New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Digital Ocean - Introduces High CPU Droplets
Nice!
We’re excited to announce the next generation of Droplets on DigitalOcean.
High CPU Droplets are optimized for computationally intensive workloads, providing maximum access to Intel’s latest and most powerful CPUs including the Intel 8168 (Skylake) and Intel E5-2697Av4 (Broadwell).
Increasing vCPU performance as much as 2.5x to 4x compared to Standard Droplets, High CPU Droplets are ideal for applications such as active front end servers, data analysis, batch processing, and CI/CD servers. High CPU Droplets are available in 5 new plans starting at $40 / month for 2 dedicated vCPUs, and can be configured with up to 32 dedicated vCPUs.
Comments
A w e s o m e !
Droplets > Buckets ?
Shitstorms > Droplets > Buckets
The question is: can it mine?
Tried it. Can barely get over 2kh/s XMR. Mining on DO's droplets will lead to either a suspension, or a large loss. Price is unreasonable, being that I was charged $6.66 US for around 7 hours of usage. Don't expect it to be useful for anything other than crunching predictions for the next deadpool.
(this was done on the largest CPU droplet, with 48GB memory, and a Xeon E5-2697 v4 @2.6 GHz, 40MB cache.)
A w e s o m e !
They suspend you even on the high-cpu droplets? Those don't sound like dedicated cores to me.
Why would anyone CPU mine on a cloud server unless they liked losing money lol
CPU mining is silly but the issue is whether you get suspended for high cpu use on a supposedly high-cpu server.
Doubt it, but seems like the wrong question to ask anyway. No one is buying those for CPU mining unless they're not the ones paying for them. If you get what I'm saying
Silly LET logic. "Mining on cloud servers". Anyway these servers are not for let, its designed for serious people with real world applications.
About the suspension , well it depends. If the suspension says, "Your usage for the past 7 hours is too high" then that might be ridiculous but if it says "We have detected that you are in bridge of TOS by using mining software" , then that's justifiable.
It is KVM and that would immediately mean breach of customer privacy as someone entered the VM. It is not external visible if you mine.
That's not what Vultr staff said.
I'm unfamiliar with the internals of it, but I'm sure you could identify miner traffic on the network. Mining software doesn't sit there spinning wheels for no reason, it sends and receives data. Deep packet inspection is kosher.
With that said, why even bother CPU mining? Isn't the difficulty so high that you need a GPU or ASIC to make any progress nowadays?
Or you could just bust them on credit card fraud because no one is CPU mining at these prices.
Github Student Pack
BTC mining VM
...?
Profit!
Well, this was even profitable in the early beginning of DO on regular droplets.
Let me put it another way:
1) What happens if I run a different task that's 100% CPU on all cores 24/7 on one of those droplets? Is that CPU abuse? (Of course I wouldn't use droplets that way, too expensive, it's why I have 5 LET dedis now [facepalm]).
2) Is DO doing something to detect mining inside droplets, or is it just about CPU usage? Of course the network traffic can be proxied between the droplet and the BTC network, so can't easily be picked up by watching the droplet's network destinations.
3) actually there are usually altcoins for which CPU mining is still profitable. Though it's saner to do it with dedis than droplets.
Shouldn't be, but do keep in mind that these are not 100% dedicated servers. It is always possible for one tenant to cause issues for another, even if difficult under certain circumstances. So while maxing out those CPUs should not be counted as abuse, there should never be any blanket statement made to say that there is no way that you can abuse it. It is plausible that abuse of the server in some way that impacts a neighboring customer (intended or unintended) may also correlate with maxing out those CPUs.
Should be the same for Fran's slices I assume. You shouldn't be able to abuse it easily but if you manage to do it, and it is theoretically possible, it should be dealt with on an individual basis with the context of that particular event in mind.
There is often a tendency to want to know all of the exceptions up front and have them clearly defined, but you and I know that the hosting industry just doesn't work like that. Every time you document all possible exceptions, something you never imagined pops up. If you define all of the exceptions and have to continually alter them, you are changing agreements with customers on a constant basis and it's not a fair thing to do. Instead, the concept of the exception should be left vague enough that it covers exceptions, but also clear enough to define "This is how I'll know it's an exception." So the real answer as to what is an exception is "I'll know it when I see it, but the basis of it is you're doing something that negatively impacts neighboring customers."
Hope that makes sense
We have hypervisor level graphs for each droplet. They're the same ones you see in the control panel if you don't have do-agent installed. We have no visibility inside the server. I can tell you if someone is CPU mining with reasonable accuracy due to simple deduction of what is available external to the droplet, but it's still a "best effort" that just happens to have a flawless success rate, rather than "I know because I can see the data." I hinted above at one sign, you could guess the other.
True. Yeah with dedi prices as they are today, it just doesn't seem smart by any measure to me.
@jarland The high CPU (32 core) one seems to be dedicated. I am able to use all 32 cores with no throttling.
Page 375 of the 791 long ToS says your VPS access and all data on it belongs to Choopa-Vultr-Cheney Holding SpA. - ALL GOOD, NO WORRY.
But yea, CPU mining unless you snoop into the actual CPU usage patterns (or the executions in depth, but dumping these will insanely degrade performance) i doubt anyone can reliably detect that is a miner and not a very efficient $whatever app.
DEFINITELY. Your data is SAFE with us!
To conclude, these servers are really not worth it. In 2-3 months, you'll have the cash to build a comparable dual-socket server. Wait a few more and it'll have a GTX 1080, and NVMe.
Not sure about the move, as @jarland said. It's not really worth it, even for a large company. I guess that it'll work for short term computations, but I'm sure that a dual E5 on a newly built server will cream whatever DO can put into the $640/mo top-of-the-line droplet.
I don't want to say that it's a terrible move, but it definitely isn't for us. It's really expensive, and for the computations it can put out, it's more worth renting multiple E5s from a reputable provider. With that, you'll even have the added bonus of high availability.
Yeah, a job where a monthly (or longer cycle) rented or colocated dedicated server is the perfect for the job won't be replaced by this for sure. I see in my mind a vision of someone who automates via API the creation of a server, the necessary computations or workload, and then spins it down again. Could be wildly beneficial to someone who needs that power but only needs it for an hour. Especially where infrastructure is large enough that those line items on the long balance sheet make a significant dent.
F
A
K
E
N
E
W
S
On our slices you can use 100% of what you pay for 100% of the time, we don't cap that. If you buy a 1GB or 2GB slice and rip the core 24/7 we reserve the right to hard cap you down to the limit the plan states.
Francisco
Yeah but you can abuse more than cores, and that abuse could correlate with high CPU usage. I'm sure that when someone causes a load of 1000 on a box outside of the VM, despite KVM "not allowing" that (it's not that flawless), and someone complains, you're going to do something
There's nothing healthy to be found in saying "dedicated cores" = "you could never possibly abuse resources." Not every single resource is dedicated or isolated so flawlessly that neighbor impact is impossible. Just less likely. I don't trust virtio enough to call it perfect. It would be dishonest to say it's impossible and that's what I mean here.
I learned a long time ago not to tell LET members that there is absolutely no way to abuse something. If anyone can find a way to do it, no matter your safeguards, that someone is a member here