I am happy and proud to announce the release of Nitro 2.0.
It has been almost two months since our last update to Nitro 1.2.9, which had fixed almost all of the problems our users were experiencing, and delivered best-in class performance for event log conversion. With Nitro 2.0 we are introducing a set of new features which will make working with log files more efficient and useful. And, as if that was not enough, we could dramatically improve Nitro’s performance and fix even more bugs.
Redesigned Analysis UI
A lot of people have told us that, while they love Nitro and enjoy the ease of use in configuring their CSV or Excel files for conversion, they found the Analysis view to be somewhat lacking. We designed Nitro as a tool for getting your data from CSV or Excel into ProM1 as fast as possible. And, for that purpose, providing a lot of information about the log data itself did not seem to be a priority.
But it’s not like we didn’t see a problem with that, and we certainly listen to your feedback2. We have been working hard on providing a more useful analysis view for quite some iterations. But each of the solutions we came up with so far was just not good enough, and we had to discard it. But now we think we have nailed the problem, and we are proud to present you a completely reimagined and redesigned log analysis view in Nitro 2.0!
Once you have converted your log data, you enter the Overview screen, which shows you an executive summary of your data. The chart in the center of the screen shows the density of events over time, but you can easily switch it to show the density of cases, or to show the distribution of case duration and the number of events per case. Nitro’s charts are also no longer static, and you can drill into the information a little deeper by hovering over them with your mouse pointer.
For frequency distribution charts, Nitro is the first process mining tool to now also provide Pareto Charts as an alternative to plain histograms. Pareto charts are, effectively, histograms that additionaly feature a curve for the cumulative percentage, which makes it easier for you to see what subset of, e.g., cases are responsible for most of the time spent in your process.
In a table at the bottom Nitro now shows you high-level information for each case in your log, including its start and end time, duration, and number of events. If you want to quickly find the longest-running case, or the one with the most events, you can easily discover those since we have included an in-line histogram of those metrics in the table.
On the left of the analysis view, you can navigate between the overview and more in-depth analysis views for activities, resources, and each attribute in your event log. The frequency distribution of the attribute’s values in the log is displayed as a Pareto chart or a regular histogram. At the bottom of the screen, a table explicitly lists all values for that attribute with their respective frequencies. As with the case overview table, we have also integrated an inline histogram which allows you to quickly spot the frequency distribution right in this table.
We have performed lots of process mining analysis projects with a number of tools, and getting a thorough understanding of the structure of your data is an essential task before you dive into in-depth analysis such as mining the control-flow graph. Back in the days, I have designed the log dialog for ProM 53. I still like that solution, and Nitro’s redesigned analysis view is nowhere near as comprehensive as ProM’s log summary. But I feel that Nitro’s new analysis view does a much better job at providing that essential, immediate feedback on your raw data.
We are of course very curious about what you think of our new analysis view! Please don’t hesitate to let us know in the comments, or drop us a mail – we are indeed listening!
A beautiful and useful result view is certainly a good thing to have, and we at Fluxicon always try to excel when it comes to usability and UI design. At the same time, we are also genuine speed freaks4, and if there is one thing we really hate then it is waiting for the computer to do its job. In fact, we think that the faster you get your job done and can shut down Nitro, turning to analysis, the better we have done our job.
The 1.2.9 version of Nitro already provided best-in-class performance, in that no other tool allowed you to convert your data faster. Well, that was just not fast enough for us and we are proud to announce that, with Nitro 2.0, you will generally convert your logs more than twice as fast! In addition to that, the memory footprint of Nitro 2.0 is at least 60% smaller than that of 1.2.9. It’s now truly about time to attack those monster logs!
In the above chart, you can see the runtime performance for parsing a CSV file and for exporting that data to compressed MXML, compared between Nitro 1.2.9 and Nitro 2.0. When parsing, Nitro 2.0 is 1.85 x faster than 1.2.9, and for export the speedup is even 2.54 x.
I have worked on the problem of storing and managing event logs for more than five years now, resulting first in ProMimport, then the redesigned log layer of ProM 5.x, and finally the OpenXES library, which powers log storage and management for ProM 6. Step by step, I had addressed a number of performance bottlenecks, and the design implemented in OpenXES is pretty fast. However, OpenXES needs to provide a generic solution, since it needs to support mining and analysis algorithms with very diverse requirements.
Nitro 2.0 is powered by Octane, which is our completely re-engineered log storage and management layer. The design of Octane approaches the problem of managing logs from a completely different angle than previous solutions, which enables it to provide that outstanding performance. One of the reasons for that performance boost is that Octane takes full advantage of the specific set of requirements present in an application like Nitro.
The above chart shows the runtime performance of parsing a compressed MXML file for Nitro 2.0 (powered by Octane) compared to OpenXES (as used in, e.g., ProM 6), and you can see why we love Octane: Nitro 2.0 is more than 8.3 x faster than OpenXES5.
For reasons that should be obvious, I cannot go into much detail of how Octane manages to speed past any competing solution we are aware of. Anyways, storing and managing event logs is quite a boring subject, since it is pure engineering – All the magic usually happens in analysis and mining algorithms. However, every major leap in that analysis and mining space has been enabled by introducing higher-performance log layers.
So now, more than ever: If you like your log conversion easy and fast, Nitro is your friend.
Full log format support
Starting with version 2.0, Nitro supports reading and writing every event log format used in practice. That means, in addition to reading event log data from CSV and Excel files, you can now also directly load MXML and XES files into Nitro. If you want to convert your old MXML logs to XES, or if you want to use Nitro’s new analysis view with your existing MXML or XES logs, no problem.
Also, most commercial process mining software unfortunately does not support the XES standard yet. And while we hope that this will change sooner rather than later, it makes no sense to wait, and we at Fluxicon like pragmatism. This is why Nitro now also supports writing logs to CSV, enabling you to analyze your XES and MXML logs with software that can only handle CSV data.
All bugs that users have reported with version 1.2.9 are now fixed in Nitro 2.0. Also, we have discovered a number of bugs that nobody has seen in use yet, and of course we have fixed them right away.
So, with Nitro 2.0 you will only get brand new bugs 🙂 And if you run into them, please do us a favor and tell use about it, either via Nitro’s built-in feedback feature, or by sending us a mail to email@example.com. We hate bugs with a passion, and we do our best to fix them and update Nitro as soon as we can. Promised.
You may have already noticed that we have adjusted our pricing for Nitro. Our previous pricing structure was aimed at power users in the enterprise space, and for some people that are professionals, but not enterprise users, that just didn’t sit right. We agree with you, and our new pricing structure includes a new plan for people like freelance consultants, who don’t need that same level of support, and whose requirements are generally more modest.
That being said, the new pricing structure looks as follows:
Nitro is free in demo mode, which allows you to convert a maximum of 10,000 events per log. This should be enough to take it for a test drive and see whether it fits your needs.
With a Professional ticket, you can convert up to 10 million events per log. This ticket includes standard support, and costs € 495 per month. Yes, that is not cheap, but neither is your time as a skilled professional. Whether you are a consultant, or you are working on a skunkworks project in your company, we think that you will easily save that money with the time you don’t need to spend customizing or waiting for other solutions.
Our Enterprise ticket removes all restrictions and grants you priority support. If 10 mio events just doesn’t cut it for you, or if you need enterprise-grade support, this is our premium solution for you. The enterprise ticket is priced at € 1950 per month.
If you can’t find yourself in the above plans, please get in touch. We know that everybody’s requirements are different, and we will try our best to find a solution that works for both of us. But we don’t like to force you into the typical enterprise sale where, in order to get a quote or even look at the product, the seller forces you into lengthy discussions with their sales staff and consultants. So, if you like it simple, use our standard plans above. If your requirements are different, please get in touch with us at firstname.lastname@example.org!
Also, if you would like to use Nitro in education or for your academic research, you can get a Nitro community ticket free of charge. We also support selected non-commercial use and projects we like with free community tickets. Please get in touch with us at email@example.com, and we will get you going.
The default log visualization of ProM 6 is also closely based on that initial design. ↩︎
Not in the “snorting amphetamines until the early morning” sense, mind you ↩︎
Just to be clear here, I am not saying that OpenXES’s performance is bad in any way. That would be kind of stupid since I wrote most parts of it myself, and not that long ago. But it shows that, when you don’t have to support a generic use case, and the many features needed for academic research, you can drastically improve performance. Which is what Octane obviously does. ↩︎