The Requirement of Unlimited Scalability Nonstop

Present transactional solutions with integrated schemas [DB2UD, DVDP] have no sufficient scalability of integration by the following reason. What if one needs to access data placed within different integration systems? The answer is obvious: either one should refer to every integration system explicitly during development, or there should be created another integration system combining these two. The first case loses advantages going from uniqueness of an integrated schema; the second case does not let combine integration systems infinitely since the scalability limit will be reached sooner or later; after that no further extension will be possible because of efficiency degradation of the extended integration system - meta-system.

Meanwhile the problem of creation of a united meta-system, being constantly extended and absorbing more and more information systems, become increasingly relevant since enterprises become more open for integration between each other and between their remote departments. High-quality integration is possible only if the meta-system has a single seamless access interface to all data, and not partial integration in the form of unconnected locally-integrated systems. It's obvious that such approach imposes special requirements to the integration system. What principles should we use for the system to create the integrated information space over unlimited quantity of information?

It's obvious that the crucial requirement to such integration system is its unlimited scalability: possibility to include new systems without efficiency degradation of the meta-system as a whole. Unlimited scalability is impossible if all meta-information, information about structure of data, is stored at one place since such centralization results in bottleneck from the point of view of efficiency and reliability. Therefore meta-information should be stored in the distributed way, which lets diversify data flows between different Net nodes without any common coordination server.

The next important architectural requirement is possibility to extend the integration system on-the-fly without stopping it. System stop for the period of integration of new systems may result in its complete inaccessibility since the integration process can take place constantly within such approach. This requirement results in independent publication of meta-information for every system being integrated, of course provided that logical integrity of information structure remains. A system to integrate should become a part of the meta-system just since the publication of its part of meta-information.

All these requirements concern to the way of storing and access to meta-information mainly. We infer from them that meta-information should be stored as a set of meta-objects forming a united space. Their publication and modification should go independently one from another, with transaction support. In spite of the fact that meta-information is stored in parts on different Net nodes, all the meta-object space should be usable as a single whole through a general object access interface. Scalability of the object space should be unlimited at that. There is only one product on the market conforming to the description in full, EntryService [FES], which is used for storing meta-information within the proposed solution of integrated structured data space creation.