ISCSI

11/18/2002

iSCSI (Internet Small Computer System Interface) is, like FCIP, a technology that allows a storage device to communicate with a host using the Internet Protocol. This permits mass storage to be plugged into a network rather than requiring it to be physically added to a computer or to a server. iSCSI has been defined by the Internet Task Force (IETF) rather than any individual company. iSCSI uses the SCSI command protocol wrapped in IP packets to control storage devices and move data. iSCSI is projected to be cheaper albeit generally slower than Fibre Channel IP. iSCSI has been demonstrated at 200Mbit/sec using Gigabit Ethernet.

Because IP packets are used, data can be moved using standard network wiring and routing. Adding or moving storage is -- conceptually at least -- merely a matter of plugging or unplugging a storage unit on the network. While the intent is to use it primarily with physically nearby units, there is no limitation other than speed of transfer that prevents iSCSI from being used with storage devices at remote locations..

iSCSI places communication modules between an operating system and network and between the storage device an network. These handle packet creation, assembly, transmission, and optionally encryption/decryption. Unlike NICs, there is minimal CPU loading because protocol drivers operate in the Host Adapter firmware rather than the server CPU. Encryption (IPsec) is an issue because there are problems interfacing it with hardware compression techniques used to speed up data transfers.

iSCSI data rates are limited primarily by the transport medium.

An iSCSI Version 1.0 standard has been agreed to and is in the approval process as of September 2002. Several companies are working on iSCSI devices compliant with the 1,0 standard. Companies including IBM and Cisco have shipped iSCSI devices already. It is unclear if they comply with the iSCSI 1.0 spec.

Return To Index Copyright 1994-2002 by Donald Kenney.