Database Reference
In-Depth Information
￿
Low stress large file read, write, and delete are tests designed to evaluate large
data transfer in the end-user-cloud database scenario. The throughput measure is
as observed by the end-user. Once again, this test aims to characterize the total
end-to-end large data handling capability by the cloud platform, taking into
consideration the various network variations. The CARE framework provides
some default test data: ranging from 1 MB, 5 MB, 10 MB, and through to 15 MB.
A RESTful Web Service based end-user is implemented for a set of target cloud
databases, including Amazon S3 and Microsoft Windows Azure Blob Storage.
Note that the CARE framework does not provide a test for the Google App
Engine, as Google App Engine Datastore does not support an interface for direct
external connection for large file access.
4.2
Application of CARE to Cloud Platform Evaluation
Providing a common reusable test framework across a number of different clouds is
a very challenging research problem. This is primarily due to the large variations in
architecture, service delivery mode, and functionality provided across various cloud
platforms, including Amazon Web Services, Google App Engine, and Microsoft
Windows Azure. Firstly, the service models of cloud hosting servers are different:
Amazon EC2 uses the infrastructure as a service model; Google App Engine uses
the platform as a service model; while Microsoft Windows Azure combines both
the infrastructure as a service and platform as a service models. Different service
models have different levels of system privileges and different system architectures.
Moreover, the connections among cloud hosting servers, cloud databases and client
applications tend to utilize different protocols, frameworks, design patterns and
programming languages which all add to the complexities to the task of providing a
common reusable evaluation method and framework.
Therefore, we proposed a unified and reusable evaluation interface based on
Contract-First Web Services and RESTful Web Services, for the purpose of keeping
as much commonality as possible. As illustrated in Fig. 4.3 , for the Contract-First
Web Services: a WSDL file is firstly built; then, the cloud hosting servers implement
the functions defined in this WSDL file; lastly, a unified client interface is created
from the WSDL file which allows communication via the same protocol, despite of
existing variants. While for RESTful Web Services, direct access to cloud databases
is made without passing the cloud hosting servers. The CARE framework currently
provides the reusable common client components, and the cloud server components
for Microsoft Windows Azure, Google App Engine and Amazon EC2.
The evaluation interface maximizes reusability of client application on the end-
user side. The Contract-First Web Service based client application is able to talk
to different cloud hosting servers via the same WSDL whereas a RESTful Web
Service based client application can talk to cloud databases directly without passing
the cloud hosting servers via the standard HTTP protocol.
Search WWH ::




Custom Search