IBM Smells Blood With New Shark

Now at 2Gbps

• By Chris McConnell
• 07/16/2002

IBM Corp’s flagship “Shark” storage system just got leaner and meaner. The monolithic storage cabinet was refreshed Monday, adding performance-enhancing features such as 2 Gigabit Fibre Channel.

The new models, designated as 800 and 800 Turbo, are the latest version of the Enterprise Storage System, which debuted in 1999. It competes directly with EMC Corp.’s Symmetrix and Hitachi Data Systems’ Lightning storage arrays for the backend storage market.

Perhaps the most significant new feature available on Shark is its support for Fibre Channel and FICON running at two gigabits-per-second, which IBM says is a first for an enterprise storage array. The improvement over the previous 1Gbps wire speed increases the number of requests a database server connected to the system can field and reduces the amount of time needed to back up.

Fibre Channel and FICON are two standards for moving block-level data over an optical connection. Fibre Channel is based around the SCSI protocol, while FICON – primarily used in IBM’s proprietary servers - is descended from IBM’s ESCON standard. Both Fibre Channel and FICON can use the same switches, plugs, and cables, and John Power, product manager for Shark at IBM, says the two protocols can now run over the same network, obviating the need for dedicated FICON and Fibre Channel networks.

Shark’s improved performance is most apparent in backup-and-recovery duties. Power says a Flash Copy, which saves data from a single point in time, is 50 percent faster on the new system, while a Real Copy, a full backup of a static file, is 30 percent faster.

Another new feature, introduced in May, is Peer-to-Peer Remote Copy – Extended Distance (PPRC-XD), which helps enterprises prepare for catastrophic events such as the destruction of a building. PPRC-XD allows a Shark to maintain a complete copy of itself on another Shark in a different location. While Fibre Channel and FICON have a distance limitation of about 10 kilometers, IBM has partnered with hardware vendors and telecommunications carriers to create services for long-distance Shark copies.

Power says with 2Gbps connections, synchronous copies like PPRC-XD are as much as 125 percent faster than 1Gbps configurations.

IBM also added a number of self-healing features from its eLiza initiative to the latest Shark. The new call-home feature in the unit compares the current state of a Shark to historical data of Sharks shipped to ascertain whether the machine is performing normally or not. If the Shark does not appear to be stable it will proactively “call home,” alerting an IBM service rep to potential problems.

The new Shark also adds two system boot drives, for a total of four. The additional boot drives can keep the system going in case an upgrade of system software goes sour. An old, known good version of the software is kept on one set of drives, while the new software is added to the other. If the new software doesn’t work, the system can still be booted from the drives with the old version.

Finally, the new stability features allow for priority queuing when backing up data. Shark can detect at the application level which data is most important for a backup and copy that data first. In backing up a database, for example, Shark will find the index file and save it first, then back up the rest of the database.

Power says about a third of Shark customers use it exclusively for zSeries and S/390 machines, another third use it for open systems running Unix and Windows, and the rest have a mix of mainframes and open systems.

IBM expects Model 800 will be available next month.