Strong Mobility in Autonomous Decentralized Systems

Decentralized Systems (ADS), which are based on the idea of abstract sub-systems intended
to address the diversity and dynamism of user demands. ADS are characterized by scalability,
fault tolerance and online system maintenance in real time systems, modern manufacturing and
production systems for e.g. railway control, steel manufacturing systems and many others.
An ADS is made of abstract autonomous subsystems; each sub-system is controlled and
managed by its own Autonomous Control Processor (ACP). These sub-systems communicate
among them using a unique method of communication called content code communication
which comprises of a Data Field (DF) and a message. In large ADS like the one in a
manufacturing system, there are various unforeseen external interferences on individual sub
systems for e.g. partial system failures, changing user requirements and many others. In order
to counter these interferences, an ADS should be mobile and flexible, since its autonomous
sub-systems should be able to migrate their computation to other sub-systems in response to
changing user requirements and environmental conditions. Moreover, fault tolerance in ADS
requires that failure of a single sub-system should not fail the entire system operation; therefore
an online replacement of a failed sub-system becomes essential for ADS. In afore mentioned
case, the job of the failed sub-system is delegated to another sub-system that requires mobility.
I have designed and developed a generic software, named as Strongly Instrumented Generic
Mobility Architecture (SIGMA) that provides strong mobility to autonomous agents which
exist in form of objects that run on different machines. The term 'strong mobility' in SIGMA
corresponds to the migration of code, data and execution state of the distributed entities from
one computational entity to another within a network. The distinguishing feature of strong
mobility is that the application continues to run as if nothing happened even if it is interrupted
or relocated to different sub-systems several times. With strong mobility we can counter the
interruptions and disturbances that are typical cases in an ADS; thus, saving the computation
and leading to increased performance and reliability in an ADS, in cases of external
interferences. However, the implementation of strong mobility is dependent on the
technological environment being used, especially the programming language, since many
modern programming languages provide restricted or no access to their execution stacks.
Currently SIGMA accesses the execution state of Java Virtual Machine (JVM) using byte code
instrumentation without violating any security constraint. Whenever interference occurs, it
transparently migrates the Java thread to another machine on the network. Right now, SIGMA
implements the concept of strong mobility at thread level only; my aim is to bring strong
mobility at all levels of an ADS sub-system. Strong mobility needs to be incorporated in
metadata processing, information semantics representation and other levels. The ultimate goal
will be to materialize the concept of strong mobility in ADS starting from individual subsystem
of an ADS and taking it to the mutual cooperation, control and communication among
all the sub-systems and analyze its prospects.