Analysis: Implications of Informix Benchmark beating Oracle MDM Benchmark

Last week IBM released a benchmark which we feel really shows the potential for smart meter data management in the real world.  It showed that IBM Informix TimeSeries software could enable the AMT-SYBEX Affinity Meterflow meter data management (MDM) application to load and validate up to 420,000 data transactions per second, a significant increase on the speeds demonstrated by any previous benchmark test.  It showed that high performance could be maintained and scaled up to 100 million meters, also more than any previous benchmark. 

The Race for Smart Meter Data Management is on

They aren’t alone in having sought to stretch the boundaries. Industry players have been working to demonstrate the potential for meter data management in benchmark tests for several years now.  The ‘rules of engagement’ were first devised back in 2006, when a benchmarking methodology known as ‘A Day in the Life’ was devised. This split the process of meter data management into two phases: Collecting and Validating the meter data for a community of X million meters and calculating billing data for 5 percent of them per day (so that within one month all meters are billed).   The headline figure tends to be the number of meter readings the system can manage per second or hour, rather than the billing calculations.

Since 2006 rising load and validate rates have been demonstrated, starting  at 23,000 transactions per second back in 2006 up to 47,500 per second in May 2011 and then in a big leap  to 277,777 transactions per second cited by Oracle in August 2011.  Using the AMT-SYBEX MDM application IBM has now benchmarked performance of 420,000 transactions per second.

How have IBM and Oracle achieved these increasing levels of database performance?

Traditionally, loading speeds into a relational database are limited by the need for indexing. This is particularly onerous when dealing with interval data that needs to be managed as a timeline of changing values. So, how has each got round the indexing overhead experienced by relational engines, structuring a time line of rapidly changing values? When we look behind the benchmark headlines it is clear that the approaches are completely different.

• Oracle has addressed the issue by accepting the need for Indexing that outweighs the raw data and focused on brute force. It has used hardware developments residing in Exadata servers to compensate for the inefficiencies of the standard relational model. Indeed its own Webcast points out that without its Exadata hardware the performance is back to the 10,000s rather than 100,000s of transactions per second.
• The IBM approach is quite the opposite. It has focused on enhancing the relational database from within its portfolio which is most suited to the task. Informix is already one of the most hardware-efficient, fastest databases around, and it has added a feature to it which removes the need for excess indexing entirely. This feature, which is called TimeSeries, in effect creates a special Index that registers the header data once and stores the data updates as a timeline in the same record.  TimeSeries and the ability to time link data is a fully integrated, native feature of the database.  Not only does this give a load and validate speed advantage but also significantly reduces the amount of data stored.

Which approach best fits the (utility) bill?

The next generation of application architects must decide which approach works best for them. Oracle has a market weight and track record around database solutions that IBM has not yet demonstrated to such an extent. If decisions by utility CTOs on which platform to choose are based only on the pitches of database-savvy salespeople then Oracle certainly  has the capability to engage the market (even though that capability is gained through hardware evolution rather than database developments).

IBM, despite a slightly chequered success with DB2 and an almost silent stewardship of Informix for the past decade, has nonetheless continued to invest constantly in the platform, and encouraged Informix engineers to continue innovating within the IBM family.  What that has meant is that Informix has remained a highly relevant database with a significant installed base of happy users.  By continuing to innovate, it has come up with by far the most elegant solution to the relational limitations when it comes to smart meter data management, and set a benchmark which will be hard for others to beat for some time to come.

There is, of course, a drift towards using all-Oracle solutions for some enterprises, and cash-rich Utilities may well buy into the promises of fully integrated applications, database and hardware appliances which link seamlessly into other Oracle-enabled enterprise applications and middleware. However, IBM is no slouch when it comes to enterprise-wide integration, and so it remains to be seen how these two giants battle it out in the Utilities space.

ISV innovators’ budgets may make the choice for them

The future is, perhaps, even more interesting to contemplate.  The potential is immense not only for smart metering but a whole new generation of “Smart” solutions that will face the same interval data challenges when helping us to manage resources and supplies of all kinds. Many of those new applications will come from ISV innovators -- so will innovators all be able to afford the Oracle route to goal? We aren’t sure. It may be that their budgets, and those of their customers, make the choice for them.

The emerging and far broader application space will not be about cash-rich enterprises but by innovative ISVs for whom the entry price of an Exadata-based solution is completely out of their budget ballpark – and often completely out of those of their customers.  If enabling a smart data management model the Oracle way carries a starting pricetag of around USD 2 million for a quarter rack of hardware with a database (when the latest benchmark to get to 277,000 transactions per second took a full rack vs Informix on a 16 core P Series) it is hard to imagine that this will be feasible for all the potential innovators and users.

At Oninit, we firmly believe that Informix TimeSeries software is most likely to offer the direction with most flexibility and potential to innovate in this space.  We regularly work with ISV innovators, development partners and our customers to explore ideas and create proofs of concept, as well as with young and first-time Informix developers who want to take advantage of Informix’s free point of entry and open choice of standard hardware platforms.

The future is smart – and we look forward to working with smart thinkers to make it happen.

Click the link to read the full benchmark report Unprecedented Performance and Scalability Demonstrated for Meter Data Management.

Read our main article AMT-SYBEX and IBM show that Informix performance is ahead of the pack for smart meter data management for the main story about the benchmark.