69 / 96
W H I T E P A P E R
© 2017 Persistent Systems Ltd. All rights reserved. 69
www.persistent.com
7 Non-functional aspects
7.1 Brief Summary
Performance
is critical to the functioning of the business at ease with speed and agility. The two main measures of
performance are
—
Response time of queries to the DW/BI system, and
—
Throughput, i.e., a count of the number of queries the DW/BI system can process within a defined time interval.
A related but
similar
topic that we will be addressing under this section is
scalability
, which corresponds to the ability
to overcome performance limits by adding resources.
Similarly, Security is a big, esoteric and complex topic which, if not acted upon, can result in range of potential
problems. The wide variety of data speeding through the enterprise and moving into and out of the data warehouse
requires unprecedented levels of protection. An end-to-end security strategy is required for the data warehouse
environment. A data warehouse environment consists of much more than just a database. The entire environment
ranges from components extracting data from operational systems, moving this data to the data warehouse,
distributing the data to other analytic platforms, and finally, distributing it to the end business user. In today’s highly
distributed, complex landscapes, the environment spans multiple servers, applications and systems.
This chapter focuses on key sub-areas as listed below:
I. Performance and scalability of various components in the DW/BI (space as well as speed); Sizing and
production deployment
II. Security – Security aspects of the DW/BI system;Auditing, Error logging and data life cycle.
7.2 Best Practices
7.2.1 Performance at Requirements and planning stage
—
Interview the business process owners to get an understanding of the business requirements
(page 72, 83-
[1]86) and
gather detailed information from the performance point of view on each requirement.
Some of
the questions could be: how many users will be using the system, what is the mix of users based on workload,
what are the data query patterns, how would the system be used by these users (on Web or Smart
Phones/Tablets), what is the current data volume and anticipated data growth in next 6-months or 2 years,
what are the SLAs anticipated for end-to-end data movement vs SLA required at each component level, what
kind of hardware is planned in the production, will the hardware be shared across departments, will it be
deployed on cloud and if yes, are there any budget constraints, is time for performance tuning or dedicated
personnel for testing and optimization accounted for in the plan etc.
—
Requirements can never be frozen and may evolve, or change, or drop as the design and implementation
progresses. Thus, managing the expectations fromperformance is a continuous process by not over-selling,