CPU allotment.

It's undefined and largely unimportant; for the vast majority of people, they're not CPU limited, because contention is low (cue zunzun with his "My atypical CPU-intensive website is the exception!" post). The host machines are all, at a minimum, dual quad-core Xeon processors, for eight cores per machine. Unless it's changed, the allocation is generally 20GB of RAM per host machine divided among linodes (so, 40x512 on a machine, or 20x1024, or 10x2048, and so on). So your minimum guaranteed share on a 512 would be 20% of a single Xeon core.

What generation of Xeon processor is used is, of course, variable. Older Linode hosts would have older generation processors, and for all we know, some machines may have Xeon processors with more than four cores. Linode is notoriously secretive about their back-end infrastructure, because it's supposed to be completely transparent to end-users; all Caker has ever said about CPU specs is that it's "enough".

If I had to take a wild guess, I'd say that a Linode host has the equivalent of a minimum of 16 EC2 compute units per host. But that's a wild guess. And EC2 doesn't do burst CPU like Linode does.

No offense but telling someone that CPU is largely unimportant when they are specifically asking about cpu speeds and when you have no idea how cpu intensive an application I intend running on the node is a bit silly.

Thanks for the answer anyway. It was somewhat helpful.

If there's another VM on the same physical host I'm using that's running a CPU-intensive app, I haven't noticed :-)

Most CMS's use MySQL, and that's by no means light – it's not as heavy as some other DB software, but it's not too light either. Also, there's a good chance the there are a lot of 'nodes on the same physical host as mine that use Apache, which is quite resource-intensive under high loads, and I haven't noticed any changes in performance. That's why Guspaz said it's unimportant -- if we can have several VM's on the same physical machine that are running (mostly) heavy apps without noticing a performance drop, you shouldn't have to worry about it.

@Eamonn:

No offense but telling someone that CPU is largely unimportant when they are specifically asking about cpu speeds and when you have no idea how cpu intensive an application I intend running on the node is a bit silly.

Thanks for the answer anyway. It was somewhat helpful.

It's unimportant because in the vast majority of cases we see here, the CPU is almost never a bottleneck. Zunzun is, I think, the only user I've ever seen complain about not having enough CPU power. Usually it's disk IO, transfer cap, storage space, etc.

You may think it's silly to say that, but experience has shown that it's true; most users come here used to shared hosting or low-end VPS hosts where they're CPU limited, and they're concerned about CPU resources at Linode. Once they get on a real linode and poke around, they realize that they're not going to be anywhere close to bottlenecking on the CPU.

@Guspaz:

Zunzun is, I think, the only user I've ever seen complain about not having enough CPU power.

Sniff sniff boo hoo waaaah more CPU power whine whine complain complain.

James

All of mine that I checked fall into two main categories:

processor    : 0
vendor_id    : GenuineIntel
cpu family    : 6
model        : 23
model name    : Intel(R) Xeon(R) CPU           L5420  @ 2.50GHz
stepping    : 6
cpu MHz        : 2500.090
cache size    : 6144 KB
fdiv_bug    : no
hlt_bug        : no
f00f_bug    : no
coma_bug    : no
fpu        : yes
fpu_exception    : yes
cpuid level    : 10
wp        : yes
flags        : fpu de tsc msr pae cx8 cmov pat clflush mmx fxsr sse sse2 ss ht nx constant_tsc pni ssse3 sse4_1 hypervisor
bogomips    : 5000.18
clflush size    : 64
cache_alignment    : 64
address sizes    : 38 bits physical, 48 bits virtual
power management:

processor    : 0
vendor_id    : GenuineIntel
cpu family    : 6
model        : 26
model name    : Intel(R) Xeon(R) CPU           L5520  @ 2.27GHz
stepping    : 5
cpu MHz        : 2260.998
cache size    : 8192 KB
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 sep cmov pat clflush mmx fxsr sse sse2 ss ht nx constant_tsc nonstop_tsc pni ssse3 sse4_1 sse4_2 popcnt hypervisor
bogomips    : 4521.99
clflush size    : 64
cache_alignment    : 64
address sizes    : 40 bits physical, 48 bits virtual
power management:

All times four, of course.

@Guspaz:

It's unimportant because in the vast majority of cases we see here, the CPU is almost never a bottleneck. Zunzun is, I think, the only user I've ever seen complain about not having enough CPU power. Usually it's disk IO, transfer cap, storage space, etc.

You may think it's silly to say that, but experience has shown that it's true; most users come here used to shared hosting or low-end VPS hosts where they're CPU limited, and they're concerned about CPU resources at Linode. Once they get on a real linode and poke around, they realize that they're not going to be anywhere close to bottlenecking on the CPU.

Jeez man you don't learn.

I am not coming here from shared hosting or low end vps hosts. I maintain 7 dedicated servers at the planet and manage over 20 vps' for design clients.

I am the one who wrote, profiled and tested to death my application. I know how much disk IO it needs. I know how much transfer it will need. I know exactly how cpu intensive it is. You don't.

I am looking for a vps to host a CPU intensive app so if you have nothing further to add apart from snide side swipes at another user then please refrain from posting in this thread. Thank you.

Hoopycat that is perfect. Thank you very much.

I would say the L5420 proc is somewhat popular. All mine have been the same. :D

processor : 0

vendor_id : GenuineIntel

cpu family : 6

model : 23

model name : Intel(R) Xeon(R) CPU L5420 @ 2.50GHz

stepping : 6

cpu MHz : 2500.088

cache size : 6144 KB

But in Guspaz defense, I will say that most of the posters posing that question about CPU speeds truly don't need to worry about it. You, having actually profiled your app, are in the minority. I don't believe Guspaz was being flippant, only speaking from a great deal of experience on this board.

He wasn't trying to be snide or to side swipe you. Without a detailed explanation, we know not how CPU intensive your app will be, so all we can do is speak from our own experience.

@Piki:

He wasn't trying to be snide or to side swipe you. Without a detailed explanation, we know not how CPU intensive your app will be, so all we can do is speak from our own experience.

Exactly this. Being rude those that offer information isn't a great way to join the community. I've stated the general case, and explained how this relates to why Linode won't give you detailed hardware stats (aside from what we can gather), I'm not sure why you're so upset to get this information.

Older linodes likely have slower CPUs. I believe my previous linode, which was a larger size (I started at 512, it became 768, and the old 360 became 512, so I resized from 768 to 512) was an older Xeon at 1.7GHz. Linode probably cycles out or upgrades older hardware. If you do hit an older box, by whatever coincidence, you can always ask Linode to migrate you to a different host. Alternatively, since Linodes are pro-rated by the day, you could sign up for multiple linodes of the same size (each would be on a different host), and keep whichever has the best CPU.

If it wasn't clear from earlier posts, Linode only puts linodes of one type on a given host; all linodes on a host will be 512s, or all will be 768s, and so on. It's a mystery to the customer base how Linode manages to offer periodic upgrades in hardware (RAM, disk, etc) without ever shutting down the hosts or migrating customers. Perhaps there's some transparent live migration going on, or perhaps they have hot-swappable (and upgradable) hardware, although that'd be odd.

@Piki:

He wasn't trying to be snide or to side swipe you. Without a detailed explanation, we know not how CPU intensive your app will be, so all we can do is speak from our own experience.

I was not referring to snipes at me. He has been polite and helpful toward me. I was referring to the two digs at Zunzun.

@Guspaz I did not intend to be rude to you and I don't believe I was. I have been reading this forum a bit this last two days and I see you are a valuable and knowledgeable member of this community and post a lot of helpful answers.

But in this case I did need to know cpu power. Linode have this in their terms of service

Misuse of System Resources: Intentional misuse of system resources, including but not limited to employing programs that consume excessive network capacity, CPU cycles, or disk IO.

So I wanted to know what size of linode to go for so as not to degrade the performance of other nodes on the same box.

My application is a personal hobby thats why it is not going on any work servers. It is a browser based wargame that has at least 50 concurrent users offpeak and 150-200 users during peak hours at weekends. It does a lot of computational work so you can see how it is cpu intensive.

Thanks to the answers here I know what size linode best suits me for this particular app.

@Eamonn:

I was not referring to snipes at me. He has been polite and helpful toward me. I was referring to the two digs at Zunzun.

Those are not digs at zunzun. He really is the only member I've ever seen with sufficiently high CPU requirements for it to be an issue, and he really does pop into threads discussing it to list himself as the exception. I'm surprised he hasn't here yet, especially since he really does have experience with using a lot of CPU power on a linode.

@Eamonn:

But in this case I did need to know cpu power. Linode have this in their terms of service

Misuse of System Resources: Intentional misuse of system resources, including but not limited to employing programs that consume excessive network capacity, CPU cycles, or disk IO.

That's more of a CYA kind of thing; each host box having eight cores (at least, that's all we know about, they could be quad CPU boxes, Linode won't tell anymore), and each linode having only four virtual threads, no one linode can completely hose a host. If it's a problem, Linode would let you know, and resizing a linode is very painless, it takes only as long as the time it takes for the linode manager to copy your disk image to a new host (it's all automated).

Network capacity tends not to be a problem either. There's a default outbound cap of 50 Mbps upstream, which they'll raise if you can demonstrate you need more. If you're using a ton of network capacity, you're very likely paying extra for it, so they're fine with that; if you are sending out 100Mbps constantly all month long, but you're paying the ~$3k in extra bandwidth each month, they're not going to complain.

Disk IO is the only thing that is a concern, really; as far as we know, linode hosts have four 15K RPM SAS drives in RAID 10, so there's a lot of IOPS to go around, but they're not SSDs, so it's impossible to truly schedule and manage IO like you can CPU usage. The rule of thumb that Linode has given in the past is, if you're averaging under five digits (as in, under 10k for IO in the linode manager), you're fine. If you're averaging in the five digits, you should keep an eye on it.

@Eamonn:

So I wanted to know what size of linode to go for so as not to degrade the performance of other nodes on the same box.

Don't worry about it, as I said, you can't degrade the performance of other nodes on the box with CPU usage (and they'll let you know if you need to upgrade later). Start with what you think you'd most like to pay (or size it based on RAM), and if it's not fast enough, bump yourself up to a larger linode (takes just a few minutes, entirely automated) to get a bigger guaranteed share of the pie.

@Eamonn:

My application is a personal hobby thats why it is not going on any work servers. It is a browser based wargame that has at least 50 concurrent users offpeak and 150-200 users during peak hours at weekends. It does a lot of computational work so you can see how it is cpu intensive.

Thanks to the answers here I know what size linode best suits me for this particular app.

Don't be afraid to experiment, since changing the linode size is painless; a lot of people overestimate what kind of Linode they need (or just don't know how to tweak their setup properly, and I'm not saying that this applies to you), and end up paying for more than they need. I mean, Apache's default config under moderate load can easily consume 8GB or more of RAM, but you can probably handle that same amount of load on a 512MB linode with some reconfiguring (since Apache's default settings are completely insane and irresponsible). So users who just leave it as-is end up paying for a ton of RAM they don't really need. I'm ranting at Apache, now, not you :P

@Guspaz:

Those are not digs at zunzun. He really is the only member I've ever seen with sufficiently high CPU requirements for it to be an issue, and he really does pop into threads discussing it to list himself as the exception. I'm surprised he hasn't here yet, especially since he really does have experience with using a lot of CPU power on a linode.

but he did…

@glg:

but he did…

Wups, he did. I missed it. His avatar blends in, doncha know.

@Guspaz:

His avatar blends in, doncha know.

Guspaz:

~~http://forum.linode.com/images/avatars/ … 74f863.png">http://forum.linode.com/images/avatars/10597044184b1542f74f863.png" />

Zunzun:

~~http://forum.linode.com/images/avatars/ … 45bfc0.jpg">http://forum.linode.com/images/avatars/14729035464d026da45bfc0.jpg" />

Ha, I only just noticed.~~~~

Too many people from the same family in that army… personally, I prefer to have a bullet proof helmet over a metal one :-P

Where's the third avatar?!!

ObUsefulContent: Again, depending on the application, RAM is probably going to be the ultimate deciding factor for plan size. There is not a strong relationship between $/mo and steal time like there is between, say, $/mo and storage capacity. However!, if your application allocates more memory than you have available, there's a good chance you won't be running into CPU bottlenecks.

So I guess I'll hijack this old thread and ask how does a person get more IO bandwidth.. or is there plenty of that too? :)

Divide the I/O across multiple Linodes, or fund development into faster physical disks and/or larger solid-state disks. (The first is easier to realize.)

zing ! :)