Should I move to Linode from Liquidweb.com?

On a linode, each user gets a fixed amount of guaranteed CPU relative to the plan size they have, so say there's twenty 1024s on a linode you're guaranteed get 1/20th of all the cpu power.

However if other linodes aren't using their full cpu allowance and you need more you can use theirs. When they start to use their allowance you'll be throttled back to your guaranteed amount.

There are no warnings from linode so far as I know about excessive CPU usage since you can't exceed your usage.

If you search the forums there's a load of posts about linode's server specs, fyi you get access to 4 virtual processors this is an example

processor    : 3
vendor_id    : GenuineIntel
cpu family    : 6
model        : 26
model name    : Intel(R) Xeon(R) CPU           L5520  @ 2.27GHz
stepping    : 5
cpu MHz        : 2266.746
cache size    : 8192 KB
physical id    : 3
siblings    : 1
core id        : 0
cpu cores    : 1
apicid        : 3
initial apicid    : 17
fdiv_bug    : no
hlt_bug        : no
f00f_bug    : no
coma_bug    : no
fpu        : yes
fpu_exception    : yes
cpuid level    : 11
wp        : yes
flags        : fpu de tsc msr pae cx8 apic sep cmov pat clflush acpi mmx fxsr sse sse2 ss ht nx constant_tsc nonstop_tsc pni ssse3 sse4_1 sse4_2 popcnt hypervisor
bogomips    : 4537.60
clflush size    : 64
cache_alignment    : 64
address sizes    : 40 bits physical, 48 bits virtual
power management:

(X4)

I tried out liquid web way back when (before the smart servers) and they were rubbish, I had issues with disk IO there and crappy support.

IMHO you should try linode, it won't cost the earth to try it out for a week/month.

@obs:

IMHO you should try linode, it won't cost the earth to try it out for a week/month.

The problem is moving the website from Liquidweb to Linode. If I don't test the server in real conditions, I won't know if it's able to run my website. And I can't be doing this every day. So I must be (pretty) sure about Linode servers.

However, your post is quite illuminating.

I'll wait for other replies (if any), but next billing cycle at Liquidweb is in a week from now, so I should make a decision in the next hours.

Thanks!

@usr01:

MemTotal: 524288 kB
@usr01:

I pay for it $60 per month
I don't think you need exactly Linode 1536, Linode 512 will be enough (same size of RAM) and it will be just $20.

@usr01:

the CPU load at my VPS at Liquidweb.com is ranging from 2.0 to 4.0
You will have 4 cores, and I'm sure it will be enough.

To compare, compilation of PHP ('make' from sources) takes 17% of one core. So it's 4% of total abilities.

At least, you should try it. I have tried out lot of hostings (all kinds of them: shared hosting, vps (OpenVZ), dedicated server, clouds with XEN-VPS) and I can say that Linode is best and amazing and awesome and lot of good words more. And it's not only my opinion, read :)

I'd try starting with a Linode 512 as well. I think you'll find it probably performs much better than what ever your VPS claims they are giving you. However, you could go with a 768 or 1024, get more resources and still save money.

Check out this comparison of some of the competition, Linode simply blows the competition out of the water in terms of performance and IMO service:

http://journal.uggedal.com/vps-performance-comparison/

@obs:

There are no warnings from linode so far as I know about excessive CPU usage since you can't exceed your usage.
While I'd agree that it isn't so much a warning as a notification, you do get notified if your average CPU exceeds the notification threshold set in the manager. By default that's 90% (as a percentage of a single core).

– David

@usr01:

The problem is moving the website from Liquidweb to Linode. If I don't test the server in real conditions, I won't know if it's able to run my website. And I can't be doing this every day. So I must be (pretty) sure about Linode servers.
I agree you can't be absolutely sure, but even some level of testing is going to be better than purely evaluating reviews or postings here, depending on how critical your site is.

If I were you, I'd still consider replicating your current site over to a test Linode, so it is running with a production stack/data, and then just doing the best you can to stress it (via a favorite stress test tool and/or site). It may not guarantee that it'll work under your exact production load, but I think it would go a long way to highlighting any possible kinks or requirements in terms of Linode size, as well as a chance to get some experience with the environment. In addition, the act of getting the test node running will serve as a dry run for migrating your setup, making any final move that much easier.

I would also certainly try to overlap this with your LiquidWeb operation, in case for some reason you opt to remain where you are, and even if that means extending your current plan one more billing cycle. Though I guess that also depends on how critical your site is for you. Hot cuts with no fall-back are nasty and fraught with peril :-)

– David

Note that the "Load average" displayed by the uptime command is NOT related to CPU, so your problem might not be CPU-related at all. Running "vmstat 1 30" will output thirty seconds of per-second resource utilization data.

From the meminfo output you pasted, I'm really worried about:

MemFree: 215168 kB 
Buffers: 0 kB 
Cached: 0 kB 

That tells me that, somehow, you have ~200 MB of memory sitting there doing absolutely nothing, and none of it is being used for disk caching. This is not good.

Here's what one of my busier Linodes has:

MemFree:           15864 kB
Buffers:            3988 kB
Cached:           229848 kB

I'm wondering if they might disable caching so that they can oversubscribe RAM a wee bit…

Magic 8 Ball says "Definitely Yes"

Thanks to everybody. I've finally purchased a Linode 1024, so I can start doing tests. I guess I will install a copy of my website and stress it in order to check if everything is OK. Just as db3l said.

And regarding hoopycat's words:

> I'm wondering if they might disable caching so that they can oversubscribe RAM a wee bit…

All VPS at Liquidweb are fully managed. I didn't touch anything. The settings are the same as in day 1.

Thanks again to everybody!!

It looks to me like his vps is running on OpenVZ, the host takes care of the disk cache.

And did you notice the MHZ for each cpu is only 400 something? probably why you have such high load average.

@hoopycat:

Note that the "Load average" displayed by the uptime command is NOT related to CPU

say what?

it's not a pure measure of CPU, but it's hardly unrelated.

@glg:

It's not a pure measure of CPU, but it's hardly unrelated.
I suppose not completely unrelated, but with Linux it's more loosely related than on other systems, and under some conditions can be pretty well divorced from CPU usage, so it's probably safer not to assume a relationship unless you look at other information.

Especially in a VPS environment, it can be very common to have processes blocked in I/O wait (due to disk contention) and Linux counts all of those towards the run queue shown in uptime. In that case, while the load may still be representative of how quickly work is getting done, it's far less related to CPU.

With a more traditional calculation, each increment in load value of 1 is roughly representative of a full CPU core during the respective time interval. But under Linux and particularly in the Linode environment, you may show higher values and yet have plenty of spare CPU. Without looking at other information that covers the same time period, which can be hard unless you were watching in advance of running uptime, you can't really deduce much about CPU usage just from uptime output.

– David

Not to mention that a load average of 1 on a single-core system is a lot higher than a load average of 1 on a quad-core system…

I never understood why people rely on load averages rather than percentages of utilization when it comes to saying how much load a system is under; it seems to me that the load average only really becomes useful once you hit 100%.