Registry & Scale free networks

Registry as implemented in current SOA architectures does not scale up in EOA.
Services or Event registries are strategic elements for SOA. Even tough information and services are theoretically always active and available, if it is not known their network location, they get often fully lost.

Taking into consideration such as criticality of registry in SOA, OASIS in UDDI version 3 has introduced replication policies on cluster that allow an higher degree of tolerance toward damages. Currently UDDI version 3 supports both clustering and mirroring of a main node. Replications are complete and for such a reason they have to be implemented into the intranet so as to avoid any network constraint. The strategy is a complete replication for failover where each node is a mirror of the master.

In EOA we have an international-wide federation of services where it is strongly suggested not to set up the registry in a single node, even if implemented as a cluster. Who will assume the responsibility of maintaining 24hrs. x7 days a system that if crashes will cause impossibility to access to the whole ecosystem? By persisting with the intention of reinforcing and consolidating the single central registry the result would be a sort of Fort Knox where also the network of services like the supply of electric energy should be redundant, replicated and should come from different power stations1. How would it be possible to guarantee service continuity in case of natural calamity?

Typical disaster recovery solutions lack of scalability if applied throughout the whole national territory: N^2resource should be employed where N stands for the number of place of failover and procedures of human intervention for network reconfiguration should be planned. An hierarchical structured network cannot easily get managed and balanced by an administrator due to its numerous parameters which change very frequently.

The alternative solution proposed in the context of digital ecosystem registries is to exploit resources spread in the territory, getting an higher degree of tolerance against failure through a self-healing architecture (self-maintaining), replicated, redundant, with a scale free distribution rather than hierarchic, gaussian (random) distribution.