Ask HN: What did you learn in building your scalable web application? | Hacker News Reader

Ask HN: What did you learn in building your scalable web application?

6 pointssravfeyn5 comments

I have built couple of client-side API mash-ups. I have only worked as an intern on a big (in terms of users) web application. I am planning to build a Django application on AWS for a large user base. Though I can build a quick prototype on Django, I just have no experience to learn from.

What are the common 'gotchas' in building a scalable application in Django on AWS? I intend to learn from industry experiences. Please tell your personal experience and give some pointers to blog posts or books.

View on HN

Unless you're talking about a million concurrent users or something, Django actually scales pretty well for medium concurrency.

Here's some basic advice though (1) Cache stuff. Django has a built-in cache framework, but sometimes I even cache stuff in the process (It's not best practice at all, mind you, and every pro will tell you to stay away from that)

(2) Use the ORM properly! For example, never do something like len(MyModel.objects.filter(arg='blah')) to do things like count entries or see if entries exist. Instead use .count() or .exists()

(3) Unless it's something complicated, I would just use something like Heroku. Easy to scale resources up and down as well as deploy. I typically use AWS S3 as my static host and Heroku as my Django host.

Thanks. Can you please explain 3rd point. How is EC2 complicated than Heroku?

Deployment with Heroku: Write app. 'git push heroku master'. Scaling with Heroku: Drag the scroll bar in the admin panel.

Deployment on EC2: Write app. Provision EC2 instance. Ensure instance has all the software you need (Framework, nginx, etc), and install it if needed. Pull code from repository. Start all necessary processes. Scaling with EC2: Repeat deployment steps (but this should be easier because you know what needs to be done with the image you've selected).

Heroku is easier and removes the headaches of running your own server. If you like running your own server - and know how to secure and optimize it - then EC2 (or linode, rackspace, etc) is a good option.

As a counter point - Try to avoid caching until all other options are exhausted. They tend to complicate matters and pave over the real problems, when they could some times have been solved more elegantly.

new relic (https://newrelic.com/) + load testing with funkload has been helpful for me so far, but we shall see...

Reply on news.ycombinator.com