The move to a virtualized data center has brought an increase in storage area networking. In this video, you’ll learn about NAS, SAN, jumbo frames, Fibre Channel, iSCSI, and other storage area technologies.
<< Previous: VirtualizationNext: Cloud Technologies >>
Now that we’ve implemented these virtualized environments and our software-defined networks, we’ve also found that we need a storage mechanism that is also network-based. In this video, we’ll look at a number of options for storage area networks.
When we start talking about storage area networks, there are two major terms we’ll hear about. One is NAS and the other is SAN.
NAS stands for Network Attached Storage. This is a storage mechanism that you are connecting to the network, usually with an ethernet connection. And in the world of storage on the network, this is the most basic kind. This is what we call file-level access. If we need to gain access to a file, or even change a part of the file, the entire file has to be transferred down so we can modify the contents and then put the entire file back on the storage device.
As you might imagine, that file-based access is not the most efficient. And that’s why generally with our virtualized environments, we’ll see things like Storage Area Networks, where you have something called block-level access. When you need to change part of a file, you simply change the blocks on that storage device, so you can minimize the amount of traffic that’s transferred through the network. And you can make those changes very efficiently while keeping everything on that storage area network.
These storage devices are generally sending and receiving a lot data. They’re often put behind file servers and application servers. And there’s a large amount of communication into those storage devices. That’s why we usually will build an isolated network that only data going back and forth to our storage area network is going to use these very high bandwidths.
The original ethernet specification limited the ethernet frame size to 1,500 bytes of payload. That’s the data that you’re putting inside of an ethernet frame. So if we’re sending information back and forth to the storage devices, we can only put 1,500 bytes in every single frame that we’re sending. It would be much more efficient if we could increase the size of that payload so we didn’t have to send a lot of small frames. We could send fewer very large frames, and have a much more efficient way of communicating between those devices.
This is why we created jumbo frames. Jumbo frames allow us to increase the maximum size of the payload to 9,216 bytes. On most of the devices that you’ll use and configure, you’ll see 9,000 bytes as, generally, the maximum. And that’s a good number that all those different devices can use as a standard for maximizing this data transfer.
Now that our frames are bigger, we’re transferring a lot more with a single frame. So there’s going to be fewer frames that have to go back and forth between the routers and the switches and the other devices on the network. And that makes our file transfers that much faster.
Whenever you’re configuring jumbo frame support, all of your network devices between those two points have to also support these jumbo frames. So if you’re going through a switch that doesn’t understand jumbo frames, those frames will not be passed through the switch at all. So you want to be sure, if you’re implementing a jumbo ethernet configuration, that you have configured your operating systems to take advantage of those frame sizes. Your storage area network has to be able to take advantage of those frame sizes. And all of the switches and routers in between must also be able to pass these jumbo frames.
One common way to connect our computing devices to our storage is through a very specialized network called a Fibre Channel network. Fibre Channel networks come in different flavors, but they have very high speeds relative to our 1- and 10-gig ethernets. Fibre Channel will run at 2-, 4-, 8-, and 16-gigabit per second rates. And it will run over fiber or copper. Although the name says Fibre Channel, it’s one that supports both of those types of physical connections.
Just as you would connect many devices together with an ethernet switch, you would connect these devices together with a very specialized Fibre Channel switch. So you need a Fibre Channel interface in your servers. These are called initiators. And you also need Fibre Channel interfaces in your storage devices. And you would usually communicate to those storage devices using a very standardized form of communication– something like SCSI, serial attached storage, SATA commands, or some other very common way of referencing the storage on that storage device.
You connect all of them together over fiber or copper to your Fibre Channel switches. And they would use those higher Fibre Channel bandwidths to be able to send that data back and forth.
Although this Fibre Channel infrastructure gives you a lot of speed, not all of your devices are going to have Fibre Channel interfaces. And that’s why we’ve created other ways of integrating with Fibre Channel networks over an ethernet network. One of these ways is using Fibre Channel over Ethernet, or FCoE. This is a way that we can use our standard ethernet interface to be able to integrate to an existing Fibre Channel network. This is one that allows us to talk Fibre Channel over these ethernet frames. And it’s something that’s not using a routable set of protocols. So we have to be on the same network as the Fibre Channel network to take advantage of Fibre Channel over Ethernet.
One way to get around the non-routable nature of Fibre Channel over Ethernet is to take all of your Fibre Channel information and put it into an IP packet. And we do that by using Fibre Channel over IP. Everything that would normally be sent inside of that ethernet frame for Fibre Channel, we simply put inside of an IP packet inside of a frame. When we do this, we’re now able to route that anywhere that we can route IP. So we can now really separate geographically our in-station devices from the storage devices themselves, and send all of the Fibre Channel information over the IP packets.
Another way to connect our storage area networks is using something called iSCSI. This stands for Internet Small Computer Systems Interface. You may be familiar with the SCSI interface standard. And although it’s difficult to find physical hardware devices with the old SCSI interfaces on them, SCSI is a very well established standard. And it’s one that we can use across the network to be able to communicate in a language that storage devices can understand. This is an RFC standard that was created by IBM and Cisco, and allows us to take these devices and disks that are on a remote device and be able to use them as if they were locally connected inside of our computer.
We’ve now effectively extended the SCSI devices that used to all be inside of our computer, and taken those protocols and put them on the IP network. Now, we’re able to load new drivers onto our systems so that we can communicate via iSCSI and send all of that information over a very standardized ethernet network. We don’t need specialized storage devices. We don’t need specialized switches in the middle. There’s no extra hardware that we would need to add to our workstations. We can simply use our ethernet network and communicate via iSCSI.