We are looking at NodeBalancer as a HA solution but are curious about the performance aspect. Does anyone have any observations on how much lag this will induce on a web application? I read an initial report on the service that said as high as 2s ? Now that can't be right....

Thanks,
Scott

Nanoseconds? A few microseconds at most. It's inconsequential and the benefits of routing request to healthy backend nodes, distributing load, being able to do rolling upgrades, etc, etc, etc outweigh.

-Chris

The time required to do backend processing on your Linodes once the request actually reaches them is likely orders of magnitude greater than the delay required to route the request after reaching a nodebalancer.

I beg your pardon, I'm not at all saying that routing through a NodeBalancer doesn't have the benefit of availability. In fact, we're already sold on it because the configuration seems easy. I just wanted to get a handle on the experience from a user's perspective. Has anyone done a "before-and-after" with a browser's waterfall graph? Or hit Webpagetest.org just for shits?

Edit: Some of our static assets come in less than 100ms with Nginx serving them up. Neverminding the app for a minute, what will happen to them?

Regards,
Scott