Different computer platforms are very different in the way they store data. Among other, differences concern data encoding, file systems, file structures, file naming conventions, and disk recording format. In a heterogeneous environment where platforms are by definition different, a distinction must be made between storage sharing, data copy sharing and True Data Sharing

In Storage Sharing, a storage subsystem attached to multiple computer platforms is divided into partitions, each partition is only accessible to the owning platform or to a certain number of homogeneous platforms. Storage capacity can be reassigned to different platforms as needs change. Some advanced form of storage sharing allows to access storage partitions owned by other platforms, only to manage the storage in that partition. This advanced storage sharing is used to copy files for backup or other purposes. Storage Sharing is performaed by several storage systems suppliers as CLARiiON, EMC, Sun, and MTI.

Data Copy Sharing involves replication of the data. One approach is to copy the data across the attachment mechanisms used to connect storage subsystems to the processor. Another approach consists in translating data in the expected format in a controller. Data is the same at the particular point in time the copy is created, but the two copies can change independently from there on, so there is no assurance that they remain identical after the copy is created. Access to data is usually prevented during replication so the copy accurately reflects all data at a particular moment of time. For large amounts of data, the time needed to copy data may be significant, and the amount of storage necessary to store the copy could be very important. Copying speed is an important factor. Some implementations may require several copy steps to get the data onto the required format. New techniques have been developed to allow to limit the time of data unavailability. An example of Data Copy sharing with high speed copy has been implemented by EMC.

If data is being shared without a copy being made of the data, then multiple computer platforms are capable of accessing the same physical instance of the recorded data on a storage subsystem. This is designated as True Data Sharing.

There are different levels of performance and complexity in the implementation of True data Sharing. The first level is when heterogeneous plarforms can access data, but only the original data owner is allowed to modify it. The next level is where a data item may be updated and re-written by mutiple heterogeneous platforms, but "one at a time". In this case, a locking mechanism must be used to momentarily prevent a platform from updating the data. The third called "Concurrent data sharing" is where all platforms can either read or update data simultaneously.

Data and format translation can be implemented via software or through hardware. Hardware will most likely be located in the Intelligent storage subsystem.

The advantages of True data sharing are numerous. With only one copy of data, there is never a need to replicate the data for use elsewhere, data maintenance is simplified and problems due to out of sync conditions are eliminated

Implementations of heterogeneous "one at a time" true data sharing exist today. It is fairly easy to implement using tape storage since the data is always recorded sequentially, or in consecutive groupings of records on the tape, so the file structure and format is the same for all platforms and the file systems are similar. Another storage device that lends itself to one at a time true data sharing for heterogeneous platforms is CD-ROM, since there are standards for the format and structure of these files. The standards require all platforms that attach CD-ROM drives to access them in the same way. An implementation by Encore in their Infinity product uses this scheme in sharing data between MVS and UNIX.

There are no known implementations of concurrent true data sharing. The most likely upcoming products will address the most popular UNIX implementations and PC servers running Windows NT

By: Farid Neema

This Paper was produced by:
PERIPHERAL CONCEPTS, INC.
351 Hitchcock Way, Suite #B-200
Santa Barbara, California, 93105
Tel: (805) 563-9491
fneema@silcom.com