The other night I was at a Taos mixer discussing with several other consultants the impact of virtualization and cloud computing. It was interesting to see others take on the the adoption and even the meaning of cloud. The most interesting part of the conversation was when Trish Palumbo, our VP of Talent Management, asked how we thought things would be different in five years. Five years is like forever in tech and it’s hard to predict what will change, but I gave the same answer that I had given many times before. With the adoption of virtualization, and now cloud computing, I had thought that the biggest change would be with the skill-set of those in IT. I had thought that IT folks that had just a single discipline, such as being a “network guy” or a “storage guy” or a “server guy” (or gal of course), would have to start expanding their skills and understanding into other disciplines. As virtualization brings the configuration and troubleshooting of a majority of infrastructure into the hypervisor and as network and storage converge, you need to have some clue about how it all works together. And I thought this was going to be the case for everyone. So that was my answer, everyone is going to need to know more about all the knobs in their infrastructure, even if just at a high level.
I have been talking about private cloud for the last two years. My last gig was working with a VAR pitching cloud solutions from VMware, HP, Redhat, Cisco, Eucalyptus, etc. But it wasn’t until recently that I actually got to work on a private cloud implementation for a large enterprise. Even though it has been only a few months it’s been a fantastic experience and has allowed me to start to see how the technologies and processes related to cloud can be leveraged by a business. This experience has lead me to believe that my answer, that everyone will need to have an understanding of the entire infrastructure, is not at all how things are going to change.
So Trish, I’d like to revise my answer.
Finding Good People is Hard
I’ve interviewed a lot of people over the last ten years and you know what? Finding a person who is competent in just a single domain is really hard. Looking back I’m not sure why I thought that just because the infrastructure was getting more complex that somehow people would become more competent. Sure there are people out there that are multi-disciplined, people who know their way around an IOS prompt, a vSphere Client and /etc files, but those people have always been around. They are usually your most senior folks and are often found in architect or lead positions. You are not going to find many of them. So what then? The infrastructure is becoming more complicated and it’s unrealistic to expect everyone to know everything. To me it seems that cloud computing constructs are going to be key in building the next generation of infrastructure.
Using IaaS to Deliver More than VMs
Today’s IaaS is mostly about allowing users to self-service provision VMs with minimal interaction with those that run the infrastructure. But what about those of us that run the infrastructure and need more than a VM. As an infrastructure architect at a large enterprise this still involves meeting with the the storage, server and network teams to get all the basic infrastructure provisioned and configured. This is all needed before that cloud stack can be installed and configured. Folks still need to carve out VLANs and LUNs. They still need to grant access to the integrated IP KVMs. It’s all needed before those IaaS users can get their VMs. These siloed teams still exist and will continue to exist. So what about pushing the IaaS concept down deep into the infrastructure layer? What if each of these teams published a service catalog of basic infrastructure?
What if I could login to a portal to provision all the infrastructure needed for a new cloud stack? I could select 4 medium sized nodes for a management cluster and 16 big nodes for a cloud capacity cluster from the service catalog provided by the server team. I could then go and select the IP networks and VLANs made available by the network team, along with load balancing and firewalling for my management stack. Lastly I go shopping for some storage made available by the storage team, choosing some small fast LUNs (or NFS shares) for my management stack, a big slow one for my images and template repository and then a bunch of medium ones for my tenant workloads. Through this portal I should be able to link these infrastructure components together kicking off automated configuration of the switches, routers, arrays, etc.
This self service approach will allow for infrastructure to be turned out at a faster pace and more importantly provide a mechanism to ensure standards are followed and minimize configuration drift. Even for smaller organizations, that may just have a single storage guy instead of a storage team, this could be used to give those on a small IT team the ability to perform basic tasks (like expanding an existing LUN or adding another LUN to a server) freeing up the storage guy to work on more complex issues.
The More Things Change …
This vision is not new and I know there are several companies working on making this a reality. HP, Cisco and to some extent VMware,with their Software Defined Data Center vision, come to mind. Really this is the next logical step for Orchestration, so any of those vendors will be deep into this space. But even in many of those solutions there isn’t a clear separation of duties. And that is really my answer. While I had thought that IT silos were going to have to go away. I now think that silos will live on and that the way the silos interact will change and become more efficient. I think solutions that enable the server, storage and networking teams to keep their kingdoms and allow them to better serve their consumers will become dominate over the next five years.
Really this is probably just another step in the plan to minimize how much us folks in IT need to talk to people. 😛