I was trying to look up the VPS/host ratio for some of the plans, and found that the entry that used to break down how many Linodes were allocated per host per type (e.g., 40 Linode 360s per host) seems to have been removed from the FAQ recently. Or did it just move somewhere else that I haven't located yet?

If removed, are there any implications that there may be changes to such allocations underway, or was it perhaps just deemed a negative when marketing against competitors who don't publish equivalent details?

For what it's worth, I found its being documented way up front very refreshing when researching VPS providers.

-- David

That is odd. Google's Cache from 6/19 still has the old version with the breakdown, so this was a pretty recent change. I wonder if Linode decided that modern servers are capable of running more then 40 360s?

And that removal is the only change to the FAQ.

I'm curious as to why and what might be going on? I certainly hope it's not more hosts per server.

Subscribing to this thread. I'm very interested in this, too.

Google Cache: http://google.com/search?q=cache:http:/ ... are-a-host

When that expires: http://web.archive.org/web/200802130036 ... are-a-host

I did some math on this a while back and Linode's VPS/host ratios were making increasingly less sense on the latest hardware. I'd guess they started getting new hardware in and did the same math. This would only change the ratios on the new hosts, though, which would make the numbers in the FAQ accurate for one set of servers but not the other until the old ones were entirely displaced (which, given Linode's size, I expect would take quite a while).

This is really interesting. I'm subscribing as well. Let's hope we'll have a formal announcement soon.

Assuming that there are changes being made, which I still don't know the edit need imply, I don't necessarily have an issue, but would hope that any new ratios were published somewhere, even if it needs more explanation than the prior one line FAQ. I believe transparency in this regard helps more than hinders.

With respect to ratios - by "less sense" I presume you mean from a business sense. I would assume that while hardware performance has increased, I don't know that net capital costs should have risen significantly (well, switching to RAID 10 would add to storage cost), so at least in theory Linode isn't taking a hit by just passing the performance gains on to us, although it's true that a higher ratio might yield more revenue while maintaining prior performance. I suppose there's an argument to be made about consistent performance among older and newer customers.

I do get a little concerned with respect to CPU and disk, as shared resources. On a recent host (May), my Linode 360 could, I think, in a worst case throttle CPU down to a bit less than .5GHz (assuming 20 Linodes assigned to each quad-core 2.5GHz xeon, minus whatever dom0 overhead on the host). Now, average case should be, and has so far for me, been much better. And I've been able to burst to a full 400% so best case is way better. But I still have some concerns of average sliding down towards worst over time. I still don't know how loaded my particular host is, for example.

I assume (hope) that any changes would be guided by stats from active hosts in terms of average saturation over time.

I'm slightly more concerned about disk contention, as it does seem to be the most constrained resource for Linodes (and other VPS solutions). Though my average performance is very good so far, I have encountered some slower periods. Perhaps the shift to RAID 10 will further improve things on that front.

Of course, this could all just be putting the cart before the horse. Heck, I'm even willing to be a glass half full guy and think that perhaps the ratios are being decreased, to provide more headroom for future increases in plan limits

-- David

There is no reason for alarm - we'd never implement any change that we thought would make your service worse. We have a track record of doing just the opposite: increasing resources and performance while increasing value at the same time. As our fleet of machines grows and as we adjust our hardware specs to take advantage of newer processors and technologies, it's inevitable that we'll have a widening disparity of performance between old and new boxes. We are now adjusting the number of Linodes on a per host-hardware-specification basis. Some Linode 360 hosts will have more than 40 Linodes on them and some 360 hosts will have less. If you factor in newer hosts having much faster bus transfer rates, more efficient cores and more of them, faster clock speeds across those cores, faster disks, faster RAID cards, faster RAID setups (RAID10), we've effectively lowered the resource contention ratios. Performance will only continue to increase at Linode as time goes on, regardless of how we allocate things on the back end.

So, nothing radical is changing, other than our move to some really sweet new hardware along with RAID10. Pay more == get less contention, just like it's always been. We'll slowly be migrating older machines from RAID1 to RAID10. The cycle of life for hosts will continue to roll on at Linode as it has for the past 6 years. Old machines will grow old, receive some TLC upgrades until it's their time to go, and then people will be migrated off onto newer machines, rinse, repeat.

I've updated the FAQ entry.

-Chris

Will the new hardware change the price?

Having to spec this sort of stuff out for work, I seriously doubt it. You can buy/build a server/workstation that's easily 2-4 times more powerful (and in some cases ridiculously more powerful) for 1/2 the price it would have cost 3-5 years ago.

That's why Linode (and all other hosting providers) are able to continually increase available resources without increasing the price to their customers, while at the same time probably increasing their profits.