There are many ways to provide access to storage across the network. In this video, you’ll learn about Fibre Channel, iSCSI, InfiniBand, and more.
<< Previous Video: Virtual Networking Next: WAN Services >>
In a previous video, we talked about the differences between a NAS and a SAN, but let’s review since we’re going to be going through a number of these SAN technologies in this video. NAS stands for Network Attached Storage. This allows you to take a storage device and connect it to any [INAUDIBLE] network. Usually, a NAS provides file-level access. That means if you need to change just a single character inside of an entire file, you need to rewrite the entire file onto that NAS device.
A SAN is a Storage Area Network, and this is a type of storage that is a bit more efficient than network attached storage. That’s because we’re using block-level access with the SAN. This is very similar to the drive that would be in your existing computer. If you need to change a single byte within a file, you only need to change that block of data. You don’t have to rewrite the entire file onto that device.
Regardless of whether you’re using a NAS or a SAN, you’re going to be sending a lot of information across the network. So you would commonly connect these to high-speed networks that can support the bandwidth required for the data that you’re transferring back and forth.
A normal ethernet frame can support 1,500 bytes of a payload, but if we’re sending and receiving so much traffic to a storage area network, then we might want to increase the size of these frames. The way to do this is to enable jumbo frames on your network. This will allow you to transfer 9,216 bytes of data within a single ethernet frame, although it’s common to configure 9,000 so it’s more compatible with the 1,500 bytes of payload. Either way, you’re increasing the amount of efficiency by six times because you can fit, now, so much data within a single ethernet frame. This is going to increase the efficiency of the traffic we’re sending through the network. Not only is the data within the packets larger, we don’t have to make as many switching or routing decisions because we’re sending fewer packets through the network.
One catch is that all of the devices on your network between you and the storage area device have to support these jumbo frames. There’s usually a configuration option within your switch, and there’s usually a configuration you would set in your operating system to turn on the ability to send and receive jumbo frames.
One popular type of storage area network is Fiber Channel or FC. This is a high-speed typology that has been specifically built for storage area networks. It is built to connect servers directly to storage devices. It can support 2, 4, 8, and 16 gigabit per second rates, and you can run this traffic over both fiber and copper connections. Even though the name of it is Fiber Channel, it does support copper connectivity.
To use a Fiber Channel topology, you would usually have a Fiber Channel switch. So you would plug-in your server with Fiber Channel, and you would plug-in your storage device with Fiber Channel. The server would commonly be using some well-known protocols to communicate to those storage devices. You may be using SCSI, serial attached SCSI, or SATA commands to send data over this storage area network.
There are also types of Fiber Channel networks that don’t require any specialized hardware. You don’t need a Fiber Channel switch if you’re planning to use Fiber Channel over Ethernet. All of these Fiber Channel commands are being sent over the existing ethernet network. It’s commonly connecting over Ethernet to an existing Fiber Channel infrastructure, and since this is done at the ethernet frame level, this is not routable traffic. But this does allow you to use the ethernet card that’s in a device rather than using a Fiber Channel adapter.
If the storage and the devices that need access to that storage are located on different IP subnets, you may want to consider using Fiber Channel over IP. This is effectively tunneling Fiber Channel within existing IP packets, and that would allow you to route this traffic from one subnet to another and still be able to communicate to a Fiber Channel device.
Another type of storage area network is iSCSI. This stands for Internet Small Computer Systems Interface. You may be familiar with the SCSI drives that you would install with SCSI adapters inside of your computer. This simply extends that SCSI capability across the network. This was created by IBM and Cisco, and it’s now an RFC standard. This would make these remote drives on the storage area network look and feel as if they are local to our computer, very similar to how Fiber Channel works. And of course, SCSI is well-supported in software and operating systems, and there are drivers already available for iSCSI in many existing OS’s. You wouldn’t need any proprietary topology, there’s no need for any proprietary switches or adapter cards, and since you’re using IP, you’re using an easily routable iSCSI protocol over an existing ethernet network.
If you need very high speeds with your storage area network, you might need InfiniBand. This is a high speed storage topology that is very focused on speed, and it has its own switches and adapter cards, very similar to the way that something like Fiber Channel might be implemented. With an InfiniBand, we can connect to the storage area network with both copper and fiber, using quad SFP connectors. You would commonly see InfiniBand used with environments that needed very high speeds and very low latency. So for research and supercomputer operation, InfiniBnad networks are very common. It’s common to see 200 gigabit per second speeds with InfiniBand networks and you can then aggregate those together in for 4x, 8x, or even 12x links.