Database Reference
In-Depth Information
accurately, due to the absence of a timer process in the cloud database. The CARE
framework thus equates this measurement to time taken to process the database
request as seen by the cloud hosting server by measuring the processing time of
the database API as the database processing time as the latency between the
hosting servers and cloud databases within the same cloud platform is negligible.
Additional terminologies used refer to different response types that are based on
the request:
Incomplete request
is a type of request where an end-user fails to send or receive.
Completed request
refers to a request where an end-user successfully sends and
receives a confirmation response from the cloud platform at completion time.
Subsequently, depending on the response, the completed request can be further
classified as:
Failed request
that contains an error message in the response.
Successful request
which completes the transaction without an error.
Test Scenarios
The CARE framework provides three key test scenarios to differentiate the can-
didate cloud platforms. While there are potentially other more sophisticated test
scenarios, the three test scenarios provided by CARE cover most of the usage
scenarios of typical cloud applications. Hence, the CARE framework provides a
set of test scenarios that strikes a good balance between simplicity and coverage.
End-user-cloud host
represents the scenario that an end-user accesses a web
service application hosted on the cloud platform from a client side application.
The response time would be the end-user's primary concern in terms of the cloud
application performance.
Cloud host-cloud database
represents the scenario that an end-user operates on
a form or an article hosted in the cloud database through the cloud hosting server.
The time taken to send the request from the end-user to the cloud host server is
excluded as the focus is on the impact of different data sizes on the database
processing time. It is especially interesting to be able to measure the database
processing time of concurrent request that have been simultaneously generated
by thousands of end-users. The database contention due to concurrent requests
will be a key-determining factor in the overall scalability of the cloud platform
in this type of scenario. Besides identifying different performance characteristics
across cloud databases, a local database (LocalDB) is also provided by the CARE
framework in a cloud hosting server as a reference point for comparison to other
cloud databases.
End-user-cloud database
illustrates a large file transfer scenario. It is conceiv-
able that data-intensive computing would be increasingly pervasive in the cloud
where a large variety of new media content, such as video, music, medical
