Wednesday, February 4, 2009

Performance Management: SOA Ecosystem

Dimensions of Performance Management in SOA ecosystem are:

1. Capacity Management
2. Performance Tuning
3. Profiling & Analysis
4. Performance Bottlenecks
5. Architectural Analysis
6. Tools and Platform limitations
7. Data availability
8. Data discrepancy


Factors to be considered while doing Performance Management

1. Workload: Different forces that acts on a system which affect a system’s performance
a. Concurrency: Concurrent usage of shared resources like processor, memory, network, connection pool, thread pool etc lead to a queue which affect response time as well as throughput.
b. Session Length: Longer session time means higher usage of resources. High level of Concurrency and Session length are sufficient to stress a system.
c. Data in Session: More data in a session coupled with High concurrency and Long Session length is deadly combination for Performance Management.
d. Break Time: This represents the time taken by user to initiate subsequent request. This parameter plays crucial role in working with concurrency and session length.
2. Usage pattern: Usage pattern represents the variation in usage of system by various actors – human and machine over a time period in real time production environment. To understand Usage pattern volumetric studies are good source of information.
3. External Interfaces: In SOA ecosystem definition of external interface may assume various meaning. One can assume that interfacing with services within an enterprise/LOB will be treated as external interface while another analyst/architect might not agree. Things become more complex with involvement of B2B and Business to SAAS cases. And never forget normal I/O operations.
4. Infrastructure: In a SOA ecosystem Infrastructure always plays a crucial role in Performance Management due to its very nature. Few of the attributes of infrastructure are: network capacity; Server sizing in terms of CPU, RAM, cache, etc; load balancing; fault tolerance; virtualization; etc.
5. Persistence Layer: The volume of data to be read and to be written to persistence layer. Persistence layer primarily consistence of three systems:
a. RDBMS
b. Hierarchical Databases ( e.g. LDAP)
c. File system

Performance Management & SDLC

No comments:

Post a Comment