Business Intelligence Network
business intelligence resources

Blog: Richard Hackathorn

Main

Monday, 22 October 2007

Mixing Old Wine in a New Wine Skin

In a briefing by Vertica today, I revisit a number of familiar database concepts, such as column-oriented store and data compression. The list of players involved with Vertica is very impressive, from Michael Stonebraker as founder and CTO to Don Haderle as an advisor. However, I wondered how a new DBMS vendor could emerge successfully in a market that is consolidating. The question that kept bugging me was...

What is new about Vertica that was not invented and commercialized decades ago?

It seems that the appropriate analogy is that of taking several old wines and blending them together in a new wine skin. Here are the old wines:

1) Column-oriented store, which is great on query performance but terrible on update/load performance
2) Data compression, which is great on size reduction (10% to 20% of the raw data) but terrible on compress/decompress processing
3) Multiple sort orders, which is great for forming multiple indexes for complex queries but terrible on duplicating data
4) Dual data spaces optimized for reading and writing respectfully, which is great for absorbing an update stream but terrible on the query engine to operate concurrently on two different structures.

Mixed thoroughly together and pour into commodity hardware running Linux. And I must say that the resulting wine was... well... fascinating.

Vertica has been quietly selling product for three quarters and has about 50 customers. Their pricing is solely based on data volume, rather than the number and size of processors. Current partnerships include Business Objects, JasperSoft, Informatica, Talend, and interestingly Hewlett-Packard.

Vertica is a company to watch and expect a launch of a second version sometime next year.

Posted by Richard Hackathorn at 1:03 PM | | Comments (0)

Thursday, 23 August 2007

ParAccel: What is New and Different?

As I was preparing for a briefing with Kim Stanick, VP of Marketing, I kept wondering what was different about ParAccel. Started in 2005, this new start-up makes some bold claims about speed, scalability and simplicity. I am not against being bold. But where is the meat?

Their DBMS engine uses compressed in-memory columnar processing with a shared-nothing massively parallel processing. Column-oriented goes back several decades. Compression is not new. In-memory databases also go back several decades. And, MPP is becoming common. So, what is new and different? It might lie in their strategy.

ParAccel is targeting the medium-sized companies (and above) who are maturing their BI system and experience pain with performance and scalability. The sweet point is between a half terabyte to ten terabytes of data. Avoiding a rip-and-replace strategy, their approach is to augment the existing system through an Amigo arrangement. Based on analytical complexity, incoming queries are routed to the database-of-record or to the ParAccel database. ParAccel is also offering their DBMS as an appliance through reseller arrangements with major hardware vendors.

ParAccel is a company to watch. I am still searching for an answer to my question, which should emerge over the coming months as they roll out their product and secure satisfied customers.

Posted by Richard Hackathorn at 7:31 PM | | Comments (0)

Friday, 10 August 2007

How Can You Make Money Selling Free Software?

I guess that I am traditional because I have believed that software only has value if you pay a lot for it. As a BI professional, you probably have similar feelings.

This morning I had a briefing by JasperSoft as part of the Boulder BI Brain Trust. Paul Doscher, CEO and Nick Halsey, VP Marketing, gave an overview of their company, product line, marketplace and competition – quite a fascinating tour. But throughout, I kept thinking to myself, “How can you make money selling free software?”

Their main product is the JasperSoft Business Intelligence Suite Professional 2.0, which consists of the typical tools for reporting, analysis, ETL plus a server for corporate administration. An unlimited all-you-can-eat subscription is currently $35,000 – which is about 10% of a limited license for major BI suites.

Quite a deal, but here is the mystery… You can download most of this software for free from JasperForge.org which is the open-source community. So, why don’t people do just that? Well, most do, especially if they are just getting started, trying to understand the tools, or evaluating whether it satisfies their company’s requirements.

The beauty is that this lengthy fiddling-around time does not burden JasperSoft, as it would in a traditional sales cycle. In fact, these fiddling-around folks often buy training and documentation. There have been over six thousand companies who have bought something from JasperSoft.

When these folks are finished fiddling around and get serious within their corporate setting, then that price tag looks very good for a supported controlled-released version with a few critical proprietary extensions. As Doscher puts it, “Our customers are ones who value their time more than their money.” In a way, the software cost is more like an insurance policy to avoid or minimize nasty situations in the future.

So at this point, I am starting to get it. But, there was more!

There is a certain power behind a viable open-source community. If I am a smart CIO, I realize that I am not buying into proprietary software from JasperSoft, but I am buying into open-source software from JasperForge. Therefore, I must examine carefully the viability of that community.

The community contributes code that continually extends functionality, tests the various releases of that code, reviews the quality of documentation, and creates localized versions in 35 languages – wow! Then, JasperSoft comes around quarterly and takes the stable versions of the various components, wrapping into their professional version. There have been over 50,000 developers who have registered with JasperSoft so that they can download source code, post comments, contribute code changes, and receive the newsletter.

There is a certain power behind future OEM customers. The open-source code is distributed under a GPL license, meaning that you can freely take anything you want. But if you redistribute it, you must do likewise by licensing all of it under GPL. For traditional software vendors, this is a curse upon their proprietary software, even if only a small portion of the code is GPL licensed.

The trick is that JasperSoft owns all rights to the open-source JasperForge code, which is not necessarily the situation with other open-source code. Therefore, JasperSoft can issue a clean OEM license, for a small fee, to permit redistribution by other vendors. Over time, the JasperSoft code has been embedded in many other BI tools.

By the end of the briefing, I got it! The first level of revenue comes from a diversity customer base for training and documentation. A second level of revenue comes from corporate or governmental customers who want to avoid unexpected hassles from free software. And, the three level of revenue comes from software vendors who will embed JasperSoft into their products.

I will watch closely to determine whether JasperSoft can build a sustainable business in open-source BI as companies in the Linux marketplace have. This could herald a new business model that would impact the entire BI market.

Posted by Richard Hackathorn at 1:00 PM | | Comments (0)

Friday, 27 July 2007

Why did Business Objects Buy Inxight?

An interesting development in the Business Intelligence (BI) marketplace was the acquisition of Inxight by Business Objects (BOBJ). As a spin-off of Xerox PARC, Inxight has been noted for its deep technology in text analysis and visualization. Although it sold packages to many corporations, the primary business model was based on OEM technology licensing to larger software vendors who then incorporates juicy tech-tidbits into established product lines.

What is the leverage of these tech-tidbits for BOBJ? Adam Binnie, Senior Director of Product Management, explained that a key tech-tidbit is Awareness Server, which has been part of Inxight’s SmartDiscovery line.

The capabilities are impressive! Using a federated search across internal/external and structured/unstructured information, Awareness Server compiles the results in a meaningful manner using content filtering and relevance ranking. As an impressive addition, the View by Concept display organizes search results into a visual tree structure of matching categories.

It is obvious that the text analysis capability of Awareness Server is world-class. There is a lot of potential here to enhance the BOBJ product line with a unified search tool, making this powerful technology pervasive within corporate BI.

Many vendors have attempted to incorporate text analysis into mainstream BI tools but often with disappointing outcomes. What will be different about BOBJ’s effort? Or, will these gems gather dust in the bottom levels of some menu selection? A key issue is the degree of deep integration of Awareness Server into the Universe metadata. Will I be able to search (query) into my data warehouse as a big semantic web using the BOBJ infrastructure?

We should watch these developments at BOBJ.

Posted by Richard Hackathorn at 10:58 AM | | Comments (0)