This
is the fourth of a six part series of posts on how SOA works within the
Agile life cycle. Here we will look at IT operations, monitoring and performance.
Once an enterprise application is in deployment and under use, ensuring its
continued availability, performance and accuracy becomes critical to the
business. IT Operations teams may be part of the overall test and development
team, or grouped as a separate function in many enterprises to offer a measure
of impartiality to ensuring that required Service Levels are met within the
live application environment.
Operations Management and Monitoring
The leading monitoring frameworks can be supported by directly testing and validating the systems under management, and feeding rich metrics and test output directly into the team’s IT Management dashboard, such as Wily/CA Unicenter, TIBCO Hawk, HP OpenView, IBM Tivoli, and others. As an example, our LISA tests can be run continuously at runtime and act as a listener for both functional errors, and server metrics from components including JMX, SNMP, Windows Perfmon and LISA’s own types of metrics.
The Operations team can set thresholds or boundary conditions on these continuous tests and have the Testing framework alert them if an issue occurs. For instance, if performance is too slow, or memory usage or a certain database table increases in an unexpected way over time, the test should report a failure to the management dashboard, and send an SMS message to the admin, along with the test case for root cause analysis and correction. LISA can also output this data as reports into the team’s management tools or data store to provide the ability to perform audits.
Performance & Load Testing with Leading Tools and a Virtual Service Environment (VSE)
Also key to the Operations team is ensuring that delivered systems will meet expected Service Level Agreements (SLAs) for customers. For today’s highly interconnected applications, the leading Test and Performance Lab solutions on the market were left out of the lifecycle of applications until near the end, when an interface or a complete system environment was available to test. A Virtual Service Environment (or VSE) provides teams using Performance Testing tools like LoadRunner and SilkTest the ability to get involved in ensuring performance much earlier in the lifecycle, at up to a 90% reduced cost per test environment.
The VSE captures and simulates the expected behavior and response times all of the dependent systems within your target environment. As component load tests are conducted earlier, teams can optimize their use of system resources, find memory leaks and fix the root causes of errors, at a much earlier stage in the design and development lifecycle, without needing access to the live system or all of the interdependent components needed in the test environment. The ability for multiple teams to conduct their own performance testing processes in parallel, prior to solution delivery, pays huge dividends in productivity and quality in deployment.
Behavioral Service Virtualization in this sense is by no means a replacement for virtualizing hardware and network resources to save costs in IT operations. But it does represent a huge way to reduce the dependencies and costs that can't be touched by virtualizing a given set of servers - after all, there are many third party or shared Services in the cloud, or heavy transaction systems and partner systems that can't be virtualized by conventional means.
Michael Vizard of eWeek writes about the role of virtualization
in his post, Network Management Gets Virtual. He
writes that many vendors are saying that, “it's become critical to extend existing network and systems management
tools into the realm of virtualization. They argue that IT
organizations don't need to invest in learning separate management tools for
physical and virtual environments and the impact of decisions in one
environment can be devastating on another unless the management infrastructure
for both are tightly coupled.”
This tight coupling can occur under LISA. Using a VSE, teams can either model incomplete components, or capture live services, and simulate their behavior on a running server. Functionally and under load, these Virtual Services hosted in LISA’s VSE act and respond just like other services, databases and systems you connect to in the live deployment, including expected response and transaction times. Without requiring access to critical live systems, the team can connect their component to the rest of the virtual architecture in LISA Virtualize, and conduct performance and load testing with their tools of choice.
It is important to note that most companies have significant investments in test lab tools, and supporting them with robust target test beds makes those labs much more productive within the software lifecycle. Of course, our LISA Test & Validation solutions can also be that load and pefromance testing engine, and LISA test suites are executable and
interoperable with the Test Management, ALM and SOA Governance processes
mentioned in this series.
In our next post, we will look at IT and SOA governance. You can download this complete series in the Whitepaper on "Agile SOA" at our ITKO LISA resources page.
Comments