NonStop Systems Introduction

Requirements of ZLE Systems
NonStop Systems Introduction—527825-001
2-12
Support for Heavy Transaction Volumes and Mixed
Workloads
Programs built using standard component models are portable: programs developed
on other platforms, such as Windows or HP-UX, can easily be ported to run on the
NonStop Kernel platform. And programming skills are transferable as well; developers
do not need to learn new skills to develop applications for the NonStop Kernel.
Support for Heavy Transaction Volumes and
Mixed Workloads
A ZLE system must be able to support an extremely heavy mixed workload, including:
•
Hundreds, thousands, or even tens of thousands of events being added per
second
•
Massive random reads in support of ZLE inquiries
•
Constant random updating and deleting of information to keep data store current
•
Bulk deletions to remove old data
The ZLE model requires a data environment and transaction and query execution
environment in which data flows into and out of the ZLE data store with zero (or low)
latency. It must be able to handle constant updating to remain current with the
operational systems. The data is used for making real-time decisions that affect, for
example, customer service. Hundreds of thousands of these decisions, which typically
require subsecond response, may be executed per second. Changes resulting from
ZLE applications must be propagated to operational systems to keep the enterprise
synchronized.
For example, several clerks in the same office might be inquiring against an inventory
database and placing sales orders in an order database simultaneously. The
application must process all these transactions at the same time. At the same time,
other users might be querying the data store for decision support analysis. The system
must be able to support these mixed workloads.
ZLE data may be subject to demand by thousands or even hundreds of thousands of
users. High volume users include call center agents, web guests, mobile phone access
agents, event and interaction capture agents, EAI agents, and operations analysis or
clerical users.
An architecture that supports zero latency operations must be able to handle massive
numbers of transactional inserts and updates, batch extracts, online transaction
processing (OLTP), and massively parallel queries against the same database tables
concurrently without degrading performance levels. It must be able to perform different
types of functions and process different kinds of workloads in parallel and around the
clock.