Larger Linodes

Having a virtualization layer between the hardware and your OS turns out to be pretty handy sometimes, especially when you don't have to worry about it. The Linode Manager and LISH are quite nice for provisioning and troubleshooting.

It's also a lot easier to split up your stuff across multiple instances. You could have your database servers on large nodes, serving stuff to application servers on smaller nodes, knowing that everything is on the same grade of hardware and that you can add application servers without waiting for someone to provision hardware for you.

I honestly hate to have to say it, but Linode is not the way to go for larger allocations. Because their pricing structure is strictly linear (a 2 GB instance is double the cost of a 1 GB instance) you can get massively better deals at other excellent providers.

One I would check out for large VPS allocations would be Storm on Demand. They offer 32 GB of RAM, 1.5 TB of disk for $350/month. (Bandwidth is extra and can be pay-as-you-go - still free inbound).

http://www.stormondemand.com

(Sorry, Linode. I love you guys but it is true about the larger plans)

Thanks both. Ghan_04, I'm in the UK and so are all my clients, looks like Storm's datacentres are US only so not an option due to latency. Looks like I may give Linode a go and just be prepared to move back if necessary (unless my current host offers a good deal to stay).

Just in case it's not clear, Linode does scale up to 20GB in size. It should also be noted that, while the costs scale linearly, it's not only the RAM that increases, but also the share of other resources like CPU and IO time. So while the cost might double from a 1GB to 2GB linode, so does your CPU time, for example.

I can comment on the performance, which I'll be honest - I've always been impressed with the consistency of. I started out with a 768 and a 384 (Apache and MySQL), now I have a 1536, two 1024's and a 512. I'll likely be moving up to a 4GB upgrading the smaller two to 2GB, and the 512 to a 1024 in a year if things keep going as they are.

Traffic for my Linodes started at around 10,000 pageviews/day and now is around a million for mixed PHP/MySQL content, sites, and applications. In fact due to a miscommunication between me and my junior designer, a page element was pushed into production with the content hosted on the two small Linodes before I had a chance to upgrade them - moving from 20,000 pageviews/day to around 700,000 pageviews/day. I still don't know how, but they held up just fine until that evening when I upgraded and added the third server.

What I'm getting at is that yes, you'll see consistent performance. I've been a Linode customer for, I believe, about 3.5 years now and that's one thing that has impressed me about them - consistency.

I have 8 8gb nodes in a small hadoop cluster on linode. Most of my issues with it stem from correctly tuning the jvm :-p

I have all 8 machines pegged at 100% (or 400% if you like) cpu for a quite a lot of the time every day, and normally am doing about 4k block/s on the disks. That said, with how batch oriented hadoop is I'm not sure I'd actually notice a little performance blip here or there.