle.
If automated IT Operations is going to succeed and deliver its promised benefits then IT Operations is going to have to get reorganized – with supporting hardware teams part of the virtualization team. Furthermore Application Operations will have to be instantiated as a function that is responsible for the actual service level delivered by the applications to their constituents.