While configuring system architecture IT professionals may not give storage the recognition and importance it deserves. In turn, this could affect performance and future costs. Today enterprises realize the value of storage and go in for enterprise SAN as a cost-efficient, reliable way to store and access data.
Storage can be direct-attached storage, network-attached storage or storage area network (SAN). SAN lies in between DAS and NAS; DAS forming the underpinnings of the storage system and NAS overlaying SAN.
Enterprise SAN takes the concept a step further in the form of centralized storage under common management and security but connected to a number of computer systems to permit data sharing. Enterprise SAN arose due to the need of large enterprises to handle heavy loads and the system can be scaled to handle over 300 GB workloads through unlimited connectivity and support for different platforms.
Enterprise SAN implementation results in better availability, better disaster recovery, improved data sharing and reliable backups under a unified administration. Enterprise relies on the multiple pathways of SAN to ensure that data is always available despite the failure of a server.
A typical enterprise SAN may be made up of a Cisco MDS 9710 multilayer director in large scale networks. The CISCO MDS9710 introduces a layer of intelligent features over a protocol-independent switch layer leading to higher flexibility, availability, security, scalability, and ease of management.
The Building Blocks Of SAN
SAN allows multiple servers to access a single storage device but not simultaneously. Another server can take over control, a useful feature when a server fails. SAN relies on various technologies such as:
Fibre Channel: Typically fibre channel SAN may use fibre channel though not necessary since the network can use ATA or SCSI drives as well. For longer distances, 1, 2 or 4-gigabit fibre optic cabling is employed. FC storage devices are connected to a server through FC switches that allow another server to access the array if the current server fails. Performance is high but expensive.
iSCSI: Internet SCSI is lower in cost and allows easy management of connections since it employs TCP/IP protocols and Ethernet switches. Apart from lower cost, it is easier to route connections through multiple servers connected to a single Ethernet switch over networks, useful for data mirroring and data disaster recovery in enterprise SANs. Drawbacks are high CPU utilization, expensive network card and cost of encapsulating SCSI protocol into TCP. iSCSI targets or servers with appropriate software can operate on any server OS permitting the use of a variety of servers that can also serve as NAS devices.
ATA Over Ethernet: This is a low-cost alternative to iSCSI, also known as AoE, encapsulating ATA into low-level Ethernet and bypassing TCP/IP. There are no CPU overheads or the need for high-end Ethernet adaptors which makes it better compared to fibre channel or iSCSI. The downside is that it cannot be routed, which is not much of a problem since software usually integrates iSCSI and AoE automatically using LAN/SAN for high throughput with iSCSI for low bandwidth use.
Writing On The Wall For Enterprise SAN?
Enterprise SAN, since it is so widely deployed, is not likely to die out any time soon but the writing on the wall is there. It needs to be trained staff to configure the complex system and costs of operation are high. Increasing preference for open source free OS like Linux that does not encourage high-speed file systems is another limiting factor.
Enterprises are looking at models that need less monitoring and administration of hardware as can be found in the appliance model. At the same time, the rise of big data calls for a rethink of traditional ways of handling storage and application-based handling seems far more attractive. A typical example is Oracle’s ZFS storage appliance utilizing integrated hardware-software architecture that permits multithreading storage and efficient use of hardware.