Database Reference
In-Depth Information
◦ Throughput is about 10 msg/sec for a 150 KB Order message. This para-
meter was never a problem as we can increase operational concurrency
by implementing parallel queues/handlers.
◦ The error rate is approximately 0.01 percent faulty message per week.
Practically, all errors are related to XML validation for inbound mes-
sages.
• One of the declared goals was to put OEBS BES back into production. After an
initial trial period, EHS implementation was considered so satisfactory that it
stayed in production for three years (since 2011).
• Because of considerable simplification of the Adapter framework, migration from
OFM 10
g
to 11
g
was gradual and relatively smooth.
Of course, there are always walls to hit. We had plenty of them, as already mentioned, in-
cluding AQ implementation in XE, the
DBMS_LOB
functions for CLOB handling below
and above 4 KB, and so on. We can explore the following two major factors that helped
implement this approach:
• Strong support from local RRD Architecture team. Because of a very strong dis-
cipline established by system owners and wide authority finally granted to the
chief architect (truly,
Necessity is the best Advisor
), implementation of this
concept became possible in such a competitive environment.
• When developing in Java or SOA Suite, we have plenty of tools for maintaining
concurrent development, implementation, and components interdependency
(Maven is the natural choice, Hudson/Jenkins, classic Ant, and so on). Unfortu-
nately, we do not have so many options for team collaboration in PL/SQL. Code
rollup is a tricky business in centralized DB environments, such as OEBS, when a
dozen teams work on the same modules (Account Payable, for instance) and we
need some framework in addition to the strong discipline. Luckily, such frame-
works were established, thanks a lot to the architect who managed to build a
Maven-like tool, which allowed us to do complete or partial installations, code
validation, and assembly on running environments practically without downtimes.
We hope that this example demonstrated how to optimize the Adapter framework (for DB
adapters) by refactoring core applications, making them more SOA-oriented. Now, we re-
turn to the middleware layer and see how the ABCS optimization can be achieved there.
Search WWH ::
Custom Search