Null Disquisition

After a few days of tinkering with EC2MPI, I spent some time polishing up a stat mech MPI simulation. The code in question is a 2d Ising model simulation using Replica Exchange. Right now it stands at around 400 lines of C++ using STL vectors (which I love). Once I know it works (or at least works well enough) I might post it up here, but for now I’m just trying to generate pretty hysteresis plots and observe the critical behavior of a 2d Ising model system. Here’s a picture with points on it.

I leave the interpretation to you. The best part of this is that I can do these MPI runs without burning a hole in my lap (the MacBook gets rather warm). -David

For my Master’s thesis, I’m going to be running a lot of MPI code, and naturally I need a place to run it. Let me first say that my university has an excellent high-performance computing center run by one of my committee chairs that is more than capable of serving my needs – but yet, I am unfulfilled. With our scheduling system, there is a “backfill” that is always available for running small jobs (like the ones I run), but for my thesis, I want to test the massive scalability of an algorithm (Replica Exchange). When I mean massive, I mean massive – think 1000 compute nodes or more.

Big ideas, people.

In order to satisfy my need for a massively parallel platform, I looked no further than Amazon EC2. As should be apparent from many of my previous posts, I have been doing a lot of work with Amazon’s cloud services – both school and work.

A few weeks ago, I started an MIT-licensed open source project on GitHub aptly named EC2MPI. Today I made a major step forward with this project which was the motivation for this post. I finally have everything configured properly and got my first no-hassle MPI cluster up and running.

The script I wrote (EC2MPI), is written in Python and presents an interactive prompt to the user. You select the architecture (i386 or x64), the number of instances, and I also have support for user-defined SSH keypairs (not AWS keypairs) for cluster security. The instances are spawned, and EC2MPI sets up the SSH keys, as well as MPI configuration. It is so freaking sweet.

I wanted to share some issues I’ve had so far while developing this and how I solved them.

Intra-EC2 communication – For this, I needed each instance to be able to talk to one another for point-to-point as well as collective communication. My solution for this was to allow the user to generate SSH keypairs which were stored in a private S3 bucket (owned by the user). My user-data script sent to the instances took care of downloading and installing the keys upon startup.

Shared storage among instances – In order to run MPI code, the nodes in the cluster need access to a shared storage volume which will contain binary files compiled by MPI. Since EC2 has no shared storage (for now), I had to find an alternate solution. The solution I settled on was to use s3fs: a fuse-based filesystem which allows you to mount an S3 bucket as a volume. Reading and writing to the shared volume is pretty slow (unless it’s cached), so for certain kinds of code this might not be ideal. However, I believe it is the best solution for now. I imagine one day Amazon will add a feature to the Elastic Block Storage volumes that allow them to act as shared volumes.

Starting up and tearing down clusters – I used Amazon SimpleDB to keep meta-data about the cluster: how many instances are in the cluster, internal/external IP addresses, etc. This is also how I define the master node and worker nodes. This will allow me to add features such as adding and removing instances from a cluster without having to tear the whole thing down. Also I did all startup config with a user-data script so the script does not have to log into each instance upon startup. This allows the clusters startup to scale well.

Check back soon for some benchmarks and more detailed write-ups as the project progresses. First, I need to get my maximum number of instances increased (right now I can do 20 max). Fast times ahead, friends.

-David

This morning I wanted to mess around with some MPI code (Message Passing Interface). We have an high performance server farm here at FSU (aptly named the HPC), and I know I could have used that to test some MPI code – but using provided facilities is so boring. Plus, I wanted to try installing MrBayes in multi processor mode and distribute it over a few machines. I don’t think the admins at the HPC would like me doing that (if it’s even possible).

Show Accidental Tutorial

Hide Accidental Tutorial

So I head on over to Amazon and sign up for my own personal AWS account. Sign up was very simple, as was generating my pk and cert files. Now, I’ve done all of this before so it’s not news to me, but I’ll review anyways. After you sign up, you head back to your AWS account page and generate and download two pem files: pk-somereallylonghash.pem and cert-somereallylonghash.pem. You don’t need these to spawn instnaces, but you will need them for other things. I like to keep them in ~/.ec2

mv pk-somereallylonghash.pem ~/.ec2
mv cert-somereallylonghash.pem ~/.ec2
chmod 600 ~/.ec2/*.pem

It’s good practice to store things like keys and certificates in 600. In fact, ssh won’t let you use a private key if it’s not 600. Next, head back to amazon and get your account Id and super sercret password. They will look something like:

SOM3ALLC4P5L3TTERSANDNUMBERS
Som3mixeD(4sl3tteR$||um8er$and$ymb0ls

Now, I wouldn’t recommend saving these on your hard disk. I keep them in an encypted email. It would also be a good idea to print them out and keep them somewhere (it’s a pain in the ass to get a new password and go and change everything). At this point we have everything we need to spawn instances, bundle a new image, store things on S3, deliver content through the CloudFront (which is awesome btw), and anything else AWS has to offer. In order to get your spawn on, we need a tool of some kind (because no one wants to use the stupid command line tools). Elasticfox has pretty much taken it’s place as the singular tool for interfacing with EC2 services. It is incredibly useful – good design, open source, has constant updates, and they have a very fast turnaround time between Amazon releasing a new feature and Elasticfox having support for it. I think Amazon has commit several developers to it, but that’s just a guess. One more annoying step before we can launch an instance and log in. We need a keypair. Elasticfox is nice enough to do this for you. Simply click on the “Keypairs” tab and generate a new keypair. Name it something meaningful, not “keypair” – maybe something like “yourname-laptop” or “yourname-work”. You can of course share a keypair across multiple computers, but you’re on your own for that. Elasticfox will prompt you to download the keypair. Obey the machine – download it and save it to your ssh folder  (usually ~/.ssh).

mv david-laptop ~/.ssh
chmod 600 ~/.ssh/david-laptop

Alright! Now we can get to it! Go to the “Images” tab in Elasticfox and pick your favorite linux distro (alestic has a nice Debian 5.0 base image: ami-67fe190e). Click the green power button to launch, make sure it has your keypair selected, and select “ok”. A few seconds later, the instance should appear as “available” in your list of instances under the “Instances” tab. Once it is ready, double-click on it and copy the public dns name, then open up a terminal and ssh into it (using our keypair of course)

ssh root@ec2-123-456-77-22.compute-1.amazonaws.com -i ~/.ssh/david-laptop

Replacing my keypair name with yours, of course. That’s it, you should be connected.

Sorry about that. I didn’t mean for this to turn into a tutorial. Stupid blog. Back to my story.

After I got the AWS stuff set up, I spawned 4 base debian 5.0 instances and set them up with MPI and installed the multi processor version of MrBayes. After a little tinkering with MPICH settings (I will use OpenMPI next time), I got it running: 4 processors spread across 4 instances. So hot. I think I’ll be changing my thesis to a project where I can use AWS. It’s just so sexy.

Oh, and it cost me $0.12