I am currently running ubuntu 12.04 32 bit on what was once called (I think) a Linode 1.5. I have 1.5 GB of ram, 8 cores, 3K GB of transfer, and 75 GB of disk space for $30/mo.

The server hosts multiple web applications, mostly wordpress but also some django, with the DB's on the same box, and a low-volume mail server. More RAM would be good for memchached but having all those cores seems like a good thing for the nginx and php-fpm processes. If I were to upgrade, I would move up to a Linode 4, 64-bit and Ubuntu 14.04.

My question is whether the overall performance is going to be better or worse in the event of a lot of traffic hitting one or more of the web apps simultaneously. Will the speedier CPUs, SSD's, and the increase in RAM more than make up for the 2x cores on the older hardware?

And does anyone know if there will be a forced upgrade / retirement of old hardware sometime soon anyway?

Thanks for any input.

Linode now bills by the hour, so spin up a new box, load your apps & data, and give it a spin around the block.

If it's better, then upgrade, if not, shut it down and keep your old setup.

Either way you'll know your answer FOR SURE for a couple of bucks.

Having people guess which is better, without knowing all the details would be about as accurate as asking a Magic 8 Ball.

_________________
Either provide enough details for people to help, or sit back and listen to the crickets chirp.
Security thru obscurity is a myth - and really really annoying.

Thanks @vonskippy. You are right of course. But it's a non-trivial thing to spin up a new box with a new architecture (64 vs 32 bits), a new OS (14.04 vs 12.04) and then load all of my stuff (each of which is a pretty customized environment) and get it all working. I guess I am lazy/busy enough that I was hoping that somebody who might have relatively similar use cases as mine and has already upgraded might have some experience they could relate.

I'd probably try to avoid that much complication, at least for basic testing, or at least at first, since you're just trying to evaluate if the platform will be a net plus for your use case. The closer to apples to apples you are probably the better. For example, 14.04 could cause changes itself, regardless of host, so you might want to avoid that change as a variable up front. For example:

• Clone your current Linode (use a backup snapshot if you have that service, or briefly bring your current Linode down and clone)
• Switch to a 64-bit kernel (ideally one close to your current kernel though some might suggest always using the latest). You can also skip this step if you wait until 32-bit upgrades become available.
• Use Finnix to tweak clone for new host (hostname primarily, but any other production-specific configuration you may need to tweak)

Then boot and stress your application to see how things go. Best would be if you can run similar requests/tests on production and testing together, but pretty much any usage you might use for which you have a feel how your current system performs should be fine. Once you have an idea of how this performs you can do upgrades to the test box as needed (you could do a 14.04 upgrade while remaining 32-bit user space), and eventually decide if doing a full 64-bit new installation is worth pursuing. In that last step - which I agree can be complex depending on your current setup and ease of restoring - you can plan on switching to your test Linode as your new production box once it's done and ready.

I took this approach with a pair of Linodes (app and db servers) I have, and after setting them up for testing, started mirroring production requests to the boxes to let them run in parallel with the main servers. The results have shown the new hosts to perform significantly better than the old, particularly with respect to worst case response processing - which I attribute to my existing worst case behavior being largely bound by I/O delays.

For what it's worth as a general comment, if you're not currently using more than 400% CPU on a regular basis on your current Linode, I wouldn't expect the drop in virtual cores to be a material issue. Of course it really depends on the overall mix of resources your load needs.

-- David