Jitterbit streamlines data migrations

Straightforward graphical tools come close to satisfying "no expertise needed" promises

At a previous job, a colleague told me how she went home one night and lovingly described the arcane parameters of our assignment to her husband: Data from four computers must be massaged into several versions of XML before being passed off into a primitive version of a Web service. She was steamed when her husband’s reaction was, “So you’re just building a data pipe.”

Mundane as it may seem, the problem of moving data from system to system is much more than “just building a data pipe.” There are deep challenges to getting the information from one system to another without being thwarted by null pointers, character format errors, database incompatibilities, and any number of other things that can go wrong. Jitterbit aims to make the job simple enough that, as its Web site says, you can “integrate your data ‘out of the box’ without any coding expertise.”

Streamlined data, no programming?

Jitterbit 1.0 comes with a server and a client, and both can be installed on Windows or Linux boxes. The packages come either as the free “community edition” protected by the Jitterbit Public License (an open source license similar to the Mozilla License), or as a “professional edition,” which is the same software but includes support and training.

In this architecture, the server does the work, and the client offers a GUI for directing the server. You create a set of “operations” that apply “transformations” to data from a source before sending it to a target. Each of these are encoded in XML and are uploaded to the server that runs them according to a schedule you set.

There’s really no traditional programming involved, which is part of the system’s appeal. You set up your sources, transformations, and targets by filling out forms and selecting options from menus in the Jitterbit UI. Many of these menus are fairly intelligent and draw their information directly from the data sources. The database source, for instance, will look up the tables and columns so that you can choose them instead of typing them in from scratch.

When you’re ready to move the data, you select one part of the data from a source, one part from a target, and then push the “map” button to connect the two. When the server runs the operation, it will copy over all data elements in this map from the source to the target.

The folks at Jitterbit have clearly built more than a few data pipes in their time. I’ve found that much of the programming involves writing code, compiling, uploading, and then finally testing all of the particulars. My favorite part of the Jitterbit interface is the testing tool, which lets you know immediately whether the current settings for a source, an operation, or a target are valid or invalid. When you type in the parameters for a database connection, you can test to see whether the link is valid with a quick push of the button. This testing loop is much simpler than compiling and deploying the application, and for me, it remains the system’s biggest advantage.

Embracing Jitterpaks

Jitterbit encourages users to share some of these integration operations with one another by providing an abstracted version of the XML that acts as source code, known as a Jitterpak. If you want to make it easier for others to follow in your footsteps, you can publish a Jitterpak with instructions for interacting with a certain data source, and others can start using it immediately.

The company is offering some basic versions that interact with Web services from Salesforce, SugarCRM, and Amazon. One nice touch is that Jitterpaks can be distributed without sensitive information, a feature meant to make it simple for users to share the essential details of linking data sources without revealing their passwords.

Jitterpaks, incidentally, are limited to sources and targets built into the current version of Jitterbit: Web sites, FTP sites, Web services, and ODBC-compliant databases. I’m hoping that future versions will include the ability to move information in and out of other file types, including e-mail, Excel files, and PDFs.

There are limitations. Although Jitterbit can pull information from Web sites, there isn’t any way to parse the often arcane and annoying HTML surrounding the data. More expensive tools such as Kapow beat it easily at extracting data from complex Web sites, and some of the sophisticated reporting packages such as Cognos make it much simpler to extract information from complicated schemas in databases. Jitterbit’s user interface is not bad, but it’s not as simple as some of the more expensive reporting packages.

I tested the system by building several data pipes, a process that took only 15 minutes. I found that Jitterbit works well when you’re sticking to the beaten path: If you want to pull data from one machine, extract strings from one part of a data structure, and place them unchanged into another standard structure, it’s very simple. The challenges crop up if you want to make a few subtle changes to the data along the way, perhaps by changing a character set — the mechanism for actually tweaking the bits as they flow by just isn’t as robust.

Simple is as simple does

Even when I try to be as unbiased as possible, I still find Jitterbit’s claim that the tools can be used without coding experience hard to buy. There are too many data-transfer glitches caused by odd character sets, faulty database connections, SQL anomalies, and other annoying facts that only a programmer would know about.

But that doesn’t mean the tool won’t make both casual and serious programmers pretty happy with a set of straightforward tools for extracting data from one format and putting it into another. It does come quite close to satisfying the “no expertise necessary” promises, even though I feel that it really helps to be a programmer to get the most from a tool such as Jitterbit.

If your data pipe needs are more complex, watch the project for new features. The Jitterbit representative I spoke with said that many of the items I wanted would be arriving with Versions 1.1 and 1.2 — and that those versions would be appearing in “weeks,” not months or years. I expect Jitterbit’s solid framework will expand rapidly as they build it out.

InfoWorld Scorecard
Ease of development (30.0%)
Performance (15.0%)
Value (10.0%)
Documentation (15.0%)
Capability (30.0%)
Overall Score (100%)
Jitterbit 1.0 9.0 8.0 9.0 8.0 8.0 8.4