An important aspect of planning the implementation
and configuration of your Sterling Selling and
Fulfillment Foundation system
for production is determining your workload and business processing
characteristics, and your performance requirements. This includes
(at a minimum) the following:
- Identifying the key or high transaction volume
use case scenarios - for example, in retail environments, you may
have an order capture, order returns, order modification and order
authorization use case scenarios. For each use case scenario, you
should determine:
- The workloads (both custom-developed and IBM®-supplied) that are
carried out
- The forecasted peak transaction volumes
- When the peak processing periods occur during the
year
- The external systems the Sterling Selling and
Fulfillment Foundation system
is integrated with
When choosing use case scenarios, you should include:
- Workloads with anticipated high transaction volumes
- Workloads that are complex (for example, orders
with large number of order lines)
- Workloads/transactions that have to traverse long
network distances (for example, user and data center in different
continents)
- High volume transactions that are integrated with
external systems
For each use case scenario, you should:
- Perform load testing to at least the anticipated
peak workload volumes.
- Measure the computing resource cost at different
workload traffic volumes.
- Estimate the computing cost per unit work.
- Identify and tune expensive workloads - This could
include ensuring all SQL are supported by appropriate indexes (Custom indexes), custom code, and so forth.
- Incorporate the cost per unit work into a resource
capacity forecasting or planning model.
- Project out the resource requirements for the peak
periods.
If you have remote users, you have to test use case
scenarios that involve screens or network based transactions across
a real or simulated wide-area network. These include:
- The use of the Sterling Selling and
Fulfillment Foundation screens
(for example, to enter or modify orders)
- RF transactions
The answers to the questions above are critical
to how you configure Sterling Selling and
Fulfillment Foundation.
System test
We
strongly advise that you schedule time and resources to test Sterling Selling and
Fulfillment Foundation system
(including all custom code, integrated external systems, and so forth)
prior to implementation. IBM tests Sterling Selling and
Fulfillment Foundation to
common or general usage patterns. Your configuration may differ greatly:
- Custom code - Need to ensure your custom code scales
and does not have longevity issues. These are issues that show up
after running the system for many days - for example, memory or connection
leaks.
- Integration to external systems - Need to ensure
that external systems can scale along with Sterling Selling and
Fulfillment Foundation.
In the right conditions, slow external systems could tie up the Sterling Selling and
Fulfillment Foundation resource
and could lead to a system slow down.
- Configuration - Need to test the Sterling Selling and
Fulfillment Foundation system
with representative data. For example, your configuration may have
much larger catalogs and ship nodes than most customers.
- User locations - Need to ensure users get responsive
service. For example, you may have large customer groups located in
a different continent from Sterling Selling and
Fulfillment Foundation.
You may also have customers who dial in to access Sterling Selling and
Fulfillment Foundation.
You need to ensure that all users get appropriate screen response
times.