How to load balance larger sites with linnodes?

What are the best ways to load balance larger sites with linodes?

In the past I've used a pair of BIG IP F5's with one as a hot failover partner to balance 4 http and 2 mysql servers. I am now looking to expand capacity further and want to consider linodes in place of growing my own setup.

Is the only way to load balance with linodes to have nodes (light-weight?) acting as linux-director load balancers then http and mysql nodes behind them as needed? What are other linode users doing for this type of scaling? All I can find mentioned in the forums is about separate production vs development servers and separate servers for different services or content type.

Will linodes physically host hardware like load balancers, vpn/firewall pix's, etc?

5 Replies

I have been using mod_proxy to randomly load balance between 2 linodes, but this ofc it not real load-sharing.

Linode's offerings are limited to virtual machines running Linux right now; they don't colo customer hardware.

You'd have to either use one or more front-end Linodes running whatever software load-balancing solution is appropriate for your case, or go for DNS round-robin.

However, if your site is actually big enough right now to justify redundant hardware load balancers and six physical servers, I'd hesitate to recommend using Linodes without knowing more about your specific application. Remember, these are basically slices of a physical host; you're saying you already use six full physical boxes.

You can do software-based load balancing via nginx, as per this example:

http://wiki.codemongers.com/NginxLoadBalanceExample

modproxybalancer for apache is rubbish, I would say don't use it. nginx and it's load balancing can take many more requests per second, and if you don't want to use that I would suggest haproxy instead, both of these can be used without round robin proxying which is crucial to performance under load.

in response to nknight:

I'm not sure that linodes will be the best in the end for standard content delivery, as you are right I am using multiple full machines already.

I'm more interested in using them primairily for failover purposes.

I'd also like to try using linodes over a vpn to add additional capacity during unusual peak times, etc.

Also, as my 'full' machines are only dell 1750s, 2x 2.3Ghz, 2GB ram, and scsi with raid on the dbs only, I am really not sure that their linode node slices might not compare better when you eliminate hsoting charges, hardware upgrades and fixes, etc.

Reply

Please enter an answer
Tips:

You can mention users to notify them: @username

You can use Markdown to format your question. For more examples see the Markdown Cheatsheet.

> I’m a blockquote.

I’m a blockquote.

[I'm a link] (https://www.google.com)

I'm a link

**I am bold** I am bold

*I am italicized* I am italicized

Community Code of Conduct