Hacker News Reader

Comment by chaxor | Hacker News Reader

It's still very bizarre to me to see people completely write off spinning platter drives as 'old tech'. They're still used everywhere! (At least in my world)

We have all of my teams data on an array of 18tb drives for a 100TB raid10 setup, and a NAS at home doing the same, etc. Even some of our OS drives at work are 7200rpm drives - and we're a computational lab. Why is everyone so intent that these effectively no longer exist? The cost for a decent amount of space with NVME drives is just far too astronomical. We're not all millionaires.

View on HN

Spinning rust is still by far the cheapest way to store bulk data, and it is fast enough for infrequent access. It's perfectly fine for archival purposes. Large disks are $0.02 / GB while large SSDs are closer to $0.07 / GB.

On the other hand, ignoring my media collection, both my personal computer and server only need a few hundred GB of storage. SSDs are cheap enough that they are a no-brainer: they are physically a lot smaller, quieter, and more resilient. The orders-of-magnitude better IO speed doesn't hurt either, even if most of it is wasted on me.

1TB NVMe drives are now less than $75. I could get a 1TB HDD for $40 or a 3TB one for $75, but why bother?

Even 2TB SSDs can be bought at < $130 now. Plus the fact that consumer motherboard usually have 2 m.2 slot. You can have 4TB of ultra fast storage on a normal consumer level computer without pcie extension cards. Which is overkill for most people. Most consumers probably don't even need that much storage space.

until they start shooting video or finetuning stable diffusion or something

video and archived neural net snapshots are pretty sequential

I don't think you need nvme ssd just to play video. For that kind of usage. SATA ssd is good enough. And neural net snapshots...? Do most people even do AIs?

yeah, that's what i'm saying about video, raw capacity matters more than bandwidth, and below the 100ms level, random seek latency hardly matters at all for video

right now most people don't use neural net snapshots but 25 years ago most people didn't use 3-d rendering, encryption, or video codecs either

I don't see anywhere in this article where spinning disks are written off.

It is sensible to consider the I/O speed of things other than spinning disks. Especially when they are exceptionally commonplace in consumer devices and developer workstations.

They're not used everywhere, most stuff desktop or servers are now based on SSD or equivalent.

Your use case of people running desktop with 100TB is niche, for $100 you can get a very fast 1TB nvme drive now days, which is fine for 99.99% of the population.

>a 100TB raid10 setup, and a NAS at home doing the same

How do you deal with these onerous constraints? Do you have a system for deciding how many copies of archive.org to keep locally?

I know you're being facetious, but here's some fun stats from archive.org.

A few highlights from the Petabox storage system:

No Air Conditioning, instead use excess heat to help heat the building.

Raw Numbers as of December 2021:

4 data centers, 745 nodes, 28,000 spinning disks

Wayback Machine: 57 PetaBytes

Books/Music/Video Collections: 42 PetaBytes

Unique data: 99 PetaBytes

Total used storage: 212 PetaBytes

Considering this data is from a year ago, it's got to be substantially larger now.

Link: https://archive.org/web/petabox.php

I'm not sure what exactly you mean - but we typically store DNA sequencing data from people that want us to analyze it, databases of proteins and scientific articles, ML models, etc. Proper considerations of how to store the data and compress it correctly help a lot, but unfortunately academics are pretty terrible at choosing the right data structures for the job, and often several copies are needed for mirroring the raw data, then the processed versions are updated.

Spinning disk are fine for some workloads. However, I've definitely seen teams engineering around slower spinning discs. NVMe isn't egregiously expensive, and if teams are needing to spend too much work and time thinking about disk speed it probably cheaper in the long run to switch.

I think you know the answer to your question, which is that normal end users who build PCs basically never use them. Power users with NAS and huge porn collections and servers, of course they still do.

Most people are conscious that they're making a performance trade off when they decide to store data in one.

And there is still demand. Otherwise they wouldn't be still produced.

Reply on news.ycombinator.com