Many a times, I came across typical questions from various clients regarding application performance assessments and validation: “we have 30% scaled-down version of production for an enterprise application and how can you benchmark application performance? How can you peruse these results and predict performance for production hardware and provide performance sign-off?”
In my view, it’s quite a common scenario faced by Managers, SMEs, Architects and QA teams involved in providing application performance services – be it plain vanilla performance testing or comprehensive Performance Engineering (PE) services. Performance Modeling is one of the approaches that can help address such scenarios, however, it takes significant time to create and it becomes more complex to implement for distributed enterprise IT applications due to multiple processing layers, 3rd party systems, varying data and so on. Also, Performance Modeling requires mathematical/analytical modeling supported with more data points for accurate prediction. On the other hand, the idealistic approach is to get a PROD-like hardware in all layers, which is expensive and has high lead time to procure required hardware, in order to carry out performance benchmarking and provide sign-off.
In this whitepaper, I would like to share and suggest a simple but powerful approach that is based on performance engineering concepts such as ‘knee point’, ‘hardware capacity saturation’ and ‘Horizontal Scalability’ in specific. It helps solve the problem at hand in a more practical and convincing way as it’s cost-effective, simple and easy-to-implement for a given enterprise IT system.
Let’s take a look at the outcome of implementing the above-mentioned Approach for an enterprise level Client-Server application in one of the largest Financial Services firm in USA. The application aims to automate paper-based processing of various documents required by enterprises to carry out financial transactions. The application is implemented using a COTS product supporting Optical Character Recognition (OCR) technology that extracts data from various documents that will be further validated by internal users thru CITRIX interface. Due to business criticality, management wanted to have a performance baseline of this application, however, there is no environment that matches PRODUCTION configuration (i.e., number of servers and hardware capacity) that can be used for performance testing except their QA environment which is a scaled-down environment. (Refer to Table.1)
The application is expected to have peak load of 100 concurrent users accessing the application thru CITRIX UI and a batch throughput of 100 batches/hour, hence performance validation is required for these 2 non-functional requirements – Concurrent User Load & Batch Throughput.
In order to find out Max User Load that can be supported by given QA environment, carried out an incremental concurrent user load tests (4, 8 and 12 concurrent user load) for UI functionalities while the batch processing takes place in parallel, in order to mimic production scenario.
At 12 concurrent user load, observed 100% CPU Utilization on CITRIX Server whereas CPU usage on COTS server ranges 60% – 100% for the 10 min duration where all 12 users are active. Besides, the UI response time started degrading from 8 user load to 12 user load as shown in Fig.1 and Fig.2 below. CPU usage on CITRIX servers for 8 user load is ~60%.
Overall, 1 CITRIX and 1 COTS server in QA can support max of 10 concurrent user load with 10 Batches/hour as CPU utilization on both servers reached 100% when all 12 users are concurrently working and 2 users failed.
Linear extrapolation of test results to PRODCUTION capacity indicates that PROD can support 120 concurrent users with 80 batches/hour (refer to Table.2) which is close to the defined NFRs of this application. Once the application went live, Production Support who take care of production operations confirmed that ~100 end-users are currently accessing the application thru CITRIX interface and they have not heard of any performance issues from the users.
Having said that, let’s take a look the benefits and challenges of implementing this approach.
While this approach is definitely helpful to have early insights about application performance and capacity needs, there are few challenges that should be kept in mind as well:
To conclude, performance benchmarking and providing sign-off for applications in scaled-down environments has its own challenges to deal with – by embracing a simple and effective approach based on core performance engineering principles such as ‘application knee point’ and ‘horizontal scalability’ helps gain early insights of application performance instead of waiting till a PROD-like environment is created. Note that this approach is not a replacement to carrying out performance benchmarking on prod-like environments rather an alternative and effective approach to gauge application performance early in the game. After all, for business and IT teams, proactive information with actionable insights is the key to manage application performance effectively!!
About the Author
Madhu Tanikella is a Senior Technology Architect at Infosys, Hyderabad. He has 15+ years of experience in providing delivery and consulting services in Performance Engineering & Performance Tuning for enterprise software systems that fall under client-server, multi-tier (Java/J2EE), SOA, Messaging Middleware architectures, Virtualization and Cloud models and custom-built applications with multiple COTS products. He has specialized expertise in the areas of NFR & SLA Validation, Workload Modeling, Performance Modeling, Bottleneck Identification and Performance Tuning (of Java/J2EE, Oracle DB, TIBCO Business Works and IBM MQ), Capacity Planning, Hardware Assessment and Sizing for distributed software systems. He also has expertise in assessing performance practices in an organization and setting up Performance Engineering CoEs for various clients in different Domains such as Financial Services, Manufacturing, Insurance, Healthcare, Retail etc. Madhu Tanikella is certified in IBM Cloud Computing Architecture and IBM SOA Solution Designer. He can be reached at [email protected]
Save the date! We will be sending out the agenda in a few days. STLCMG’s...Find out more