Archive for September, 2009

Let’s talk about performance

Tuesday, September 15th, 2009

Some users are asking for benchmarks so we thought we’d share what you can expect in terms of performance. Monkey Analytics is running on Amazon EC2 servers – 1.7GB RAM, 1GHz Xeon/Opteron equivalent machines with plenty of disk space. Free users may be sharing such a machine with other free users, but paid subscribers will have effectively dedicated machines.

If you run an Octave session and execute tic / toc commands before and after your scripts, you will observe that computations are fast, as you would expect given these specs.

However, if you measure the time to submit a command and receive results you’ll simple commands can take 4 seconds to complete. We are improving this – on our development servers we return simple commands in under a second, and we’ll keep pushing on this until it’s unnoticable.

In summary, our computational performance is great, but our perceived performance has room to improve.

We should launch a faster interface within the week – so stay tuned!

More fun monkey graphics

Saturday, September 12th, 2009

Everyone loves monkeys, right?  That’s the theory at any rate.  And we think they’re even more lovable when they are acting geeky.  Without further ado – here are some new graphics we created for our User stories page:

You know the kind - two monitors arent enough for the kind of analysis these monkeys do

You know the kind - two monitors aren't enough for the kind of analysis these monkeys do

These monkeys wont stop until theyve pounded learnin into your brain. (Bring bananas, not apples to class)

These monkeys won't stop until they've pounded learnin' into your brain. (Bring bananas, not apples to class)

Minor update – computation error handling

Friday, September 4th, 2009

Not the sexiest of features, but quite important. We weren’t handling errors generated by user computations properly – and this lead to some annoying debugging by you, our users. We are sorry – we know how frustrating that is! You may now rest assured that if your scripts or commands are generating errors you will now see them.

Do you really want to setup your own EC2 server (version R)

Tuesday, September 1st, 2009

We asked this question earlier in regards to Matlab in the Cloud; now that we have launched support for R, we’re asking again in the context of R in the cloud. – third result is a neat post detailing running R on EC2 servers. Neat, but as we stated before, we believe that many of you will prefer to get down to computational brass tacks and not futz about with system administration.

Update: Robert Grossman’s original article offers more detail on running R on EC2, and has some interesting discussion as well.

Agree? Disagree? Let us know in the comments.