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.
Comments
Huh?
No, we absolutely did not ? Support ticket reference, please, or it didn’t happen.
As somebody already pointed out, the problem with NVMe is that you have to shut the server down in order to replace a failed one.
I don't have any such issues and don't recall any in the years past, are you using clouvider directly or one of the hosts in their DC?
I have a number of customers I know for a fact use sendgrid without issues, I hope you have some sort of technical troubleshooting process to back that up as jumping in on random threads and stamping your feet offtopic to get attention makes you look like a prize tool.
SSD is enough for now, maybe next year nvme become cheaper
Not entirely true. There are hot swap solutions available.
Good to hear but I suppose the most common configurations/offers aren't hot swap capable.
it depends, the newer, higher end solutions tend to have this option available.
NVMe not die easily!
It depends on the use case, but for 99% of the time I'll take HDD please. I'm a sucker for lots of gigabytes!
NVMe is a SSD with a faster interface, therefore it dies like every other SSD.
Not quite for example p3700 come with something like 15 DWPD and 2mil hours if I remember I don’t see any SSD come that lifetime.
Well at least with good modern enterprise grade ones they die in read only mode 99% of the time so data recovery is a breeze.
HDD is a big storage technology now, nothing more, its performance is just dog shit compared to even crappy SSD's never mind NVMe SSD's
NVMe, latency is king. SSDs aren't the fragile things they once were, improved flash and controllers offering things like wear leveling means it's a moot point. Also, don't build things that depend on single points of failure - use multiple servers/disks so one having an issue isn't a big deal.
Of course. I wasn't defending HDD, I just wanted to say that every kind of device can die, so even NVMe require redundancy.
Is this the dumbest thread in the history of threads? Pick 1 clearly superior technology and ask people if its superior? I'm expecting we'll see some offer threads from the OP soon.
It is a matter of quality, not because it is NVMe.
Can’t fully agree with that either. If you take a look at endurance rating, you’ll find that similarly priced SATA SSD tends to be rated for less endurance than similarly priced NVMe SSD, manufactures tend to anticipate more writes on the faster drive.
Can be, but not because of the technology itself, I suppose it is because of more modern production process. I would never use them without any form of redundancy, anyway.
Yep, agree on that 100%, likely a business reason - what use to you would be all this performance if the drive would die 5 times sooner
.
to be honest I think NVMe is a real thing for providers to begin with, not neccessarily customers.
because it is faster, it simply allows a higher density on the very same server without running into the IO bottleneck.
from that PoV what is supposed to be changing for a customer? I'd say not much - either you are with a provider who's already cramming a lot of people onto a single node. with NVMe he most likely will do so even more - so performance might be a bit better for IO, but network and cpu might then become a more serious bottleneck...
or you are with provider who doesn't oversell that much in the first place. your performance probably is already decent with whatever setup you are on - again NVMe will only enable the provider to balance the ressources better to oversell a bit more without damaging your good performance anyway. not much changes.
believing that a provider puts in more expensive NVMe disks and does not either raise prices or number of clients per node more likely is wishful thinking ;-) ;-)
TL;DR; in most cases I'd still rather go with SSD/HDD instead of NVMe to not be put on even more overcrowded servers at all.
I'd still rather go with SSD/HDD instead of NVMe to not be put on even more overcrowded servers at all.
No agree the CPU will die before doing too much overselling with NVMe.
Is power consumption differences significant in a server? and between different models?
Both NVMe and SSD typically user considerably less (typ. -50% to - 75%) power than spindle HDD.
SSD is the future. My main rig will be spindle-less soon, planning to get one of those pci-e m.2 adapters (takes 3 of them) and use exclusively m.2 instead, saves me cables and space.
I was talking about the OP's post, SATA vs NVMe SSD
There’s many taking part in this discussion... At the end of the day it’s a forum and people have the right to ask open questions.
It’s not a case of what’s superior, it was an open question to see what consumers prefer.
title:
Come again?
A title that attracts attention... arrest me officer.
The title stands, what’s people’s general opinion on NVMe, is it going to be industry standard in the next couple of years, but yeah maybe some can’t get there head around it.
Most got my jist by reading the thread.
Well if you get second-handed Intel NVMe SSDs from eBay that may be possible, but it's another story (and the endurance rating of the SSDs are probably less so the VPS can be unstable at times).
For the typical server model, a NVMe or AHCI/SATA SSD will definitely consume way less power than spinning rust.
However, a lot of mobile platforms still stick with AHCI or SATA based SSD's because the power management controls there are much more mature compared to NVMe drives and really power management on the NVMe's isn't all that unified.