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.
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.
Extending the Network
IP storage represents a relatively new class of storage subsystem. It uses available disk technology that communicates via an IP address. This is a new implementation, but several vendors have rushed products to market. In the near future, observers expect vendors to blend IP storage with NAS and SAN subsystems.
The approach offers several benefits, including the fact that it is not limited by physical configuration. However, it will require host-processing power, and questions remain on the speed.
IT managers may feel caught. Promising storage technologies will take time to develop standards. The lower cost, network attached and directly attached subsystems will only provide immediate relief. Some, as a result, use storage service providers. These companies essentially offer on-line storage and data management. Although IT managers express concern over this type of arrangement, it does offer a solution that may make financial sense.
As the demand for storage explodes, the search for storage solutions will continue. However, users frequently focus on the initial cost of storage subsystems without considering implementation and on-going forget to calculate the total cost of introducing new storage technology and expanding existing storage. At the very least, a change in storage will impact backup systems. Larger SANs systems require independent administration software and personnel to configure and maintain the devices.
In addition, as storage systems expand, the task of securing the subsystems against intruders and ensuring reliable access becomes more complex. IT managers can expect improved security and management utilities as this market grows.