Have you ever had that helpless feeling when the mechanic tells you your car needs a new flux capacitor? You are not sure what a flux capacitor does but it sounds like it is going to be expensive. It's a pretty similar feeling to when your IT department comes to ask you for a lazy couple of million dollars to do cloud-this or Hyper Converged-that. In this article we'll help you navigate some of the current terminology around Hyper Converged and Web-Scale Computing.
The Googles and Amazons and Facebooks of the world started building their own massive computing platforms based on cheap, commodity infrastructure. The first part of Web-Scale Computing is to have massive farms of cheap infrastructure where it doesn't matter if servers fail. The second part is that you can just keep adding these cheap infrastructure building blocks as you need them. The third part is that you have some special software running over the top of your farm of cheap infrastructure to make it all work together. When Google talk about the Container for Data Centers it is not some fancy you technology, it is something that looks like a shipping container they drop into the data center and fill with cheap commodity servers.
To understand Converged and Hyper Converged infrastructure, it is useful to understand what non-converged infrastructure is. Basically non-converged is where your network and your servers and your storage all come in separate boxes, and are possibly all from separate vendors. The challenge for your IT department is to make sure they all work well together.
Converged infrastructure is where some of your infrastructure vendors get together to put their network and servers and storage together in a single box, making sure it all plays nicely together. There are some very good cost, performance and operational reasons to do this. Products like VBLOCK from VCE and Exadata from Oracle are really starting to gain traction with big corporates.
This is a relatively new market that goes one step further than Converged Infrastructure. These new-ish technology startups are not only putting your server, network and storage together in a single box but are also adding in technology from other devices that de-duplicate and compress data and optimize network traffic. This Hyper Converged infrastructure runs on commodity x86 hardware. As with Web-Scale Computing, these vendors also have special software that glues the systems together.
Obviously not every business can build Google or Amazon or Facebook sized Web Scale Computing platforms, however with Hyper Converged infrastructure businesses of any size can start getting some of the same benefits, using commodity x86 server building blocks in clusters that can easily scale out and provide a single pool of network, server and storage resources to run systems on. New Vendors playing in the space include SimpliVity, Nutanix and Scale Computing.
Some of your data is used every day, some is almost never touched. Instead of having all of your data on one type of storage, you have the data you use every day on faster and more expensive storage, and the data you almost never use on slower and cheaper storage. When you stop using data all the time it moves to the cheaper storage, and when you start using data more it moves it to the faster storage.
This term was coined in the mainframe era when an upgrade involved using a forklift to remove the old mainframe and bring in a new one. It has a similar meeting nowadays, where a system is difficult to upgrade because you have to move it to new server platforms and re-install and upgrade your software.
Hardware you buy just for the name. The vendors sound like they are big and safe, and it is what everyone else is buying.
In practical terms this means adding in extra building blocks of infrastructure when you need more capacity. In the context of this article, Scale-Out is desirable as you can easily add cheap, commodity capacity when you need it, as opposed to trying to Scale Up your existing systems that may have no more room to grow and will possibly require outages to your systems to be upgraded.
You might remember on your old computer you used to have an operating system like MS-DOS where you would have to go type in commands to do stuff. Along came Windows and your computer looked slicker, you could just click on things and you didn't have to be as clever to get your computer to do what you wanted. Software-Defined infrastructure is a little like that, it hides a lot of the complexity of your infrastructure systems and makes them easier to use.
For example, where you previously had a highly qualified (and highly paid) Storage Engineer working on your network fabric, assigning LUNs and controlling data replication policies you now have a simple software layer that someone less skilled can operate, and hides the complexity beneath.
Of course that is not all Software Defined is. Software Defined also pools, virtualizes and controls your network, server and storage resources. It allows you to aggregate, divide up and control your IT resources.
North-South traffic is data that passes from a server in your data center out to a client computer, for example a laptop on someone's desk. East-West traffic is data that passes between servers in your data center. The reason East-West traffic is important in the context of this article is that server virtualization and cloud computing is seeing an increase in large-scale movement of data between servers.
Traditionally servers would run a single dedicated operating system. Sometimes these servers would be very lightly utilised, using only a few percent of their total CPU capacity. In an effort to get better use out of these servers, software was developed to allow you to run multiple instances of an operating system on a single server.
It's a bit like the difference between cars and buses. You can use a car whenever you need it and don't have to share it with anyone, but a lot of the time it sits idle in your garage. With a bus, multiple people can use the same transport and it is on the road most of the time. Because of the utilisation, bus trips are cheaper than using a car.
x86 is used to denote an Intel CPU, however in the context of this article it has come to be used as a term for cheap, commodity servers that are made by many different vendors and are based on an Intel CPU.