Centralization of production best practices

[dougwilson]

Moving issue from expressjs/express#2767

[DCKT]

It should be great to talk about

• compression
• caching
• security on production ?

[crandmck]

Great, thanks. I'll put this in the backlog.

@dougwilson @hacksparrow Can you suggest any other topics that you think should be addressed?

[dougwilson]

Exception handling (expressjs/express#2769)

[crandmck]

From [https://github.com/strongloop/expressjs.com/wiki/Near-term-roadmap](recent discussion), include:

• Known configuration problems that can lead to security issues.
• Give tips on how to configure things properly

[crandmck]

From a discussion with @ijroth and @chandadharap, @ritch said:

"In environment x, how do you scale an express app?"

X will most likely not be PaaS and other heavily managed things. It will most likely be a managed or unmanaged VM (eg. cloud or your own non-cloud / hardware). Services like Amazon, Azure, etc provide their own load balancing solution complete with https (and tls or ssl). I would recommend using those if you are on one of their clouds for anything small.

Some use cases, such as banks, require https at every node. This usually means that you need to run something similar to a load balancer within each VM or other UoS. For the first iteration we can probably skip mentioning this IMO.

At some scale it becomes economical to run or roll your own load balancer. This is where writing your own balancer in node/go/c++ might make sense, or using something like nginx. This approach also make sense if you are not in a cloud with a load balancing service at all, where you don't have the options above. Also in those environments (usually legacy, hardware / non-clouds) you might want to look at a hardware load balancer like f5 (or you might already have them).

You also need something that manages your node processes and plays nicely with load balancers. You should run 1 node process per core and usually 1 pm per app per VM. You should set your process manager up to start applications when the VM starts / restarts.

[crandmck]

Added performance best practices in #503