Data!

Hey everyone,

So Jolie and I have data collected (again) with SpecJBB as part of our trace set. We ran all of our traces through our classification algorithm and now our task is getting our data into LaTex tables and then analyzing them. It’s not terribly exciting, but hopefully we’ll get finished sooner than we did last semester, since Jolie made a parsing script that gets the relevant data we want to look at. It’s also easier because we can just copy-paste the data into the tables I made last semester. So yay! The faster we get this analyzing and data collecting done there’s more of a chance for us to actually write a technical paper of our research for the past year. 

Also, Dr. Rivoire and I are in the process of finishing up our Grace Hopper application. We turned in the actual poster submission, but now we’re doing the scholarship application so I can get funding to actually go to the conference. This pretty much involves a letter of recommendation (Rivoire’s court), a 500-word essay on why I want to go and what I’ll do with the experience (yay feminism!), and my resume. Pretty much the hardest part — my resume’s pretty dinky considering I’ve only had two real significant jobs in my lifetime, but at least those jobs were awesomesauce. Making a resume is so difficult though. No me gusta.

Anyway, just thought I’d update you all on the happenings around here, but that’s pretty much it (unless Jolie wants to talk about her conference experience, maybe?). Toodles!

Conference Preparations!

Hello everyone!

Jolie and I are still hacking away at our abstracts for our respective conferences that we’re going to (Jolie at Northridge, me at GHC). Jolie’s pretty much done with her two paragraph summary and I believe she’s started creating the basics of her poster. Rivoire and I are still going through my paragraph to tweak the wording and whatnot. I have an outline for my poster, but I still have to write the overall, 2 page abstract.

Since Jolie is graduating, she’ll be going to Northridge this semester. I’m not graduating until next fall, which is just as well, because the GHC conference isn’t until October, I think. It’s a little strange that GHC wants the abstracts submitted so early for a conference so late, but at least I’ll get everything done earlier rather than later.

Anyway, we’re in the process of completing our last run of SpecJBB (should be finished already, actually) on our first machine, Lolcat. We started installing Spec on Rageface at the same time so we could immediately switch meters and jump into collecting three runs again. Installing Spec is a bit of a hassle since we had to update the version of Java JDK and JRE we were using. Plus the installation itself takes a super long time, so we just left it alone while we worked on our abstracts.

Tomorrow we plan on running Spec on Rageface for real. We have to configure the scripts we use, but that should be somewhat easy since we already did so on Lolcat.

Well, that’s all for updates now. Toodles!

We’re Back!

Hello everyone!

I hope you all had a nice winter break (or, for those of you who aren’t in school, a nice vacation and Christmas/New Year’s). Jolie and I definitely did.

Now we’re back and we’ve got a lot of stuff planned. First, we’re going to start making runs of SpecJBB, our new benchmark, on our different machines. We had a trial run to see just how long Spec actually took without any restrictions on time and it ran for about three and a half hours. It definitely seems like a long time, but it’s still reasonable enough for us to actually start gathering data. We’ll fuddle with different parameters (we’re looking specifically at how much memory Spec gets) once we have our initial runs.

While Spec is running, we’re going to start our preparations for our different conferences. This semester Jolie, while she’s still a student, will be going to the Consortium for Computing Science in College and I will be going to the Grace Hopper Conference in the fall. So we will be preparing abstracts and posters to submit to these conferences. We’ll be going over work that we’ve done last semester as well as last summer — our work has a lot of components, so there’s no shortage of topics to choose from.

Anyway, sorry the update is so short — hopefully we’ll have more to tell you later!

–Rachelle

Gathering Data Like You Wouldn’t Believe

Hey everyone, Rachelle here.

So Jolie and I have spent the last week trying to collect more data for our classifications. We’re using Random Forest to figure out if using MMSD normalizations (mean, median, standard deviation) with a single other statistical feature (ie kurtosis, skew, etc.) will have any significant changes in our error rates. However, trying to collect all those calculations has turned out to be quite difficult. Jolie ran half of the calculations on her computer and they didn’t finish until seventeen hours later. Half of the calculations took nearly twenty four hours. Crazy right? 

I was supposed to do the other half of the calculations, but we (Rivoire, Jolie, and I) decided that we should get preliminary results first to decide whether or not this is the path we even want to go on.

We also decided to cut our power traces into smaller chunks, to see if our classification algorithm could sort the traces with smaller data sets. If an application suddenly cuts short before it finishes and we have only its chopped power signature to figure out what application it was, it’d be nice if our classification algorithm could actually work with that smaller data set. So we cut down the traces by 20% and 50% to do this.

However, we hit a bit of a snag when we realized that the 50% trace results looked a lot like the 20% trace results. Like, exactly alike. Come to find out that we’d accidentally miscalculated the 50% data set, so now we have to run the tests all over again, which is what Jolie is up to right now. 

In the meantime, I’m going to be applying for a grant to get SpecJBB2013 — a benchmark which we have from 2005, but which won’t work with the OS on our machines. Also, when the data collection is finished, Jolie and I will be going over it together to determine which direction we should go in next.

That’s all for now! Toodles.