Well, our big redesign of Connect, version 4.0, is finally live. Anthony and Rich are knocking out some final issues, but we cut over earlier this afternoon.
Luckily, it’s the weekend. So, our users aren’t using right now.
Unluckily, it’s the weekend, and everyone on the team has plans.
Next week, I’ll fill you in on the new stuff. So, tune in or out, depending on your interest level.
This is by far our biggest release, probably larger than when we switched to the Mix JRuby code line last Summer. Connect is just under two years old. Born in June 2007 as the IdeaFactory, we’ve managed to iteration four times, and the feature set is surprisingly large.
I’m continually impressed with Rich and Anthony’s ability to accomplish so much with so little.
As we realized this week, we’ve set the bar very high, and now that we have a lot of users, we have to move away from the startup mentality a bit and plan more.
For perspective, last June, when we switched to the JRuby code base, Connect had about 36,000 pageviews from 3,000 odd users. We’ve served 40,000 pageviews this week to just over 2,800 users, and Connect was down for about 12 hours while we tested the production migration.
Lately, a typical week would have topped the numbers for all of June 2008 and then some.
All these users have been busily creating data. I don’t have a solid number for the objects in Connect, but if I had to guess, I’d say more than 100,000. Anthony and Rich have completely rearchitected the data model in the new version, so that’s a lot of rows to touch.
Increased usage is both boon and bane for us because it means we have users doing work on Connect, which in turn means we can’t just mess around and bork it up without notifying them.
I’ve never been through a product testing feeling confident that I have a rock solid product. Testing is hard. I deal with it.
We went through about two weeks of beta-testing with the help of some excellent volunteer users and had the code pretty close to where we wanted it yesterday.
The plan was to let everyone know we were taking Connect down first thing today and to expect it to be live on Sunday morning.
I know, bummer, but we have more users overseas that domestically. So, we plan for India’s and Europe’s schedules.
Still, we thought this would minimize the downtime.
We were looking good, until I fired up a VM and decided to take a look in IE7.
Stop me if you’ve heard this one. Connect was unusable in IE7 and 8 due to all the AJAX we’ve introduced. As much as I’d like to skip coding for IE, we can’t. 26% of our users are on IE7, and we can’t abandon them.
So, we had an emergency huddle to decide what to do. Once again, Rich and Anthony rocked it out, and by the end of yesterday, we were ready to go on IE too.
I went to sleep preparing for a morning of testing.
I awoke to find that R&A had gone off script and taken Connect down early.
Then I get a call from a guy in EMEA. Uh oh.
Long story short, next time we need to do better.
Hard to believe that people here actually use Connect for real work, every day.
Seems like we’ve accomplished our mission, but along the way, we’ve become a product team, which is exactly what we started out to avoid. Plus, we’re not only a product team; because Connect is a web app, we’re also an operations outfit too, which is scary. I’ve carried a beeper. Do not want.
It’s not so bad. Planning won’t kill us, and it’ll be a while before we attempt a big release like this one. The operations piece will work itself out too. Frankly, Connect has very few outages (knock on wood). Not bad for a little JRuby app running on souped-up desktop boxes.
The tough part is going to be balancing the innovative stuff we want to build with the demands of our users–the classic product team dilemma.
We’ll get through the growing pains, and in the end, it’s all worth the effort.
We’ve managed to turn the ship a smidge, and I’m happy with our progress.
I have the coolest job at Oracle.
That’s my mantra for next week, when a full-scale, Facebook-style user revolt erupts.