In-Depth

DataMirror Tweaks Mainframe CDC Tool for Performance, Reliability

DataMirror last week trumpeted what it describes as “significant” performance improvements in the latest version of its Transformation Server for z/OS

• By Stephen Swoyer
• August 1, 2006

It’s not every day that DataMirror Corp. releases a new version of its Transformation Server for z/OS. It’s been more than a year, in fact, since DataMirror shipped Transformation Server 4.7, the last major revamping of its flagship mainframe data integration tool.

DataMirror last week trumpeted what it describes as “significant” performance improvements in the version 5.1.7 release of Transformation Server for z/OS, a Change Data Capture (CDC) tool that enables bidirectional synchronization between z/OS and other platforms.

The retooled release includes a new single-scrape feature and enhanced auto-restart capabilities that DataMirror officials say deliver improved throughput and scalability as well as better reliability.

Transformation Server for z/OS uses DataMirror’s log-based CDC technology to capture changed data from database recovery logs, which helps circumvent the overhead and risks associated with trigger- and table-scan-based approaches. DataMirror positions Transformation Server as a real-time (or “right-time”) technology proposition: Deltas are captured as they are written to the database logs and replicated to distributed RDBMSes or middleware message queues.

Transformation Server isn’t exactly an immature technology offering, however. So just how improved could it be? Significantly, if DataMirror officials are correct. Beta testers report that the revamped Transformation Server uses significantly less CPU time while achieving several times the throughput of its predecessors, according to the company. There are a couple of reasons for the improvement, DataMirror representatives claim. First, thanks to Transformation Server 5.1.7’s new single-scrape capability, only one read of the DB2 logs takes place regardless of how many integration processes are running concurrently. The result, officials claim, is reduced demand on DB2 and the DB2 logs.

“To achieve the best performance and scalability, it’s essential to minimize demand on DB2 and the DB2 logs when scraping the logs for transactions,” said DataMirror chief scientist Herman Wallenburg, in a statement. “The new single-scrape feature ensures that the DB2 logs are read only once, even in the most complex implementations with large numbers of independent integration processes, thereby minimizing contention on the logs and dramatically enhancing throughput and scalability.”

Elsewhere, Wallenburg says, Transformation Server 5.1.7’s auto-restart capability should help boost its reliability and availability. “[T]hese large scale environments demand products that automatically adapt to and recover from network outages and scheduled or unscheduled system shutdowns. Transformation Server for z/OS’s auto-restart capability detects these situations and automatically resumes replication as soon as connectivity is restored.”