How FlightCaster (YC S09) built multi-platform scalable apps on Heroku (YC W08) | Hacker News Reader

How FlightCaster (YC S09) built multi-platform scalable apps on Heroku (YC W08)

infoq.com

46 pointsjaf12duke14 comments

I find the non-free options no Heroku to be vastly more expensive than hosting options elsewhere such as Slicehost, Dreamhost and Amazon ECS.

They provide a valuable service in terms of taking administration issues out of the picture and letting you just focus on developing your app.

But my feeling is this is not targeted at hackers who are surviving on rice and beans but rather those who are more likely to either have well-paying day jobs that can subsidize their startup hobby on the side or have enough VC capital to not bother with the mundane details of hosting administration and its associated time and financial burdens.

Not that there's anything wrong with that - Heroku has come up with an ingenious business model because it really does satisfy a need in the vanity hosting market where cost is not an issue - but belonging to the rice and beans category myself at the moment, I just find that I'm sadly not their target demographic - as much fun as it is to use Heroku's free option for little toy apps.

Deploying to Heroku really is a delight and I think we will see more services like them start up over the next few years.

If you are looking for a free tool for rails deployment, check out my project http://wiki.github.com/wr0ngway/rubber (The Quick Start section gives a good feel for it) - while the EC2 account isn't free, its not too bad, especially if you aren't running a full time instance. I've also done some abstracting out of the ec2 bits, so theoretically you could write a slicehost/other adapter if you were so inclined.

This is cool. How much tinkering would it take to use this for Sinatra deployments to EC2?

Not much, that was one of my focuses in the current release as we needed to deploy a sinatra app. I think the only thing that missed the release was generating a Rakefile into the project root for non-rails projects. Here's what ours looks like: http://gist.github.com/189256

Ping me on the rubber mailing list if you have any questions - I'd be interested in hearing how this works out for a non-rails app for someone other than myself :)

well-paying day jobs that can subsidize their startup hobby on the side

My well-paying hobby subsidizes the day job at the moment. ;)

Seriously though: if you're charging money for whatever you're selling, servers via any of the above methods are dirt cheap. Its not quite on the same level as "economize on what brand of pens I use", but it would be impossible for me to meaningfully affect my life via switching hosting providers, whereas it is very possible to meaningfully effect my life via doing productive work on the business.

Accordingly, I suggest optimizing for whatever provider or system gets you back to solving problems for your paying customers. (I use Slicehost, for what it is worth.)

Since I know the FlightCaster guys read HN, I'd like to make a suggestion of something I'd love to see.

When I take a flight that's not on Southwest and more than an hour in length, I like to do the first-class / business class upgrades when I check in because they aren't that much more expensive (usually $50 to $100 per leg). It's one of my few guilty pleasures.

Anyways, it would be pretty cool if there was a way for me to predict which flights have a higher likelihood of having those types of flight upgrades available (which is to say flights that are statistically less likely to have a full first or business class).

I have no idea if you guys even have access to that kind of data, but whenever I book travel now I guess as to the likelihood of being able to upgrade on the day.

Great suggestions icey! It's one we've heard from quite a few travelers. Our focus is on giving travelers the tools they need to better manage their travel in real time. Right now, that means delays, but integrating upgrade availability is on the roadmap for the near future...

Stay tuned--and keep the suggestions coming. Always appreciated!

Most airlines have a separate fare class for upgrades like this. For domestic flights on AA, for example, the fare class is "X". So you can use any tool that shows you fare inventory (like ExpertFlyer) to check this.

And of course, your chances of upgrading are greatly improved if you have status on the airline. This is very easy to get. (AA is doing a double EQM promotion right now, which means one transpacific flight gets you Gold status. With this, you check a box when making your reservation, and if there is availability in the next class of service, you are automatically upgraded 24 hours before the flight. Much better than gambling on the rare "LFBU" kiosk upgrade.)

Anyway, statistics are nice, but it's pretty easy to ensure you get the upgrade. I fly about 50k miles a year, and have only missed one segment in the last two years. (And that was my fault; I checked in without enough "upgrade stickers" and was removed from the upgrade list. I got back on about an hour before the flight, and missed the upgrade by one spot. Oh well :) I have also been upgraded for free on a variety of generally non-upgradeable segments this year, too. A very nice undocumented bonus of having status.

I used to be able to do this when I was flying back and forth to London every 8 weeks; but I don't fly enough now for any carrier to really care about my business. I probably only fly about 3,000-4,000 miles a year anymore.

Is anyone else bothered by the fact that InfoQ interviewed a team of really smart guys doing exciting and challenging (and computationally intensive) data analysis on a sparkly new-ish cloud host that's supposedly doing great things for scalability, and wasted so much time talking about HTML5 and mobile web UI? Aren't there a million web designers out there they can interview instead, or am I just missing something?

I know they had a "Part 1" (which my computer is utterly failing to access right now or I'd verify the following), but from what I remember, it was about how they use Clojure, and there's an entire scalability angle to get that I'd rather hear.

I'm more frustrated that the transcriber apparently didn't want to take the time to proof their copy. That interview is hard to read. Many typos, missing words etc.

I only skimmed this one but I didn't see any replies from Bradford Cross, and he's one of only two guys on their team doing the Clojure/Hadoop/etc back end. So maybe they were talking to the front end guys this time?

Sigh... just RTF part 1 before moaning: http://www.infoq.com/articles/flightcaster-clojure-rails

It goes extensively into detail about their use of Clojure. The fact that the 2nd part's title is "Building FlightCaster's Frontends for the Web and Smartphones" could've tipped you off about what's inside...

We also found that Flightstats was reporting an imaginary airline (Oceanic Airlines).

I guess someone at Flightstats is a Lost fan.

Reply on news.ycombinator.com