Enterprise Storage Solutions

Data is the most important asset of any organization, and must be stored efficiently and securely. There are three broad techniques for storing data: DAS, NAS and SAN. Another concept has emerged, recently, called storage virtualization. Here, we’ll try to cover the three topologies with a brief insight into virtualization.

DAS (Direct Attached Storage)



Direct Attached Storage or DAS is the most common and basic form of storage among the three options. It generally comprises of storage devices that reside within the network server or are connected directly to it through a separate cable using specialized bus adapters. The storage medium could be hard disks or tape drives. These disks can be put into various configurations such as JBOD (Just a Bunch of Disks) or RAID (Redundant Arrays of Independent Disks). While JBOD is a set of separate and independent disks, RAID is an architecture that typically includes 2 to 8 disks for faster access, data striping and fault

tolerance.

The most common types of media and protocols used in this scenario are Small Computers System Interface (SCSI- pronounced as “skuzzy”), Fibre Channel and Serial Storage Architecture (SSA). SCSI is a high-speed reliable interface and can carry data up to 160 MBps, but it has limitations such as limited scalability and distance constraints. Fibre Channel, on the other hand, uses serialized data transmission over either copper or fibre optic media. The storage devices can be attached to Fibre Channel enabled servers through point-to-point topology. SSA uses the serial SCSI-3 protocol to connect networks of disks within the same system. 

DAS is easy to install and manage when compared to other storage technologies. It is also one of the cheaper storage option if not the cheapest. Other benefits of DAS include good performance and flexibility, that is, it can be attached to servers on a multitude of platforms. With benefits,however, come limitations and DAS is no different. DAS systems are not scalable enough, and can connect to one server only. If that particular server is down, the storage is not accessible to users. This also results in under-utilization of resources since the storage capacity available to one server may not be available to the other server. These limitations, and a bid to overcome them, led to the development of NAS.

NAS (Network Attached Storage)



NAS unlike DAS is more of a device and less of a part of the network infrastructure. Typically, a NAS device may consist of a server and a stripped down version of an operating system optimized for storage; and a storage that is accessible and shared across the network. NAS devices are like the Plug-and-Play hardware in your machine. Just plug them into your existing LAN and they’re ready to store. They support all major network protocols and, hence, are absolutely apt for heterogeneous networks. Unlike DAS and SAN these devices use file I/O instead of block I/O. The NAS server receives the request as a file to be accessed and its offset rather than an address to the storage device. The OS on the NAS servers carries out this

function. 

NAS provides various benefits including ease of implementation and the fact that it allows organizations to exploit their existing network for storage purposes. Other benefits are greater scalability that can be achieved by adding NAS devices when needed, easier backup options that can be built in into the NAS server and easy manageability. The flip side, however, is that the networks on which NAS is used are not optimized for storage, and usage of the same network for storage may cause serious network congestion. 

But the NAS system is also not foolproof. Some major limitations include problems in data integrity since the Ethernet protocols are designed for messaging applications and data packets may be dropped in times of busy traffic, resulting in additional traffic. These storage devices are also not suitable for the I/O demands of some database applications since these applications work on block I/O.

SAN (Storage Area Networks)



Storage Area Network or SAN is a dedicated, high-speed network behind the servers for storage. Various servers and storage devices may attach to SAN making it an any-to-any network. A SAN generally works on a Fibre channel or SCSI network. SAN allows heterogeneous servers to share a pool of storage devices that can either be in the same vicinity or kilometers apart. Since SAN is a separate network in itself, it does not affect network traffic as much as NAS does- this makes the traffic really fast. Other benefits of SAN may include easier and improved backup and recovery options, and even easier scalability than NAS. SAN can be implemented using two topologies over the Fibre Channel network: Arbitrated Loop and Switched Fabric.

An Arbitrated Loop can be best compared with a uni-directional token ring topology. It supports up to 126 Fibre Channel enabled devices. In this topology, the devices send a signal to indicate that they want to gain control of the medium. Once a device gains control the others have to wait till the I/O cycle of the device is not over. All the devices share the same available bandwidth, which is a major hindrance. It is generally implemented from small SAN configurations.

Switched Fabric, on the other hand, is a switching topology in which the devices do not need to share the bandwidth until they need to access the same storage device. Thus, it is much faster than the arbitrated loop and gives better performance.

The biggest hindrance in SAN implementation is the cost — of the equipment involved, Fibre Channel switches, hubs etc. as well as the costs of maintaining virtually another network. The latest in the list of standards that have been discussed above is iSCSI (Internet Small Computer Systems Interface). It became an official standard in February 2003. This claims to eradicate the problems faced by SAN and Fibre Channel implementations, namely, cost and access to remote devices. iSCSI is an IP-based protocol that encapsulates SCSI commands in TCP/IP packets and enables I/O block data transport over IP networks. iSCSI aims at bringing the functionality of SAN to the current LAN networks to provide a high degree of interoperability.

Geetaj Channana