Storage over the Internet - iSCSI emerges

Storage is an essential part of any company's infrastructure. Rapidly increasing storage capacities and network speeds challenge storage system performance, whether it's at the enterprise level or below. IP storage - sending block-level data over an IP network - addresses the requirements of a range of environments from single server to computer room, Internet data center, campus and WAN.

The Internet Engineering Task Force (IETF) is currently considering two approaches to transmitting storage data over IP networks. Tunneling translates Fibre Channel control codes and data into IP packets for transport from one Fibre Channel storage-area network to another remote Fibre Channel SAN. Storage tunneling assumes the use of Fibre Channel technology.

---

How it works
Subscribe to the Tech Update newsletter
  Here is a weekly newsletter to help you stay abreast of new networking standards and technologies by providing down-to-earth explanations of how they work.

---

The other approach is IP storage, known as iSCSI, which transmits native SCSI over a layer of the IP stack. ISCSI will let a corporate network transfer and store SCSI commands and data at any location with access to the WAN or, if transmitted over the Internet, to locations with access to the Internet. It will also allow smaller localized SANs to be built using the common Ethernet infrastructure. Hence, iSCSI enables SANs to be implemented by a broad, mainstream market.

SCSI is a commonly used industry standard protocol for storage devices. Using the SCSI protocol, drive control commands and data are sent to the drives. Responses and status messages, as well as data read from the devices, are passed through SCSI controllers.

In a system supporting iSCSI, a user or software application issues a command to store or retrieve data on a SCSI storage device. The request is processed by the operating system and is converted to one or more SCSI commands that are then passed to software or to a card. The command and data are encapsulated by representing them as a serial string of bytes proceeded by iSCSI headers. The encapsulated data is then passed to a TCP/IP layer that breaks it into packets suitable for transfer over the network. If required, the encapsulated data can also be encrypted for transfer over an insecure network.

The packets are sent over the network or the Internet. At the receiving storage controller, the packets are recombined and, if necessary, decrypted into the original encapsulated SCSI commands and data. The storage controller then uses the iSCSI headers to send the SCSI control commands and data to the appropriate drive, which performs the functions that were requested by the original computer or application.

If a request for data has been sent, the data is retrieved from the drive, encapsulated and returned to the requesting computer. The entire process is transparent to the user.

Although the SCSI commands and data preparation can be accomplished through software using standard TCP/IP and off-the-shelf network interface cards, a more likely scenario will involve specialty iSCSI interface cards that will perform this function. In a scenario that involves encapsulation, decapsulation and TCP/IP through software, performing these functions on the host processor requires many CPU cycles to process the data and SCSI commands. By moving these functions to specialty hardware, impact on system performance can be minimized.

An initial draft of the iSCSI standard and early iSCSI products are expected during the first half of 2001. This month, companies participating in the IETF's standards development will meet with the goal of advancing progress toward a standardized method of transferring data and SCSI commands between computers and storage devices, using IP networks.

The pervasiveness of IP networks will let data be moved over LANs, WANs or across the Internet using this new IP storage protocol.

Ross is marketing manager of IP storage products at Adaptec. He can be reached at brent_ross@adaptec.com