CrossNodes Briefing: Network Storage
Demand for storage space continues to increase. Users incorporate more graphics, audio, video, and animation in their documents, and customer databases and related files grow. At the same time, companies also need to capture and leverage data with such applications as data mining and CRM to increase revenues. The benefits are compelling, but it demands storage space.
The cost of storage is dropping, but IT managers also are seeking a better strategy that will support immediate storage requirements and allow expansion in the future. Vendors now offer an array of options that the manager must consider before investing in additional disk storage. These options include directly attached storage, network-attached storage (NAS), storage area networks (SANs), and IP connected storage (SoIP). Each type of technology offers benefits and risks that must be carefully considered by the manager.
Attached and Ready
The simplest method of expanding storage is to add a drive directly to the server. With the escalating density of disks and the falling price-per-byte of storage, many companies find this a cost-effective, immediate solution to storage problems.
Most network servers implement attached storage in a Redundant Array of Independent Disks (RAID) subsystem that incorporates several disk units into a single drive. The subsystem can increase disk access times by writing 'stripes' of data across several disks, and the server can access data in parallel. In addition, the RAID configuration can support full or partial redundancy. Seven levels of RAID exist, with Level 6 offering the most security.
- Level 0: Stripes data across all disks. This configuration does not provide error correction or redundancy.
- Level 1: Maintains a copy of each disk on another disk in the array. This is also called mirroring, and this configuration provides the highest level of redundancy.
- Level 2: Stripes data across multiple disks, but it also logs error-checking code to recover from faults.
- Level 3: Maintains data across multiple disks, but saves data as bytes on several disks rather than stripes of data. Parity information is stored on another disk.
- Level 4: Operates like Level 3 except it stripes large blocks of data across several disks instead of bytes. Parity information is stored on another disk.
- Level 5: Like Level 4, it stores blocks of data on several disks, but parity information also is spread across multiple disks.
- Level 6: Implements a Level 5 configuration, and includes such redundant equipment as controllers, power supplies, and fans.
This provides alternatives, but it does not represent a scalable answer. At some point, the server will reach a maximum capacity, and the manager will be forced to consider an alternative approach. Further, directly attached subsystem require processing cycles from the server to operate.