Information Technology Reference
In-Depth Information
Scope-awareness.
To support a large amount of
services, defining and grouping services in scopes
facilitates service search. Location-awareness is
a key feature in pervasive computing (Weiser,
1991) and location information is helpful in many
service discovery cases. Location information may
be integrated with service discovery protocols.
For example, a project at MIT (Chakraborty,
2000) integrates Cricket into INS to provide
location dependent service discovery. Another
example is Jini, in which location information
is an optional attribute for services. Moreover,
administrative domains are another kind of scope,
which is supported by many protocols. Often, in
enterprise environments, services are arranged
in administrative domains. These geographical
location information and administrative domain
information may be set as attributes of services.
Much research has proposed locating objects
in a wide area. Some of them use a single direc-
tory hierarchy, and others use multiple directory
hierarchies. No matter if there is a single hierarchy
or multiple hierarchies, the difficult problem is
to express the service information at different
levels of the hierarchies. First, what services
need to be listed in upper level directories? Sec-
ond, what service information to store in lower
level directories and what service information to
store in higher level directories? To avoid being
a bottleneck, upper level hierarchy directories
should be concise. Filtering and aggregating ser-
vice information is necessary when building the
upper level hierarchies. Third, updating service
information in the upper level hierarchies may
overwhelm the directories, when many services
update information at the same time. Service status
changes and mobile services moving all cause
the directories to be busy updating. In SSDS,
service information in non-leaf level directories
is created by using Bloom filters to achieve high
compression ratio. Nevertheless, the directories
need to be built again and again over time, since
the algorithm is not able to remove stale services.
Another example of locating mobile objects in a
wide area is Globe (Steen, Hauck, Homburg, &
Tanenbaum, 1998).
QoS-awareness.
Providing users with better
services and balancing services usage are nice
features for service discovery protocols. For
better service matching, service requests may be
directed to less loaded services or better resource
price ratio services.
Service attributes are defined to match client
requests more precisely. Nevertheless, most pro-
tocols only support static attributes. Sometimes,
dynamic information about the current status
of a service should be taken into consideration,
for instance the current load of a service. Much
more communication traffic may be generated
and directories may be more busy handling an-
nouncements. To reduce the directory's update
and network overhead, services may wait for
clients to query.
At the service side, sharing the loads and bal-
ancing them on different services is also preferred.
Few protocols define application metrics-based
load balancing. A good example is INS. Applica-
tions define their metrics and service lookups are
based on the metrics.
Cross-Layer Optimization
Most service discovery protocols are designed as
application layer protocols. These protocols enable
clients and services that run on different hardware,
software platforms, and network protocols stacks
to interoperate with each other. Nevertheless,
these protocols may not be efficient for dynamic
and resource constrained environments such as
ad hoc networks. An active service discovery
research area is to explore the cross-layer design
to optimize the performance of service discovery
protocols and adopt design approaches used in
wireless sensor networks and ad hoc networks.
Topology based.
Ad hoc or sensor networks
may form three types of topologies:
flat
networks,
backbone
based networks and
cluster
based net-
works. In flat works, all nodes perform the same
Search WWH ::
Custom Search