MeshU 2008

I am very excited to say that in Toronto we have an incredible community of professionals in the social networking + digital media space; it's not quite New York or Silicon Valley - but we're flourishing up here. MeshU 2008, precursor to the annual Mesh conference held at Mars, was a great showcase of this talent, and a clear demonstration that we have the critical mass to put on a world class event in this space. MeshU was a one-day event for social media professionals, with management, design, and development tracks. Topics ranged from cloud computing, to managing high performance teams, to web monitoring best practices, to how to design for love.

I split my time between the management tracks and the development tracks. I've been spending a lot of time recently looking into cloud computing as a delivery vehicle for the backend services we're building as part of the Fadow platform. Canada has real leaders in this space with folks like Avi Bryant and Reuven Cohen, who both gave excellent talks in this area. Avi in particular, articulating the pros and cons between relational and non-relational data storage -- including some real life examples of when you can (and should) look beyond the traditional providers. I've spent a large part of my career working on m-Commerce and e-Commerce solutions, and the prevailing theme is always to minimize database I/O; you can easily scale stateless web servers, but when those requests need to go back to shared storage then scaling becomes much more difficult. Cloud computing platforms like Amazon's S3 and Google's BigTable give you an alternative to Oracle/DB2/MySQL, allowing for a much more scalable solution, but with drawbacks in terms of data consistency and feature sets that you really need to be cognizant of. Avi also gave us a cursory overview of Maglev, a distributed cache for Ruby (think of it like a distributed map + shared dynamic methods + transactions). Really cool stuff, for the kinds of applications that many of us are creating these days.

One of the great things about these keynote sessions was the bits of trivia that you pick up: Facebook runs 1,800 MySQL databases in its datacenter. Amazon and Microsoft are both deploying 10,000 servers a month into their cloud computing initiatives (yes, that's right, MS has a cloud computing initiative -SSDS - that I had not heard of). The pace of growth in this space is not surprising; as someone who is charged with designing a massive, highly scalable, fault-tolerant, application platform, cloud computing offers a very compelling value proposition. It means I need to think about my design in a slightly different way, but the benefit is that I can provision as much or as little computing resource as I need. If I need 300 (or 3,000) servers to run a simulation, I can get them. For an hour if I need to. Under the old computing paradigm this would have been unthinkable.

For me, the most valuable lessons of the day came from the management track session on "watching websites" by Bitcurrent's Alistair Croll. Alistair has made a career out of figuring out how to monitor websites: what people do; could they do what they tried to do; why did they do it; and how did they do it. The slides from his session are posted, and I will post his great list of monitoring resources separately. The room was packed with folks who design, build, and run websites for a living, but there were words of wisdom for all of us. Some key takeaways: only 30% of the audience use synthetic tests to ensure that their site is up, the rest presumably wait for customers to complain. Most web analytics platforms use client-side javascript, which is loaded last by the browser - meaning that although you are tracking your novice users, the expert users have probably clicked through before the tracking can be fired (further demonstration that web analytics data needs to be interpreted -- not accepted as fact). And of course, I can never complain when a speaker incorporates South Park into their presentation; in this case, Alistair relating the typical web business to underpants gnomes. For those who missed that episode, the gnomes' business strategy is: Phase 1: collect underpants, Phase 2: ???, Phase 3: Profits. You ultimately need to know what's going on with your customers day to day, and continuously ensure that you are delivering the online offering that matches their needs. There is a plethora of tools out there to help you with that, and the onus is on all of us to use them effectively.