The Boulder BI Braintrust hosted Confio this week. Confio is fairly small (25 people focusing on product development sales and support, with most other infrastructure outsourced), and privately held with angel and operating revenue funding it. The value proposition of their product, Ignite Performance Intelligence, (I’ll call it IPI from here) is to deliver information about database performance using a more BI-like interface. Prospective buyers (and today’s customers, who number in the hundreds) are DBAs and application performance specialists.
IPI is used with Oracle, DB2, SQL Server and Sybase databases, and there is an add-on for Oracle business applications. IPI collects statistics from internal tables and puts them into conventional BI cubes for the reporting interface to present to users. Confio claims less than 1% overhead on the database to do so, which is credible, and it installs no internal agents – it simply runs queries. Its customers are using IPI on transactional applications roughly 2/3 of the time, with the other third being BI applications.
While other application monitoring products see the database as a “black box” and typically measure when it’s called and when it responds, IPI looks inside, and measures 14 dimensions. It measures wait time for all the steps of execution inside the database, traces them by time and date, and links them back to the user. This permits performance tracking over time (every Monday at 10 we have the following problem), identification of files or objects contributing, etc. The presentation of the information is certainly more BI-like than many other performance monitors. Confio has excellent reference quotes from customers on its site about the savings they achieved using the product.
Confio has challenges, though. It fails to present its information in the context of what matters most. IPI needs a front end that establishes KPIs for the IT staff, defining them based on SLAs, executive clout, operating costs, or whatever the governing rules are. Presenting a fairly typical dashboard as an entry point would be far more useful than an undifferentiated ranking of statistics about elapsed time or other measures. But understanding what it means, stoplighting or exceptions against established benchmarks would add a layer of value and move Confio to a nearer-to-real-time mode. (Prediction is a harder problem, but would also be useful.)
DBAs are busy; they will not spend their time simply picking the thing with the biggest bar on the graph and tuning it. And they often don’t have the buying power; Confio acknowledges that once DBAs take the download and apply it, Confio needs to work with the DBAs to sell the value of what they have achieved internally to a buyer. Confio needs to put that into the product itself. Prioritization is the heart of BI, and if Confio truly aspires to applying BI thinking to performance management, they need to attack it next. So far, what it offers is backward looking; that is something it shares with a lot of BI, but it’s not what Confio should aspire to.