Download the authoritative guide: Enterprise Data Storage 2018: Optimizing Your Storage Infrastructure
Using the Network
Network-attached storage (NAS) subsystems offer an interesting, cost-effective method of expanding storage. These devices look like a separate server on an IP network. The devices support an independent processor and operate under a shell of an operating system. Typically, NAS devices store data as application files rather than breaking it into blocks of data, and vendors now offer models with hot-swappable power supplies and redundant features designed to improve reliability.
Although NAS devices are growing popularity, it is slower than the larger Storage Area Networks (SANs) favored by large enterprises. Further, NAS subsystems are less expandable than other storage solutions.
Going It Alone
For many large companies, SANs represent the best approach to consolidating data and improving performance. The Fibre Channel subsystems operate as a fast network that bridges to the servers. This permits managers to control data flow, and the Fibre Channel technology provides performance benefits. Fibre Channel standard defines a high-speed serial interface that supports data transfers up to 100M bytes per second, with an increase to 200M bytes expected in the near future. In contrast, RAID subsystems use SCSI connections that operate at 20M bytes per second or Ultra-SCSI links that run up to 40M bytes per second.
SANs generally transmit very large blocks over the Fibre Channel, allowing the subsystems to send data without generating multiple interrupts. Fibre Channel the best suited for programs that require large data transfers, including video and multimedia applications. IT managers also tout the fact that SANs are scalable to meet future storage demands.
Fibre Channel offers a layered architecture similar to networks with a protocol-mapping layer that supports network, common services, framing, encode/decode, and physical layers. It offers three service classes:
- Class 1: Defines a temporary connection for the communications session. This requires that packets arrive at the receiving device in the proper order.
- Class 2: Creates a switched connection and allows packets can be received out of order. The configuration requires acknowledgment from the receiving device when blocks of data arrive.
- Class 3: Operates like Class 2 services but does not require acknowledgments.