Content Copyright © 2008 Bloor. All Rights Reserved.
This blog was originally posted under: The Norfolk Punt
Well, Sybase is telling a good story at TechWave – but the devil will be in the implementation detail, so it’s good to talk to customers. But aren’t the customers at such events carefully selected? Of course—but there’s a difference between selecting the top 10% of maturity/ability in your successful customers (things might not go so well for less able, or less lucky, customers; but that’s always true) and bribing customers to say nice things about your products with special deals—and you can usually tell the difference by talking to them, especially if you can take things off script.
So, I talked to Bodo Saar, Senior Consultant, Methoden & Qualitätsmanagement in der Softwareentwicklung about his implementation of PowerDesigner for a large but unnamed financial institution. This was replacing a previous requirements and analysis environment , which used Word, Excel and Visio; plus some specialist data modelling tools (a surprisingly common scenario I suspect). This previous environment was characterised by a lack of any real automated support for consistency and completeness checking, leading to problems with the quality of the requirements and analysis models, which nevertheless took lots of time to produce and manage.
What is clear from talking to Bodo is that PowerDesigner can address this situation and support a sophisticated, high-maturity requirements modelling and analysis process effectively, using an in-house developed framework and producing a high-quality specification that can be handed over to a separate group for design and implementation. It is certainly capable of a lot more than just data analysis.
The test results after design and implementation steps are handed back to Bodo’s group for testing and what is particularly impressive is the way in which they have customised PowerDesigner to support traceability from test cases back to requirements, using the in-house framework. According to Bodo, PowerDesigner is capable of supporting a rich set of dependencies both within models and between models and, in short, does the job. I am impressed by its apparent flexibility and process/framework neutrality.
What is less clear is whether PowerDesigner is the only tool capable of doing this, these days anyway, since Bodo doesn’t have experience of some of the possible competition (Word, Excel and Visio may be common requirements tools, but are hardly “state of the art”) and wasn’t involved in the initial selection of PowerDesigner. It also appears that it is being used in a “high maturity environment, with effective senior management direction and “buy-in” for organisational and process restructuring, via a technical platform implementation “task force”—which is probably the ideal environment for such a tool.
Nevertheless, I was still impressed, since at least some other tools I know tend to impose process on their users, and PowerDesigner has demonstrated its flexibility and process neutrality in the real world. And Bodo says that PowerDesigner supports the IT implementation end of Requirements Modelling particularly well compared to other tools he’s used, which have grown out of Business Process Modelling with technology implementation support bolted on.
I also talked to Asif Rahman (Director Information Application Development, Loan Performance ) about Sybase IQ. This was introduced in his organisation as a higher-performance replacement for SQL Server, without compromising the rich functionality users liked with the SQL Server solution.
Once again, it is clear that Sybase IQ does the job but not that it is the only product that could (the choice was also influenced by non-technical factors such as Sybase’s good reputation in the financial world). Besides, SQL Server may have improved since Sybase IQ replaced it at Loan Performance. Although, so probably has Sybase IQ; and Asif is far from confident that even the latest SQL Server Enterprise Edition tools could cope with his demands (and the more I look at how good Sybase IQ is at a detailed technical level, the more I’m inclined to agree with him). Nevertheless, the performance improvements Asif reports (of the order of 7 times) are impressive, and his other reasons for choosing Sybase IQ are persuasive:
Sybase’s commitment to the product and its customers: it’s “more partner than vendor” Asif says;
- Its use of standard SQL, meaning that modification of existing queries was minimal (but this might not have been the case if migrating from, say, Oracle);
- Its “multiplex” scalability, just by adding extra hardware to handle extra queries;
- Its ability to handle extra columns in existing tables as required, rather than requiring the addition of extra tables solely for performance reasons;
- Its ability to handle both large and small queries at the same time without special tuning (Asif has an extremely mixed and ad-hoc workload). Asif says that in his experience, Netezza, for example, deals with very large queries by allocating all the CPUs available to them, which kills concurrency and performance on other queries running at the same time, although this would be fairly easy for Netezza to address in firmware.
- Its use of standard hardware.
So, did Asif find a downside with Sybase IQ? I asked about load times and this doesn’t seem to affect his usage of Sybase IQ (although Billy Ho, Senior Vice President, Product and Technology Operations at Sybase tells me that improved load times is the enhancement most wanted by Sybase IQ customers generally; this is addressed in Sybase IQ EEE 12.7, which should be available Q3 this year).
However, although he appreciates Sybase IQ’s scalability (just add extra nodes and queries will distribute themselves across them), he would like the capability of running very large queries across several nodes at once (Billy Ho tells me that this shouldn’t be hard to do, once higher priority enhancement requests are out of the way). He did find that Sybase IQ’s column processing capabilities took some getting used to initially—if another attribute of an entity becomes important to your analytics, you can simply add another column to a Sybase IQ table without worrying about the possible performance implications—but you could argue that having to worry about the number of columns in a table is simply an implementation issue (fault) with other, less sophisticated, database technologies.
So, overall I found the user endorsements of PowerDesigner and Sybase IQ at TechWave pretty plausible. There was certainly more behind them than just a scripted presentation, when I talked to these people face-to face.