Archives ||  About Us ||  Advertise ||  Feedback ||  Subscribe-
 Home > Cover Story
 Print Friendly Page ||  Email this story
Standards and interfaces-A turf war

Different vendors tend to create proprietary standards in a bid to gain turf from competitors. And a gamut of available interfaces which evolve rapidly add to confusion among users. An overview of standards and interfaces may diffuse your worries. by Soutiman Das Gupta

While vendors and standards bodies battle for turf, its important that you select the one best fit for your enterprise

Technology standards, in an effort to 'standardize' have usually managed to confuse the manufacturers and users of the technology. Network storage

standards and interfaces are no exception. The 'standards folly' of the various available and emerging storage technology have resulted in some vendors pushing proprietary standards based gear into the market and others following whichever standard is popular at that time while screaming "I'm open."

Some companies have deployed a mix of storage standards, some follow a single standard and are not happy with it, and some have no clue about which one to follow. In case you are any of the above, the standards and interfaces explanations below may help clear some of the fog surrounding standardization worries.

EMC, HP, and other vendors formed an assemblage of storage vendors called the Fiber Alliance who proposed to implement SANs (Storage Area Networks) as FC (Fiber Channel)-based networks.

Elsewhere, an association called SNIA (Storage Network Industry Association) planned a proposal of its own. The Fiber Alliance was not interested in deviating from its stand on FC for SANs. In course of time, SNIA emerged as a more popular body for defining standards.

InfiniBand, another inter-processor communication standard developed by companies like IBM, Sun, Compaq and Dell is also in the market hoping to play a significant role. InfiniBand's aim is to overcome any bottleneck in systems design by increasing the speed with which data can move in and out of the processor.

Many interfaces
The various standards committees drive the adoption of interface protocols to allow any peripheral device that follows a particular standard to be used interchangeably. Here are some of the popular ones.

IDE (Integrated Drive Electronics), also known as ATA (Advanced Technology Attachment) or Ultra DMA is generally the least expensive hard drive interface. Many computer motherboards include ATA controllers and cable connectors that typically control the C-drive that contains the OS. ATA is a slightly slower drive interface, but is easier to implement and is priced low. This makes it a popular choice for desktop PCs and low-end RAID systems.

The original ATA interface was 16-bits wide and supported two hard drives at a maximum transfer rate of 8.3 MBps. ATA-2 boosted maximum throughput to 16.6 MBps for a maximum of two devices. From there, ATA designations blur as companies create names like Fast ATA or Extended IDE (EIDE) to describe proprietary feature additions to the ATA-2 standard. Accordingly, these designations are more marketing terms than official standards.

ATA/100 disks which are now available, offer data transfer rates of 100 MBps and improves reliability using the cable plant and CRC (Cyclic Redundancy Check) introduced with Ultra ATA/66.

Serial ATA, a new offering has lower signaling voltages and lesser pin count than the earlier parallel ATA. It will be faster, more robust, and have a much smaller cable. Serial ATA will also be completely software compatible with the earlier ATA.

SCSI (Small Computer System Interface) is widely used in mid- to high performance servers and storage devices and offers faster transfer rates than ATA/IDE.

SCSI generally offers faster throughput and uses less CPU horsepower during operation. This makes it more efficient in demanding multiple initiator applications for multi-users and uses. This is significant because it allows the processor to run more efficiently by making it perform more commands at the same time. SCSI can support up to 16 devices on a single bus and IDE offers only two.

The first SCSI standard, now known as SCSI-1, was adopted in 1986, and was originally designed to accommodate up to eight devices at speeds of 5 MBps. Since then, SCSI has been refined and extended numerous times, with the introduction of Fast SCSI (SCSI-2) at 10 MBps, Fast Wide SCSI (SCSI-2) at 20 MBps, and Ultra SCSI (SCSI-3 or Fast-20), which provides data transfer rates of up to 40 MBps.

The FC standard is designed to provide high-speed data transfers between networked devices. It makes use of a circuit/packet switched topology capable of providing multiple simultaneous point-to-point connections between devices. It provides powerful networking capabilities which allow switches and hubs to enable the interconnection of systems and storage into tightly-knit clusters. These clusters can provide high levels of perfor-mance for file service, database management, and general purpose computing.

FC is able to span distances up to 10 Km between nodes and allows very high speed movement of data between systems (up to 4 GBps) that are greatly separated from one another.

It can be deployed in point-to-point, switched topologies, or arbitrated loops (FC-AL). FC nodes log in with each other and the switch to exchange operating information on attributes and characteristics.

Point-to-point is the simplest topology connecting two FC devices that communicate at full bandwidth. A switch fabric is a very flexible topology which enables all servers and storage devices to communicate with each other. It also provides a failover architecture in the event a server or disk array ceases to operate.

FC-AL (Fiber Channel-Arbitrated Loop) is an enhancement to the FC standard that supports copper media and loops containing up to 126 devices, or nodes. Like SSA, FC-AL loops are hot-pluggable and tolerant of failures.

The FC-AL interface is robust enough to allow multiple devices to be removed from the loop at one time with no interruption in data transfer. In addition, the interface attaches sophisticated error

detecting codes to each packet of user data. These codes are checked at the receiver's end, which requests a re-send if there is any discrepancy.

IP storage refers to a group of technologies that allow block-level storage data to be transmitted over an IP-based network. The IP storage protocols provide the means to encapsulate these block-level requests for transmission over the IP network using TCP (Transmission Control Protocol). This allows the direct block-level requests used by SANs to take place over an IP-based network.

The IETF (Internet Engineering Task Force) is currently working on three IP storage encapsu-lation protocols. They are iSCSI (Internet SCSI), FC/IP (FC Over TCP/IP), and iFCP (Internet FC Protocol).

iSCSI is an emerging standard that defines the encapsulation of SCSI packets in TCP which are then routed using IP. This development allows block-level storage data to be transported over widely used IP networks. It enables data access from anywhere and effectively eliminates the physical boundaries of the storage network.

iSCSI enables block level data to be accessed over a standard Ethernet/IP network whether it resides on a direct attached SCSI-based device or a FC SAN. With iSCSI, enterprises and SSPs (Storage Service Providers) can build global storage networks and manage them from a central location using existing IP network infrastructures.

FC/IP transports FC frames over an IP infrastructure. It provides mechanisms to allow islands of FC SANs to be interconnected over IP-based networks to form a single unified FC SAN fabric. The extended FC SAN fabric continues to use standard FC addressing. Essentially, IP tunnels are set up between FC/IP end points. Once these tunnels are in place, FC devices view these extended links as standard FC links and use FC addressing.

iFCP encapsulates FC frames to be sent over the IP infrastructure just like FC/IP. Because of this, the IETF chose to specify a common FC encapsulation format. The main difference between the two protocols lies in their addressing schemes. The FC/IP protocol establishes point-to-point tunnels that can be used to connect two FC SANs with Ethernet and create a single, larger SAN.

SSA (Serial Storage Architecture) is a high-speed serial interface designed to connect data storage devices and other networked devices. SSA was developed and promoted as an industry standard by IBM.

Although the basic transfer rate through a SSA port is only 20 MBps, SSA is dual-ported and full-duplex to allow a maximum aggregate transfer speed of 80 MBps. SSA connections are carried over thin, shielded, four-wire (two differential pairs) cables. These are less expensive and more flexible than typical 50- and 68-conductor SCSI cables.

SSA networks can be constructed using string, loop, or switched topologies. SSA loops can contain up to 126 devices. Devices are hot-pluggable, and some measure of fault tolerance exists due to the redundant connectivity inherent in a loop topology. The devices can also be separated by much larger distances than possible on a SCSI bus.

While vendors and standards bodies battle for turf, its important that you select the one best fit for your enterprise and make the vendor promise to support the technology in future.

Soutiman Das Gupta can be reached at

- <Back to Top>-  

Copyright 2001: Indian Express Group (Mumbai, India). All rights reserved throughout the world. This entire site is compiled in Mumbai by The Business Publications Division of the Indian Express Group of Newspapers. Site managed by BPD