 |
|
|
What are concurrent users? Part 2
In part
1, we clarified the terminology used in describing web performance
and load, often the first point of ambiguity. We can now really start to
examine the meaning of 'concurrent users'. From a technical perspective,
sizing a system is based on many factors. Ultimately, the execution of
end-to-end performance testing will be based mainly on how many
'concurrent users' you have and what they are doing. To get to that
point, it is necessary to analyse the business requirements first…
The Business Requirements
Before you can determine how many concurrent users you
need to emulate, you need to understand what the business wants to
achieve and how that is going to translate into 'concurrent users'.
What is the business interested in? In most cases: Revenue and Profits.
How this is relevant depends on the type of business and application.
For now we will concentrate on an e-commerce application and how the
principles broadly apply…
The business needs to sell a certain number of units per year (i.e. #
users completing an end-to-end purchase), to sell these there will be a
conversion ratio of total visitors to purchasers. Marketing and business
analysts should have defined these figures within the business plan,
business case or product strategy. Sales need to be achieved with as
little cost as possible, so, the cost of the capital investment in
hardware required to support this throughput is of particular interest
to the business. Demographics of the users will be important too. If the
site/application is targeted predominantly at one particular group, they
will have distinct usage patterns. For example, they may visit the site
mostly during lunch time, Monday-Friday, thus a 2 hour daily peak might
be expected with 80% of the total weekly visitors compressed into these
times. Finally, you need to make assumptions about the typical session
duration.
Also, don't forget planned or possibly unplanned 'special promotions'
that the system must also support which may dramatically increase the
usage of the site.
Real users - not virtual users
With this information, you can determine the number of
visitors per peak-period. You must also determine what they will be
doing to see if they should be considered as 'Active' .'Active' implies
they are doing something such as:
-
Manually performing a function on a web page (client
function).
-
Executing a function (server or application
function).
-
Physically reading a downloaded document in their
browser window (possibly holding open connections).
-
Waiting to receiving data (HTTP request &
receive).
A person who has a web page showing on their web browser,
but who may be out to lunch, or have the browser window minimised, is
not an active user, assuming that their HTTP connection is dropped.
Whether the connection is dropped can depend on the connection time-out
setting on the server or even the brand of browser used to access the
application... This will be explained in part 3.
Also in part 3, we will address the translation of the above factors
into technical issues, the behaviour of concurrent HTTP connections (not
the same as concurrent users), considerations for implementing the load
using a test tool and the method of calculating 'concurrent users'.
Nov
2002
Subscribe
to etest associates newsletter for
more articles like this one.
top
|
|
|
Services...
security
load/stress
performance
monitoring
browser
compatibility
system
testing
dispute
resolution
localisation
usability
functional
UAT
macromedia
zone
more...
Quick
links...
"We gained the complete trust of our clients
[ITV & Channel 4], both commercially and technically. etest
associates provided a comprehensive and professional service throughout,
and I would happily use them again in the future."
Adrian Pryce
Platforms Manager
I T N - New Media
"... highly knowledgeable
and credible experts on e-business testing methods, techniques, tools
and commercial realities."
Neil Thompson
Test Manager
Barclays
Business Internet Banking
"I found etest associates a very helpful
company in their responsive and flexible approach that kept pace with
the ongoing needs of our development programmes. They have been able to
provide a broad range of testing services some of which were very
specialised. When asked to recommend an independent testing company I
have no hesitation in suggesting etest associates."
Mark Cunningham
Project Manager
Force Solutions Group
more testimonials...
|
