Evolution of Enterprise Storage

What’s your company’s most important asset? The obvious answer is data, as a company can be in serious trouble if anything goes wrong with it. Hence the importance of storing data efficiently and securely. Data storage techniques can be summed up into three topologies–DAS, NAS, and SAN.
Let’s try to understand and evaluate them.

DAS

DAS or Direct Attached Storage comprises storage devices such as hard disks or tape drives connected directly to a network server. It could be inside the same box as the server, or reside outside and be connected through a cable. This is the most basic method of storage, and the most widely used, as it’s simple, and doesn’t require too much time to set up. A DAS setup can have a single hard disk, multiple independent hard disks called JBOD (Just a Bunch of Disks), or an array of hard disks configured for fault tolerance, known as RAID (Redundant Array of Independent Disks). The most common types of media and protocols used in DAS are SCSI, fiber channel, and SSA (Serial Storage Architecture).

DAS is easy to implement, and gives a reasonably good performance. It also has relatively lower acquisition and administrative costs. However, it has its limitations. Its capabilities are limited to the server it is connected to. So if the server is down, users can’t access their data. It can’t be connected over long distances, and the storage capacity is also limited compared to NAS and SAN. DAS devices also pose management problems, as you have to manage the data on a server-by-server basis.

NAS

An attempt to overcome problems associated with DAS led to the development of NAS (Network Attached Storage). Unlike DAS devices, which connect to a particular server, NAS devices connect directly to your existing Ethernet network and are independent of the server. So, if the server fails due to some reason, data is still available to users on the network.

NAS devices can be placed anywhere on your network and can be accessed by any user. Its function is similar to that of a network printer, which can be accessed by anyone when required. NAS devices support most network protocols such as NFS for UNIX, CIFS for Microsoft, FTP, and HTTP. This makes them useful for heterogeneous networking environments.

NAS devices comprise a number of hard-disk drives and come with their own OS and management software. They can be used for a variety of applications including Web caching for proxy servers, backup, databases, print spoolers, or simply as file servers. Storage capacities for NAS devices can range from 2 GB to over 2 TB (terabytes). Another benefit of NAS is that it allows organizations to set up a storage solution using their existing Ethernet backbone, without investing in a separate network.

On the flip side, since NAS operates on networks that were primarily designed for data transmissions and not storage, performance issues such as network congestion arise. Also, if you exceed your storage capacity then you must add another NAS device.

SAN

The latest technology in the area of storage is SAN, or Storage Area Networks. As the name suggests, a SAN is a separate network linked to your company’s main network via a high-speed interface like a fiber channel or SCSI. This sort of a solution is useful for companies having high transaction volumes like banks and customer-service oriented organi- zations who need quick access to data at any point of time.

A SAN consists of multiple storage systems and servers and is much faster than a NAS system. That’s because unlike NAS, the various storage devices in a SAN are connected through a high-speed interface. Fiber channel typically has a data transfer rate of 1 GB/sec. SCSI transfer rates are lower than fiber-channel, about 80 MB/sec. However, in future these devices are likely to communicate over Gigabit Ethernet using iSCSI or Fiber channel over TCP/IP (FCIP). (For details read the New-age Storage article, page 48, in this issue).

The various storage devices in a SAN interface with the company’s main network via switches and hubs, and can be simultaneously accessed by multiple servers and computers. All SAN components are controlled using SAN management software, which allows users and system administrators to remotely control its functioning.

SAN architectures

Fiber-channel SANs can be deployed in two topologies: Arbitrated loop and Fabric, also called switched fabric. The arbitrated-loop topology can be compared to the token ring topology and enables you to connect up to 127 fiber-channel enabled devices. However, instead of sending a token, devices in this topology send a signal that tells other devices that they want to gain control of the transmission medium. When one device gains control, other devices have to wait until that device’s I/O cycle is complete. The main disadvantage of this topology is that servers connected in this topology have to share the available bandwidth.

On the other hand, in switched-fabric topology, servers share the bandwidth only when other servers need to access the same storage device. However, if each server needs to use a separate device, it can make use of the entire bandwidth. This topology provides a virtual dedicated connection between the network servers and the SAN, and lets you interconnect up to 64,000 devices. The switched-fabric topology offers better SAN performance than the arbitrated-loop topology. 

Different SAN solutions are available from different vendors. Some offer complete solutions with all necessary products, such as hubs and switches. Others, however, provide SANs that include only the hard disk drive enclosure with the specialized OS. You need to add other SAN components such as bridges, hubs, and switches later on. (For more information, read the next article on the components of SAN.)

Neelima Vaid