Tag RapidAnalytics

Posts: 7

RapidMiner: A Leader in Advanced Analytics

Last week the news broke that Gartner Research placed Rapidminer in the Leader Quadrant for Advanced Analytics. I can't say that I was surprised at the news, but I was extremely happy to know that we share the Leader Space with names like IBM and SAS.

The report is a great read about all the competitors in the space and the breakdown of their pros and cons. It's a very comprehensive report and I urge any interested readers to download the free report off the Rapidminer site. You can access it here.


The RAPi Project Update

With every few steps forward on this project, I inevitably take a step back. On Friday night I successfully created a 2 node Pi computer. I tested it with the MPICH2 script and it indicated that both nodes, RApi0 and RApi1, were talking to each other. This was a HUGE accomplishment for me.  Just to get this point was a pain in the butt because of SD card corruptions and faulty image writing, but the time was well spent.

Saturday evening I then started the process of installing Apache, PHPmyAdmin, and MySQL, as the backbone for getting Rapid Analytics 1.3 to work. Everything installed well and then I downloaded and installed Rapid Analytics.  I went through the configuration quickly and set the hostname to RApi0. Then I set my parameters to load Java and run the Rapid Analytics JAR file. The server started within 6 minutes! WOW! Awesome! But then I couldn't get to view the configuration screen in the browser. I was pointing the URL to http://localhost:8080.  At this point it was 1AM and I was tired, I decided to power down and restart tomorrow with a fresh mind.

Sunday morning I restarted both RApi0 and RApi1 and fired up JAR file again. This time, it churned and finally crashed out. Out of Memory Exception. Damn! What went wrong?

My initial thoughts on what went wrong was this, and I welcome any of my readers to chime in with suggestions (please!):

  1. I installed MySql, Apache, PHPmyAdmin, and Rapid Analytics on my head node, RApi0, and when I powered up on Sunday morning, MySQL and Apache started on RApi0. I believe that this could be an issue because it didn't load into the shared resources of RApi0 and RApi1. I welcome your thoughts!
  2. Setting the hostname in the RapidAnalytics 1.3 to RApi0 meant that I should've pointed my browser to http://RApi0:8080. Ugh.

My next steps is to power up the 2 node Pi and manually start up Apache and MySQL once both Pi talk to each other, then start the JAR.


The RAPi Project

My initial setback in running Rapid Analytics on a Raspberry Pi might end up being just a small stumbling block in the end. The only thing that's preventing Rapid Analytics from running on a Pi is hardware, a single pi is not quite there to run both X Windows and Rapid Analytics effectively.

Yes, I could put Rapid Analytics on a bigger and more powerful machine, but that defeats my minimalistic approach to this project. I want to run Rapid Analytics on the most cheapest and minimalistic set of hardware out there, and using a Pi is a great way to do it (cheap and damn minimal).

After a conversation with my IT director, he suggested clustering the Pi's together. There is a great article out there about a student, Josuha Kiepert, that built a 32 node  Pi cluster. That cluster runs at 10 Gigaflops/sec. That's INSANE!

I started pulling together information on hardware and software to do this.  I think I can pull this off with 4 Pi's clustered together using the Tomcat software. Tomcat is a system to run java in a distributed fashion across several bits of hardware, in this case 4 Pi's.

I put in another order for 2 more Pis with power and ethernet cabling, an ethernet switch, and some more SD cards.  Total cost is $140 shipped, plus the other 2 Pi's @$35 each + cables and SD cards.  I'm looking at potentially having a capable Rapid Analytics Pi cluster for under $400.

I'm calling this the RAPi project!


Rapid Analytics: On a Raspberry Pi – Part 3

Update: After talking with some IT folks, I realized that I might just be able to make Rapid Analytics on a Raspberry Pi (RAPi) happen. I need to go back a split the memory better on the Pi AND possibly make mini clusters of Pi.  How cool would that be? Stay tuned for more on this project.

It is with heavy heart that I must report that my Rapid Analytics on a Raspberry Pi experiment has failed. It did not fail from a installation or configuration standpoint - that went extremely well - it failed when I tried running it.  It was loading into memory and proceeded to hit a resource brick wall. After 40 minutes of just churning the CPU and going nowhere, I killed the process.

This experiment failed purely from a hardware and memory resource aspect. I knew that Rapid Analytics is resource intensive, and that the odds were against me, but I decided to try it anyway and share my results with  you.

Here are a few things I've learned from this experiment:

  • Raspberry Pi is a very capable computer,  you can do a lot with it provided you figure out a way to keep your resource fingerprint small;
  • You can run MySQL and Apache from the Pi quite easily and installation is a breeze,
  • The ARM optimized version of Java works great on the Pi;
  • Rapid Analytics 1.3 can be installed and configured, but hardware limitations prevent it from running;
  • You can't use the run.sh to start up RA, it gives you a VM not supported error, you have to use: sudo /opt/jdk1.8.0/bin/java -Xms256M -Xmx496M -jar run.jar,
  • I got an appreciation of what Java can do and how it works, and;
  • I cant help myself, I'm addicted to Rapidminer and Raspberry Pi.

photo(A successful install but an ill fated run on the Pi - sigh)


Rapid Analytics: On a Raspberry Pi - Part 2

I had some great success last night installing Rapid Analytics 1.3 (RA) on the Raspberry Pi. I originally tried to do all this via SSH (headless), but I was running into the X11 error on my windows machine. In the end I had to connect the Pi to my large screen TV and start X windows from there.

I got RA to initialize into the configuration routine right after I opened a terminal window but got stopped in my tracks when it asked me for what database I want to use.  Shoot, I didn't install any!

photoSo I spent the rest of the evening downloading and installing MySQL, PHP, and Apache before running out of space on my 2 GB SD card. I have too much crap on the card from my tinkering projects to finish the install. Double shoot!  Looks like I need to buy a 16GB card now!

I learned two things so far from this project: 1) you'll need the ARM optimized Java version for the Raspberry Pi, and 2) always get a larger SD card.

Although the RA configuration routine initialized quickly on the Pi, it remains to be seen if RA can run well on it, so stay tuned for more RA and Pi (RAPi?) fun.


Neural Market Trends is the online home of Thomas Ott.