jamese wrote:
Is a 1024 plan double a 512 plan in every way (i.e. CPU)?
Not exactly. Disk and memory are, since they are fully allocated to a Linode, but CPU is different. The 1024 should be close to double in terms of minimum guaranteed CPU, but a Linode is rarely getting that little CPU resource, so odds are it won't be as much a difference. Any Linode plan has access to up to 4 cores worth of CPU in terms of burst capabilities, but the higher plans have fewer guests sharing the physical CPUs so the minimum if everyone was CPU bound is higher, and maybe odds of being able to burst to the full 4 cores is slightly higher, but it's definitely not like 2x the probability or that the higher plan can burst to 2x the CPU.
Oh, and while disk space is fully allocated, the aggregate I/O to the disks is the same among the other plans, so I/O rates are shared similar to CPU and, as in that case, should have a higher minimum baseline on the higher plans due to fewer nodes, but also as with CPU, both plans would have the same peak burst capability.
There's debate as to whether you're actually better off on the lower plans since maybe the odds favor those nodes being users who don't need much resource, whereas a larger plan is likely being taken by people who already know they're going to be needing the capability. I don't know that there's any way to know for sure.
Quote:
Would it basically come down to the question of maintenance when deciding whether to put multiple sites on a single large linode vs. giving sites their own linode?
I suppose this might all count under an umbrella of "maintenance" (a single machine is certainly a bit simpler to operate), but I think of it as more a trade-off in terms of better redundancy/reliability vs. increased cost/effort to synchronize. I think there have been some other scaling threads here discussing some of the tradeoffs. For example, it's going to be easier to scale a common database by growing vertically (larger plan) than trying to synchronize or distribute it among multiple nodes. But other applications (like static web serving) would scale horizontally (more machines) easily and achieve better redundancy/reliability that way.
For myself, let's say I started with a single Linode with web server, some other applications, and the database. First growth step might be to offload the database to a separate Linode using the private network to communicate. Then that dedicated database Linode would grow vertically as long as feasible. If the front end applications needed room, I'd probably partition them among multiple Linodes, perhaps adding a load balancer node in front.
Probably until I had at least two machines in some way, I'd prefer that as a first step rather than a larger plan since it buys me some redundancy, even if only to use each machine as a backup for the other in an emergency.
There's lots of ways to slice the pie though.
-- David