A brief introduction to converged infrastructure

Sometimes, it’s better to think inside the box. Bundling different IT components together into a single unit may just solve some of your computing problems, if you plan it right. Welcome to the world of IT convergence.

Depending on which vendor or analyst you talk to, it’s known as an integrated system, a unified computing system, or converged infrastructure. Whatever name you use, they are all trying to do something broadly similar: simplify the procurement and management of different IT components by bundling them together.

Converged infrastructure is the latest development in a long journey towards making IT infrastructure more efficient. Virtualization got us some of the way there, because it enabled us to consolidate our physical servers. We ended up with smaller numbers of larger boxes, running large numbers of virtual machines.

That was all very well, but it came at a price. Management was difficult, and IT departments ended up with new challenges. ‘VM sprawl’ meant more machines to manage, while storage and servers still had to be manually networked together and configured for different workloads.

Unless you’re a real pistonhead, you wouldn’t dream of creating a car this way. Instead, you’d go and buy one, pre-built and configured, with all the different parts working together in concert. This is what vendors hope you’ll do with IT infrastructure. Instead of buying separate CPU, storage and network components and configuring them all, you can opt for an integrated experience in a box.

Just as with cars, the converged IT solution can be preconfigured according to your particular workload. When buying a vehicle, you might choose a hatchback with good safety features if you’re a family type. Conversely, you might buy a yellow Lamborghini if you’re young, single, enjoy annoying people with loud engines, and don’t mind too much about fitting a month’s groceries in the back.

Similarly, you’ll look for a different converged IT solution depending on your endgame. You might choose one configuration over another depending on whether you’re into online transaction processing (OLTP) or analytics, say.

The concept must have traction, because the market is growing. Last September, IDC said that the integrated infrastructure and platforms market grew 33.8 per cent year on year during the second quarter that year. First half revenue ballooned 35.9 per cent. People are buying. Why?

One number to call, and one person to yell at

Benefits

There are several benefits. One of the biggest is easy deployment and reduced deployment times. Bolting together storage and compute resources for particular applications, and possibly networking the two together, can be an annoying and time-consuming process. IT departments - especially those that have relatively little internal skill or human resource - may appreciate a pre-packaged box with everything inside, ready to go. If a business department is eager to get working on a new project, it can be a useful way to cut deployment times.

Because these devices arrive pre-built and pre-configured, there are also fewer things for the IT department to maintain. If the system fails to perform as planned, or suffers an outage, there’s one number to call, and one person to yell at.

When everything is running smoothly, these systems also often come with management interfaces that make it easy to manage the whole thing from a single pane of glass. That reduces the human resources overhead, helping to drive down the total cost of ownership (TCO).

Converged infrastructure systems are also predictable, and reliable. They are built on a reference architecture that has been pre-tested against the kinds of workload required by the IT department. This saves the IT department having to develop and carry out those tests before to find potential flaws in the proposed system.

Then, there’s the more efficient use of resources. One of the biggest benefits of a modern, hyperconverged system is the ability to break down silos. In a traditional infrastructure where storage and CPU power are managed separately, IT departments may have allocated different physical computing and storage resources to different applications and business departments. This can lead to duplicate sets of resources (a RAID array here, and another one over there) that are each under-used.

With a converged system, the virtualised resource is not only shared between multiple applications and departments, but also managed automatically, further honing the efficiency of the system.

What’s next?

Converged systems are evolving, with modern variants focusing management on the virtual machine, with commodity computing resources (typically x86) and disks managed in the background. Increasingly, you’ll see them offered as bolt-together nodes, enabling them to scale out.

Typically, you’ll see this referred to as hyperconvergence. It unites storage, compute, and networking in a single box around a hypervisor that does all of the infrastructure management for you. The computing is virtualised. The storage is virtualised. There’s strong software layer across everything that both abstracts and manages everything.

There are potential drawbacks. The closer the knit between compute, storage, networking and virtualization, the harder it is to use these components independently, and indeed, hyperconverged systems aren’t really designed to be used that way.

There’s always vendor lock-in, too. Ideally, organizations want to negotiate prices for different components between different vendors. That becomes very difficult with modern hyperconverged systems. It also becomes difficult to upgrade different components at different types in a hyperconverged environment. You typically don’t just swap out discrete components with these things. They’re designed to hide their inner workings from you.

There are always trade offs in any system, though. Converged systems will nevertheless be good for a particular set of customers. Small to medium-sized businesses like them, because they take a lot of the gruntwork out of system building, but they can also be good for larger IT teams that want a quick fix for discrete projects. What kinds of project? Consider these:

• Virtual Desktop Infrastructure (VDI), in which multiple thin clients share a single back-end resource.
• Regional/remote office builds, creating a simple-to-manage, quick-to-deploy set of locally-accessible IT resources.
• Unified communication, for the provision of Internet telephony, videoconferencing, messaging and collaboration services in a single box.
• Analytics, for the processing of data from transactional systems.

It’s no coincidence that wherever you see ‘cloud in a box’ discussions, you’re sure to see a healthy dose of hyperconvergence peppering the conversation. They’re one and the same, and worth a look to see if they can easily solve a discrete computing problem in one fell swoop. ®