IP-SAN, the networked storage wave-of-the-future?

Computer Technology Review, June-July, 2005 by Ed Ginty

This article deals with a subject possessing definite lines of demarcation, strong favoritism and intense opinion: Networked Storage. Whether SAN, NAS or IP-SAN, no one networked storage technology is better than the other for every client and for all purposes, as each possesses its own unique attributes and advantages.

This article focuses on the newest networked storage technology, IP-SAN, which is gaining considerable momentum within organizations of all sizes. In many cases IP-SAN is the perfect fit as a total storage solution. In others, it is proving an excellent tool when used as an integral part of a more global, mixed-architecture infrastructure solution for improved backup/restore, disk staging and efficient/economical data replication.

Networked Storage Architecture--A Primer

Networked storage architecture has enabled organizations to effectively share, consolidate and manage data resources, more so than with DAS (Direct Attached Storage). The shift from server-centric to networked storage has been dependent upon technologies designed to transfer data to and from storage repositories as fast as, or faster than, direct attached technologies, while overcoming the limitations inherent in parallel SCSI, the main interconnect for DAS.

All data is stored on disks in block form (without file-system formatting), whether originating from applications as blocks (as in most database applications) or as files. Parallel SCSI transmits data to storage in block format with limited usefulness for networks because cabling is limited to 25 meters. In addition, it cannot connect more than 16 devices per network.

Fibre Channel is currently the dominant infrastructure for SANs. SAN technology places storage resources on a dedicated high-speed Fibre-Channel network. This protocol technology grew from the need to provide high-performance transfers of block data, as well as the need to overcome the connectivity and distance limitations associated with DAS. Common Fibre Channel installations connect devices up to a distance of about 10 kilometers. More costly solutions extend that distance to several hundred kilometers without imposing practical limits on the number of devices to which is can be attached.

Unlike SANs, NAS transfers data in file format over an IP network. NAS devices use the Server Message Block (SMB) protocol to transmit database block data. However, this is less efficient than using the Fibre-Channel SCSI-based protocol.

Is IP-SAN the Next Paradigm Shift?

IP-SAN represents a new generation of networked storage. It combines the best attributes of Fibre-Channel SANs (functionality and scalability) with the best attributes of NAS (ease-of-deployment and ease-of-use) providing virtually all the capabilities of Fibre-Channel SAN and NAS devices combined (file and block-level storage) but at significantly lower cost. IP-SANs scale to Petabytes in size and can be managed as one realm, locally or remotely. Like NAS, IP-SAN takes advantage of existing GB Ethernet network infrastructures, although for security and performance purposes it is usually implemented as a separate network off the main network infrastructure.

Like their SAN and NAS counterparts, IP-SAN offers useful and cost-saving capabilities, such as: dynamic storage allocation (expanding storage on the fly); RAID striping and mirroring (for increased performance and reliability); snapshots (for fast backups and restores, as well as near-production development); Replication (block-level--only the changes are replicated locally and to the remote site).

Key IP-SAN Components

1. iSCSI Client/Host

The iSCSI client or host (also known as the iSCSI initiator) is a system, such as a server, that attaches to an IP network and initiates requests and receives responses from an iSCSI target. Each iSCSI host is identified by a unique iSCSI qualified name (IQN), analogous to a Fibre-Channel world-wide name (WWN). To transport block (SCSI) commands over the IP network, an iSCSI driver must be installed on the iSCSI host (iSCSI drivers are included with Microsoft's iSCSI initiator). A Gigabit Ethernet adapter (transmitting 1,000 Megabits per second--Mbps) is recommended for connection to the iSCSI target. Like the standard 10/100 adapters, most Gigabit adapters use Category 5 or Category 6E cabling. Each port on the adapter is identified by a unique IP address.

2. iSCSI Target

An iSCSI target is any device that receives iSCSI commands. The device can be an end node, such as a storage device, or it can be an intermediate device, such as a bridge between IP and Fibre-Channel devices. Each iSCSI target is identified by a unique IQN and each port on the storage array controller (or on a bridge) is identified by one or more IP addresses.

Business Cases for IP-SANs

Reduced Buy-In and Scalability Costs

Well-designed IP-SAN solutions preclude clients from buying more infrastructure than is required. By separating controllers (the brains) from the disk shelves (the data repositories), end-users can design solutions that fit their current requirements without having to invest prematurely. As additional performance is required, controllers are added. These controllers scale linearly in performance and load-balance across the entire realm. (Should only storage capacity be required, additional drive shelves are added without the need to purchase additional controllers or chassis and without the need to bring the system down.