Bayesian Cognitive Model in Scheduling Algorithm for Data Intensive Computing 175
issues, in forms of capturing and accessing data
effectively and fast, processing it while still achiev-
ing high performance and high throughput, and
storing it efficiently for future use [25]. Program-
ming for high performance yielding data intensive
computing is an important challenging issue. Ex-
pressing data access requirements of applications
and designing programming language abstractions
to exploit parallelism are at immediate need. Ap-
plication and domain specific optimizations are
also parts of a viable solution in data intensive
computing. While these are a few examples of
issues (such as data capturing, management, and
scheduling techniques), research in data intensive
computing has become quite intense during the
last few years yielding strong results.
2.2 Compute Resource Scheduling
In large-scale parallel and distributed computing
environments, many static, dynamic, and even
hybrid algorithms have been proposed. At the
same time, some issues related to distributed
scheduling, center scheduling, autonomy schedul-
ing, intelligent scheduling and Agent negotiation
scheduling are also in exploration. In static algo-
rithms, BNP-based ISH [3], MCP [4]andETF
[5] algorithms are suited for high speed and low
delay small distributed networks, which is con-
trary to the large-scale environment; APN-based
MH [6]andDSL[7] algorithms run very well in
large-scale distributed system and communication
delay and time cost are considered, but cannot
meet the requirement of computing nodes in trust.
In dynamic algorithms, dynamic job scheduling
is considered, and jobs load balancing and shar-
ing can be guaranteed by autonomy and intelli-
gent scheduling [8, 9]. In hybrid algorithms, jobs
uniform distribution and communication over-
head reducing are emphasized and load balanc-
ing is achieved by considering the computation
that a node performs [10, 11]. In [10], based
on DSL (Dynamic Level Scheduling), Dogan et
al. proposed a bi-criteria heuristic called RDLS
(Reliable Dynamic Level Scheduling). In [16]
Hakem and Butelle proposed BSA (Bi-objective
Scheduling Algorithm), a bi-criteria heuristic that
outperforms RDLS. However, none of these algo-
rithms takes the characteristics of nodes’ behavior
into account in data intensive computing, such as
uncertainty, unreliability and deception, and the
scheduling length and trustworthiness of nodes
cannot be considered synchronously.
2.3 Bayesian Models of Cognition
Bayesian models are becoming increasingly pro-
minent across a broad spectrum of the cognitive
sciences. Just in the last few years, using Bayesian
models people have addressed such problems as
animal learning, human inductive learning and
generalization, visual scene perception, motor
control, semantic memory, language processing
and acquisition, symbolic reasoning, causal learn-
ing and inference, and social cognition, among
other topics [12, 13]. Behind these different re-
search programs is the most compelling com-
putational question that we can ask about the
human mind. Bayesian models give us ways to
approach deep questions of human cognition that
have not been previously amenable to rigorous
formal study. The Bayesian framework for prob-
abilistic inference provides a general approach to
understanding how problems of induction can be
solved in principle, and perhaps how they might
be solved in the human mind.
3 Bayesian Cognitive Trust Model
3.1 Basic Concept
Trust is the core of relationships in social net-
works. Trust is the evaluation of certain entities’
reliable behaviors. The trust degree of a certain
entity is always decided by others’ recommenda-
tions. DISC system and social networks have great
similarities: a computing node displays a mes-
sage, reflecting the characteristics of its behavior
when it cooperates with other nodes; a node has
sufficient choices; and the node is duty bound to
offer recommendations to other nodes.
Thus, the node can evaluate the copartner
through its behavior (e.g., the node’s ratio of suc-
cessful execution). Nodes can also exchange and