NonStop Systems Introduction

Requirements of ZLE Systems
NonStop Systems Introduction—527825-001
2-2
Data Integrity
The ZLE system must be able to update operational data continuously in real time; this
requires continuous availability of the data store. Critical business applications cannot
afford even a second of data store downtime.
Without fault tolerance, computer systems can experience downtime for the following
reasons:
•
A hardware failure such as a processor failure or communications line failure can
halt transactions in progress.
•
A software failure such as the failure of an I/O process that manages a disk drive
can halt transactions in progress.
•
The entire system can be taken out of commission while customer engineers repair
a faulty component.
•
A particular application can be taken down so that operators can configure and
install more disk drives, communications lines, and processors for the application
or start additional copies of the application.
ZLE systems cannot be susceptible to any of these causes of downtime. The system
must be designed to keep applications up and running by eliminating these causes.
Furthermore, the ZLE system must continue to provide service even when existing
applications are brought offline for maintenance. It must continue to offer data store
access even during data store reorganizations and backups, system reconfiguration
operations, and so on.
Data Integrity
Data integrity means that neither the failure of a particular transaction nor the failure of
a hardware component such as a processor or disk drive can corrupt the data in the
database.
One of the most important roles of a ZLE system is to keep all of the operational data
bases in synch. Consider a sales order transaction. Suppose that for some reason
(such as a program error) the transaction stops after it has added a new order to the
order database and updated the data store with information about the order, but before
the hub can push the updated the information out to the inventory and accounts-
receivable databases to reflect the order. The result is that all three databases have
lost data integrity.
It is true that the order database contains a record of the new order, but the problem is
that the new order data is inconsistent with the current inventory and accounts-
receivable data. If the new order record states that customer A has ordered five boxes
of stationery, but the inventory record for stationery does not show that five boxes have
been deducted from the quantity of stationery on hand in the warehouse, then either
the order record is wrong or the inventory record is wrong. But there is no way of
knowing for sure.
For another example, suppose a customer calls a company’s credit card call center to
charge a purchase. The call center credit application then queries the data store to