Databases Reference
In-Depth Information
• The complexity involved in managing transactions flowing between
different sites.
• Communications delays caused by insufficient capacity during peak
time loads.
• Managing data consistency and concurrency control between sites.
These obstacles explain why there are no true distributed database
products. The remainder of this chapter addresses communications prob-
lems and requirements of distributed database system.
DATA COMMUNICATIONS REQUIREMENTS
A model of the decision-making steps for determining distributed data-
base communications requirements is shown in Exhibit 1. The first step in
this process is to choose the applications that the data communications
networks will support. Determining a single set of communications require-
ments for all distributed DBMSs is difficult, because architectures and fea-
tures depend on whether they are aimed at high-volume, online transaction
environments or decision-support and revenue-generating applications.
Each of these environments requires different network characteristics,
depending on the number of distributed queries or updates made and the
level of fragmentation and replication. Because of the complexity and diffi-
culty of establishing the relationships between these factors and data com-
munications requirements, detailed discussion of this matter is not pro-
vided in this chapter. Once the type of application has been identified,
performance considerations can be evaluated.
PERFORMANCE CONSIDERATIONS
To provide an adequate level of performance, efficient response time is
critical. Response time consists of processing time and transmission time.
In a distributed environment, transmission time has a large impact on
response time. The first step in evaluating network performance is to set
response time objectives. If data is fully replicated or is fragmented at
many sites requiring frequent accesses between sites, a low response time
is difficult to achieve. For systems that have low fragmentation and do not
replicate data fully, response time is not as critical. Once response time
objectives have been determined, factors that directly affect performance
(e.g., propagation delay, network capacity, media, and communications
equipment) should be examined.
Propagation delay is perhaps the most significant obstacle to distrib-
uted DBMSs. Low delay is especially critical in systems that replicate data
at different locations, or in situations in which a query is made that brings
together fragments from several databases. Circuit propagation times of 20
milliseconds or more will cause throughput delays for data. With fully rep-
licated systems, any update made at a local level must be propagated to
Search WWH ::
Custom Search