Database architecture fertilizers: Just-in-time, just-enough, and autonomous growth

Abstract

Ambient application environments call for innovations in database technology to ful-fill the dream of an organic database, a database system which can be embedded in a wide collection of hardware appliances and provides an autonomous self-descriptive, self-organizing, self-repairable, self-aware and stable data store-recall functionality to its environment. The envisioned setting consists of a large collection of database servers holding portions of the database. Each server joins this assembly voluntarily, donating storage and processing capacity, but without a "contract" to act as an obedient agent for the user in coordination of access to all other servers. They are aware of being part of a distributed database, but do not carry the burden to make this situation transparent for the application. Applications should be prepared that updates sent to a server are either accepted, rejected with referrals, or only partly dealt with. An active client is the sole basis to exploit the distributed system and to realize the desired level of ACID properties. The query language envisioned for this system avoids the trap to allocate unbounded resources to semantically ill-phrased, erroneous, or simply too time-consuming queries. It limits the amount of resources spent and returns a partial answer together with referral queries. The user can at any point in time come back and use the referral queries to obtain more answers. The topics presented are part of ongoing long-term research in novel database technology based on and extending MonetDB1. © Springer-Verlag Berlin Heidelberg 2006.