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.
how slow are hdd for server
EpicEscapist
Member
in Providers
Hey, I’m looking at a dedicated server from OneProvider with the Intel Xeon E3-1270 v6, 32 GB of RAM, and 2x 2 TB SATA HDDs. I’m pretty interested in getting it, and honestly the only thing holding me back right now is the HDD storage. How slow are those drives in actual use? I’d mainly be running Linux, hosting websites with Nginx, and probably some VMs and other services. I saw there’s a 2x 450 GB NVMe SSD option, but unfortunately it’s sold out. I don’t necessarily need crazy-fast storage, I just don’t want the HDDs to be terribly slow or make the whole server feel sluggish. What kind of performance should I expect from them?
Comments
2x hdd raid 1 around 150-220mb/s sequential, 80-150 random 4k iops. budget ssd does 20k-90k+ iops, nvme 100k-500k+.
in ur use case, its damn slow. iops matter for vms and databases.
Those are just resold from OVH - better go for the original, even with their stupid setup fees. And go for a NVMe one - they will pop up once but probably only for seconds or minutes before being oos again.
Very slow. You can even feel how slow it is while doing apt upgrades/installations
Either wait for the nvme stocks on that server or regularly check out https://heymman.com. You can get the E3-1240 v3, 32gb ram, 1.92TB Samsung PM863A SSD server for $22.
HDDs are pretty slow compared to NVMe or even SATA SSD.
But you can use dm-cache to improve performance if you have both an HDD and an SSD. It will dynamically cache the most frequently-accessed blocks on the faster storage. It can improve performance dramatically.
God. that's a really hard to answer for you. You really need to test it out for yourself, there's no way around it.
I'm personally perfectly content, maybe because I come from the 90s with them, with raid 1 HDDs on a dedi as Linux just runs and I really don't have anything that requires quick access (I've that hosted somewhere else on ssds)
If we would talk windows, ssds all the way. Holy crap just no.
FTFY
I mean, good to know they resell other providers servers. Looking for someone who owns their own stuff that's in Canada/the US and that's dedicated and that's around $30.
@SolidVPS ?
I don't see any dedicated from them, unless I'm just not looking far enough
If it's for your own stuff and light IO use it's fine but if you've ever used NVMe storage then you'll feel like you're going back to the stone ages using HDD storage. Just do an apt/dnf update in Linux on NVMe vs HDD
And try to use dokploy on the hdd. The provider selection (where you select the source of the application. Like github, gitlab yada yada) loads in 5 seconds. Not even kidding.
This looks like a crazy deal from wholesaleinternet or Nocix
never used HDD before, so that's why I was wondering mainly
also don't know if that's bad, but I don't plan to run windows on the server itself, just maybe a windows server
Ouch if you want to run Windows it'll be even worse.
even in like a virtual machine?
It's worse in a virtual machine because the HDD is shared by many people.
Why can't you use a small SSD as a caching drive alongside a larger HDD?
no, I mean, like, I'm renting a dedicated server that has an HDD, unfortunately I can't do that because like I said, the SSD option is sold out, otherwise, if that wasn't the case I wouldn't have thought twice
That's crazy to never have experienced running things off an HDD. I remember setting up RAID0 on 10/15K drives. Hearing those little platters spinning up to 100mph felt different. Or when you really wanted that extra performance and you had plenty of RAM to spare - setting up RAM disks so you can swing your big epeen with some awesome benchmarks.
That was then. SSDs became cheaper and a HDD won't hold a match to it. You're not confined to the diminishing speed depending where on the platters you are (further to the edge is faster and more dense while closer to the center you'll see a 50%+ drop in perf)
Running a pure HDD setup these days can be painful. I recently tried installing/running windows on a HDD and was bluntly reminded how much of a POS it truly was. When we got SSDs and they can trim nearly instantly, HDDs you'll be running that defrag for hours. Then we got SMR HDDs, and you'll be reminded even more how craptastic they are, especially for more demanding workloads.
I'm not saying you'll have a bad time with all HDDs, but you'll probably not have a good time. Especially with windows.
There is a lot of factors to take into effect on mechanical disks. Physical size, rotation, on-drive cache, SATA vs SAS, regular maintenance (defrag), defects monitoring, and so on.
Not sure how much it'll help with 32G of RAM, but maybe you could do something [stupid] like using dm-cache on a ramdisk. Just don't reboot
No no no! That would be horrible lol. With 32 GB RAM you should just let the page cache do its thing. Use a program like
preloadto automatically page in often-used applications into memory on boot.anyone know any places with SSD servers? The ones I saw I like from 2010 and DDR three which I don't know how viable that is in 2026
not the servers, I mean the processors were 2010, forgot to clarify
Is it viable in 2026? Well that depends on your workload.
Well where's the fun in that? Cache more on unbuffered non-ECC memory for extra excitement.
It can host a website easily, but running a VM on an Atom is going to be painful.
probably, but probably not suit me then because that's one of the main things I planned to do
A lot might depend on your RAM usage. You might well find a lot of the time the data you need is still cached, so it might not hit the disk much.
Actually, my primary dedi is an old OVH machine with 2xHDD with stripes / mirror depending on how important that VM is. It holds the master copy of all my websites, git repositories, e-mail, DNS, runs certbot, and probably several other things I've forgotten. It is set up to serve content as well, but it's not listed in any DNS for my websites and that will just be switched on in an emergency.
I've also got a dedi from Hetzner with 2xNVME and it's significantly faster, and the VMs on there are the primary web server and application services, as well as my build server and all of the backend jobs that require CPU. I've got a third from fiberstate, again with NVME and slowly adding services to it, but it's still mostly idle.
But honestly, I do a lot of work on the HDD based OVH, and it's fast enough for me, but because it's not being hammered, just the master copy of stuff that other machines clone from, it's probably only ever max 10% CPU or HDD utilisation.
I'll put it like this: I remember quite well when long ago I had a "stoneage" dedi with Xeon way before even E5v2 and HDD ... and ran a quite well visited community on it. In fact back then it was considered quite fast.
Of course looking from today even an SSD looks shitty and a four core old Xeon as well.
But it did the job, even well
very slow ...