drmike
100% Tier-1 Gogent
What is that a quad CPU E5-2620?
6 cores on that CPU.. But i see 24 cores on your output...
Reserving Disk Space you Purchased from the Oversold VPS providers
- Thread starter drmike
- Start date
What is that a quad CPU E5-2620?
6 cores on that CPU.. But i see 24 cores on your output...
@jcaleb lol, I magine them swapping out 'them burned out SSDs like Tony Stark was swapping out his smoking Polonium powered Arc Reactor from his chest in Iron Man 2.
Hyperthreading. 12 threads per processor.
jarland
The ocean is digital
Dual 6 core with HT. I love that CPU.
Yes, it's HyperThreading so it's 12 cores. Linux shows 24 because that's what HT does, lies to the OS that you have more cores than you do basically. Even Intel said that HT by itself takes a minuscule area of the CPU die. On the first P4 that it was released on it took only 15%... compound how much transistor size haz decreased since then and it may take ~1% to 2% of the die area per core. Just sayin'.... B)
jcaleb
New Member
that will not be economical
@jcaleb I know, it's all assumption and gossip and who really cares. The bottom line is that if you care about what you're hosting and you rely on it you go with something more reliable. Heck, even CVPS offers Xen at reasonable prices, so why not just go with that instead?
jcaleb
New Member
are their servers unused CC servers?
@jcaleb that I don't know, however I think that lots of people make lots of assumptions about them and I don't understand why. Just vote with your wallet
- because a product doesn't sell the provider will discard or change it.
Last edited by a moderator:
peterw
New Member
Hard but plausible rules. How many "second accidents" do you have?
Meh, @Francisco at BuyVM knows his stuff when it comes to virtualization and he does things well. It's the providers that are trying to exploit OpenVZ in order to make a quick buck that screw the pooch and give OpenVZ a bad reputation. A rotten apple can spoil the entire barel... that's how the saying goes.
drmike
100% Tier-1 Gogent
I *wish* they'd stop counting hyperthreading as cores. Tired of looking at CPU spec sheet to decode things -daily-.
They need some clear processor labeling already: Hiss!
Nice machine @jarland. I need to Start running some bigger RAM servers for projects.
earl
Active Member
I wonder if linode offering 8 cpu cores of which how many of that is just hperthreading.. comparing linode 8 cores to digitalocean 1-2 cores I found this result form a review source
UnixBench results
DigitalOcean1G:
UnixBench (w/ all processors) 1387.1
UnixBench (w/ one processor) 1386.6
DigitalOcean2G:
UnixBench (w/ all processors) 1873.1
UnixBench (w/ one processor) 1183.7
Linode1G:
UnixBench (w/ all processors) 1860.7
UnixBench (w/ one processor) 491.4
Rare but they happen. We've had a few people come into IRC screaming about getting suspended a 2nd time. They then straight up say "I didn't feel like fixing my table scans/super abusive scripts". They were sent packing.
We're pretty reasonable and the monbot system helps out a lot. It keeps things in check and in a few weeks when I commit the next revision it'll monitor IO abuse on OVZ.
You pretty much have to be Robert Clarke to get canned around these parts.
Francisco
If you want to use the resources to be a jerk, fine, but look at a provider that accepts jerks for the higher price.
We dont accept jerks.
I believe in maximizing the output fromt he limited resources.
I.e. if you purchase a car during warranty you do not come up with schemes to leave the car running around in circles in your yard just to test if the car will last during said warranty.
We have OVZ servers using 10% of the resources. KVMs even 5%.
Is this good ? I dont think so, this is why we came with OVerZold brand. There usage varies between 30 and 70 percent, usually, with occasional spikes from abusers.
All are overcommitted.
It allows minecraft servers, hosting, many usage types to scale up during day or evening in that part of the world and down during night while others use the free resources.
If you need price increase, you do not need to force it on providers, you can achieve the same goal by buying from Linode/Amazon, they have higher prices, therefore better services and you can allocate there all the disk you want, all the ramdisks, etc.
You are unhappy with low prices ? Fine, you can go find higher.
You like pollution, energy and hardware waste ? Fine, buy american, allocate allt he resources you want. Complain about 200 MB/s dd result and force new underused drives upon providers, it will cost more, but at least you can waste the resources as you like.
Just because is cheap does not mean it should be wasted. Someone's livelihood in India may depend on those 2 dollars. If you can buy 200 hamburgers a day, doesnt mean you need to stuff them in the garbage bin, perhaps donating to the hungry is a more beneficial idea ?
This can only be imagined by americans, everyone in the world tries to use less resources for the best output, they try to use more fo rthe garbage dump of for harming themselves (getting fat).
We dont accept jerks.
I believe in maximizing the output fromt he limited resources.
I.e. if you purchase a car during warranty you do not come up with schemes to leave the car running around in circles in your yard just to test if the car will last during said warranty.
We have OVZ servers using 10% of the resources. KVMs even 5%.
Is this good ? I dont think so, this is why we came with OVerZold brand. There usage varies between 30 and 70 percent, usually, with occasional spikes from abusers.
All are overcommitted.
It allows minecraft servers, hosting, many usage types to scale up during day or evening in that part of the world and down during night while others use the free resources.
If you need price increase, you do not need to force it on providers, you can achieve the same goal by buying from Linode/Amazon, they have higher prices, therefore better services and you can allocate there all the disk you want, all the ramdisks, etc.
You are unhappy with low prices ? Fine, you can go find higher.
You like pollution, energy and hardware waste ? Fine, buy american, allocate allt he resources you want. Complain about 200 MB/s dd result and force new underused drives upon providers, it will cost more, but at least you can waste the resources as you like.
Just because is cheap does not mean it should be wasted. Someone's livelihood in India may depend on those 2 dollars. If you can buy 200 hamburgers a day, doesnt mean you need to stuff them in the garbage bin, perhaps donating to the hungry is a more beneficial idea ?
This can only be imagined by americans, everyone in the world tries to use less resources for the best output, they try to use more fo rthe garbage dump of for harming themselves (getting fat).
I've been nearly bit by that several times in the past. Also dodgy 3ware firmwares.Microrant: It just kills me when I see providers adverise "RAID 10 Protected Storage"
Rebuilding a failed RAID-10 from a finnicky array is really bad.
wlanboy
Content Contributer
Totally agree with your arguments.
I know some providers that are using timezones to load balance their clients. This is working if your clients are spread around the world.
The managment might be not that easy, because you have to migrate clients if they change their "hours of load".
But why should a client demand the whole resources if he/she needs it only between 5 p.m. and 8 p.m. CET?
It is like the idea of car sharing. Why buy a car if you only need it two days a week?
Here is a simple way to look at things:
- provider should make sure that safeguards are in place for customer not to be able to abuse node
- CPU abuse, I/O abuse, network abuse and so on are wrong and provider has to deal with abusers
- disk space is a finite resource that a customer is paying for, and as such he is entitled to use all of it, regardless for what; occupation of disk space that a customer has paid for cannot be considered abuse under any circumstances
Last edited by a moderator:
1. Cpu, bw, ram, disk, ddos response, support tickets with dumb questions, etc. are all finite resources that the customer pays for. Despite this there are usually not codified limits on most of these except for ram, bw, and disk (and bw is way overpromised for most users). Instead, hosts rely on behavior of normal clients doing normal things statistically averaging out to result in costs that the host can set prices by. Some people use all their ram or disk (rarely both), others underutilize, it evens out. Trying to codify too much invites obnoxious behavior. Better to tell people to just be good neighbors.
2. Ram and bw use tend to be low or bursty for most users and steady (high) for some. E.g. my ram usage is usually low but sometimes I run a compilation or something that uses a lot temporarily. Someone running a database probably uses their whole ram allocation most of the time. This all averages out.
3. Cpu also tends to be low/bursty or steady/high, but steady/high cpu users are basically not allowed on vps's. I run multi-day analysis tasks at 100% cpu which is why I got into dedicated servers.
4. Disk is not really like the above. There are few bursty users of disk space. Clients tend to use a steady and maybe gradually increasing amount, that may or may not be much of their allocation. But it won't fluctuate much. So here, hosts can observe total usage and add capacity as needed. I had thought LEB wisdom of last year was "don't oversell disk" but I guess it is a normal practice. I have a $15/year Ramnode vps with 50gb disk, that's probably oversold. I have it almost empty at the moment.
5. Having underprovisioned ram/cpu/bw on a host node results in crappy vps performance which is the usual complaint against overselling. Having underprovisioned disk has no effect unless the host actually runs out of space, which results in vps's not being able to save files, creating disruption. I think this it's relatively rare for any host to ever let this happen. They monitor total usage and buy more drives if they have to. So "reserving" space isn't protecting from shortages, it's just increasing host costs and therefore (indirectly) prices. Thanks a lot bro.
6. Hosts generally offer multiple sizes of plans, so if you didn't buy the smallest one, that probably means you probably DO use more of some resource than the smallest one offers. It's unusual though to use more of EVERY resource than the smallest one offers. If I buy a 500GB storage kvm, chances are I'm going to really use most/all of the space, or else I'd have bought a 250GB or a normal VPS. But there's less likelihood that I'll use the 512mb ram that comes with it. If I buy a 128MB $15/y vps, there's good chance I'm running a vpn/proxy/small server and don't need a lot of disk, but maybe I do. So it's nice if they say "50gb allocation" and it's there if I need/want it, like the 500gb bw allocation (I might use 10gb tops). Again it's fine for hosts to figure out the average case and provision for it. Storage server disks will probably be more fully utilized than normal vps's, etc. I have a 1gb ram, 150gb disk vps at ipxcore and I'm using most of the disk but very little cpu and ram, because it's mainly a storage server. If ipxcore observes this is typical for users of these plans, it's fine if they provision their box accordingly.
7. The oversold underperforming vps's from certain notorious hosts are just crap to begin with, I don't see why anyone buys them. None of this discussion really applies to them much.
8. I do like the idea of lots of people storing cryptographic random data on VPS's, and transferring it over the internet between servers, as a way to confuse PRISM. But I think it's a bit unsociable to actually fill VPS allocations of substantial size this way.
2. Ram and bw use tend to be low or bursty for most users and steady (high) for some. E.g. my ram usage is usually low but sometimes I run a compilation or something that uses a lot temporarily. Someone running a database probably uses their whole ram allocation most of the time. This all averages out.
3. Cpu also tends to be low/bursty or steady/high, but steady/high cpu users are basically not allowed on vps's. I run multi-day analysis tasks at 100% cpu which is why I got into dedicated servers.
4. Disk is not really like the above. There are few bursty users of disk space. Clients tend to use a steady and maybe gradually increasing amount, that may or may not be much of their allocation. But it won't fluctuate much. So here, hosts can observe total usage and add capacity as needed. I had thought LEB wisdom of last year was "don't oversell disk" but I guess it is a normal practice. I have a $15/year Ramnode vps with 50gb disk, that's probably oversold. I have it almost empty at the moment.
5. Having underprovisioned ram/cpu/bw on a host node results in crappy vps performance which is the usual complaint against overselling. Having underprovisioned disk has no effect unless the host actually runs out of space, which results in vps's not being able to save files, creating disruption. I think this it's relatively rare for any host to ever let this happen. They monitor total usage and buy more drives if they have to. So "reserving" space isn't protecting from shortages, it's just increasing host costs and therefore (indirectly) prices. Thanks a lot bro.
6. Hosts generally offer multiple sizes of plans, so if you didn't buy the smallest one, that probably means you probably DO use more of some resource than the smallest one offers. It's unusual though to use more of EVERY resource than the smallest one offers. If I buy a 500GB storage kvm, chances are I'm going to really use most/all of the space, or else I'd have bought a 250GB or a normal VPS. But there's less likelihood that I'll use the 512mb ram that comes with it. If I buy a 128MB $15/y vps, there's good chance I'm running a vpn/proxy/small server and don't need a lot of disk, but maybe I do. So it's nice if they say "50gb allocation" and it's there if I need/want it, like the 500gb bw allocation (I might use 10gb tops). Again it's fine for hosts to figure out the average case and provision for it. Storage server disks will probably be more fully utilized than normal vps's, etc. I have a 1gb ram, 150gb disk vps at ipxcore and I'm using most of the disk but very little cpu and ram, because it's mainly a storage server. If ipxcore observes this is typical for users of these plans, it's fine if they provision their box accordingly.
7. The oversold underperforming vps's from certain notorious hosts are just crap to begin with, I don't see why anyone buys them. None of this discussion really applies to them much.
8. I do like the idea of lots of people storing cryptographic random data on VPS's, and transferring it over the internet between servers, as a way to confuse PRISM. But I think it's a bit unsociable to actually fill VPS allocations of substantial size this way.
Last edited by a moderator:
Even worse, you want to rent the car for 2 days a week but use it 24/7 and be pissed if the rent-a-car company shows you the door.
Last edited by a moderator: