Best Practices for Flawless Converged Infrastructure
Doug Oathout,Living simply is easy. Moving from complexity to simplicity is not: An interview with HP's expert on Converged Infrastructure.
In the process of moving from legacy IT infrastructures focused on discrete compute, storage, and networking to a converged infrastructure that brings those three elements under one centrally managed roof, there are plenty of chances for poor decisions and missteps. Most of these can be avoided through the careful application of best practices specific to converged architecture deployment. Rather than conjecture, we went to the leading authority on converged IT and sat down for an in-depth conversation with Doug Oathout, vice president of converged infrastructure and alliance marketing at HP.
Tom’s IT Pro: Let’s cut right to the meat: best practices. When businesses want to move into converged architecture, how should they proceed?
Doug Oathout: We recommend a stepped approach to convergence—five steps, actually. The first is to standardize. You want to standardize your converged storage, converged systems, and a flex network. With standardization, you’re able to deliver components in a way that facilitates much faster and efficient deployment. But you’re also laying in the foundation, or the building blocks, for a converged infrastructure that will ultimately take you to a Cloud.
The second step is to optimize. By this, we’re talking about virtualizing and optimizing the environment—putting more than one workload on a system, having a shared storage environment. With this, you can run multiple applications on the storage environment, whether it be a converged storage platform like 3Par or a P4000. You get more usability and more efficiency by having a shared storage environment.
The third step is to automate. Now that you have a converged, optimized infrastructure, what you do is automate the processes for deploying applications. They could be new applications, or maybe you’re migrating old applications onto a converged infrastructure. Once you start automating those processes, you can stand up applications in minutes instead of weeks or months.
Next, set up a self-service portal. Preferably, you want to do this during the migration. The self-service portal allows the lines of business to become select services from the IT team. So whether that’s an ERP service or just a Web application service, they can select the service, and then IT can deploy it very, very quickly.
Lastly, establish your IT as a service broker. Decide whether to deploy your converged infrastructure within IT’s existing infrastructure or as an external IT infrastructure in the Cloud, like with an Amazon or whatever.
So again, IT starts with standardizing, then they optimize, then automate, then they set up the self-service portal, and finally become a service broker. It’s a simple process. It works very well, and it helps our customers migrate what they have. They don't have to throw out what they bought during the last three to four years. They can build on what they've got and migrate over time to a converged infrastructure which then gives them the Cloud capabilities they're looking for.
TIP: At a high level, that all sounds fairly straightforward and simple—maybe too simple. In the real world, organizations must make occasional and probably repeating missteps when going converged. Where do people tend to fall down in this process?
DO: That’s true. A lot of customers don’t standardize. What happens is when they get to the automation phase, they’re stuck. Without standardization in a heterogeneous environment, it’s very difficult to automate. With our HP software portfolio, and with the IBM portfolio, you can do it, but it’s not simple. The simpler your infrastructure is, or the more common modular building blocks you use, the simpler the automation. And when the automation is simple, we can set up the service catalog very easily. Whether it’s going to be a SAP deployment, a Web deployment, or whatever, it’s a snap to set up because you know the infrastructure there. Some customers make the mistake of trying to deploy best of breed technology for every application, and when they get to the automation phase, they can’t do it. There are too many moving parts. It’s too difficult.
The second issue we see with customers when they go to deploy a converged infrastructure is they don’t take storage and networking into consideration. If you’re just consolidating servers, you don't get the full benefit of a converged infrastructure. You need to have a shared storage environment. You need to have a unified, “wired once” network fabric so that you can take advantage of moving workloads. You can take advantage of bringing resources on and off as necessary to support end of year workloads or new work loads. If you don’t have the shared storage environment or the wired once technology in your converged infrastructure, you might deploy faster, but you won’t get the efficiencies and the flexibility you might need to move workloads around or bring more resources to bear on an application.
TIP: Can we make this tangible? Do you have a real life example that illustrates the difference you’re describing?
DO: Cap Gemini did it the right way. They have an SAP outsourcing practice, and they deployed a converged infrastructure. So now when they go sign a SAP outsourcing contract with a customer, they can deploy the SAP environment for the customer to test and run on within two hours. Since they’ve got the converged infrastructure in there, they have their best practices for SAP deployments built into their automated system. Prior to this, it would take them six to eight weeks, not two hours, to stand up an SAP infrastructure for a customer.
TIP: And someone who did it wrong?
DO: Well, without naming names, let me just give you a generic case. We see a lot of customers virtualize their servers without virtualizing the network. When they don’t virtualize the network, that virtual server implementation can only be associated with that server. So if that server has some sort of fault or breakage, they have no way to migrate that workload without restarting it or recovering it onto another server. So there’s no mobility. We see customers falling into that pitfall all the time, and that’s just one of those things where you’ve got to take into consideration the virtualization of the whole environment when you’re deploying the workload.
TIP: Belts are still pretty tight, and a lot of companies are looking for ways to cut corners. When it comes to converged architectures, what’s safe to trim and what’s not?
DO: [laughs] Well, that’s the thing about converged architecture. You don’t really have to trim because it scales from small to large. The architecture will allow you to scale all the way down to five servers and a shared storage environment hooked up through our Virtual Connect technology. So even if you’re a small company with five to 10 servers, you can take advantage of a converged infrastructure in that environment. However, in a small environment, one of the things you can’t cut back on is the management. The best value you can get out of a converged infrastructure is the efficiency, but if you’re not managing the environment, it doesn’t give you the ability to keep costs out. You wind up managing the virtual machines the same way you manage a server, so you don’t get a ratio of 90 virtual machines per IT admin. You wind up with 16 virtual machines per admin.
The objective of a converged infrastructure is to give you a holistic management approach. That’s really the value, and that’s where companies shouldn’t cut. Now, if you're going to trim, one way is in the networking side of things. You can use Virtual Connect or you can use other people doing maybe Fibre Channel over Ethernet. You can use multiple protocols across one wire. The key is to have a high-speed interconnect where you can wire up your infrastructure once and reduce the number of switches in your environment. That’s probably the best and easiest cost savings. If you have a smaller number of network switches, you have a smaller number of cables and a smaller number of ports to manage. The best way to do that is with technology like iSCIS or Fibre over Ethernet, which will allow you to wire up the environment once.
TIP: What’s the most important point we haven’t talked about?
DO: You want to be open. You want to make sure your converged infrastructure solution supports the operating systems you need, the applications you need, and it can all run on the converged infrastructure you choose. Also make sure that you’re choosing a supplier who has the service to support your environment, whether it’s mission critical services or break-fix services for a branch office. Make sure you’ve got somebody who’s fluent in handling services along with the infrastructure.
William Van Winkle has been a full-time tech writer and author since 1998. He specializes in a wide range of coverage areas, including unified communications, virtualization, Cloud Computing, storage solutions and more. William lives in Hillsboro, Oregon with his wife and 2.4 kids, and—when not scrambling to meet article deadlines—he enjoys reading, travel, and writing fiction.