Ask HN: Is there a website that tracks excessive writes to SSDs in OS/app betas?

[amichail]

If not, why not?

That way, people can choose whether they want to install various software betas based on how much SSD wear they are willing to accept.

And since this data would be made public and available in one place, software developers would be more careful about excessive logging in their betas.

[CharlesW]

> If not, why not?

At least on Apple devices, this is not something anyone has to worry about.

For example, on my 2022 Mac Studio daily driver, on 24/7, DriveDx (https://binaryfruit.com/drivedx) reports my SSD lifetime left at 94%.

(I'm aware of the February 2021 "wear-gate" scare involving the first M1 Macs. This turned to be incorrect SSD-wear reporting.)

[radicality]

On my personal computers I keep track of this and ensure nothing goes write-crazy.

On my work laptop though (M3 Pro, 1TB), I’m at ~70% health. Iirc the lifetime writes are at >700TB. I do develop on it, but a lot of it is due to the various endpoint security software that’s always there logging every single possible thing but I can’t do much about it. I expect the ssd to die first in this laptop at this rate, so best I can do is ensure everything always synced/backed up from my local work laptop.

[yellow_lead]

With some badly written apps like Codex though, this may be a concern [1].

[1] https://github.com/openai/codex/issues/28224

[amichail]

What about people who want their laptop to last 10 years?

[justthehuman]

Then you probably want something that has replaceable parts and not an apple product (e.g. framework)?

Not trying to discourage you from building your thing, but just figure that's worth considering.

Also, I could see why developers would want excessive logging in their betas, so that's kind of a catch-22 (without the verbose logs it might be harder to fix bugs found in the beta)

[ValentineC]

For Macs, treat a renewable AppleCare subscription as part of the cost.

[JohnFen]

You can replace your SSD without replacing your laptop.

[waffletower]

I think people have forgotten what year it is and how ubiquitous SoC and unified memory designs have become.

[amichail]

On some laptops, it requires a complete motherboard replacement.

[JohnFen]

Really??

I would consider such laptops not fit for purpose and wouldn't buy one in the first place. Of course, that doesn't help if that's the sort you have. Sorry.

[bot403]

It helps reduce size and weight since you can just directly place ssd chips on the motherboard instead of planning for connectors.

Just depends on what your priorities are.

[amichail]

MacBooks are like that.

[baal80spam]

Well if 4 years decrease SSD life by 6%, then 10 years should decrease it by ~15% - so you have plenty of wiggling space.

[esseph]

Then you probably want enterprise level drives, motherboard, ECC RAM, etc.

[justthehuman]

I know I do!

Last time I looked at Laptops that supported ECC it was a very short and very expensive list. At least I could get enterprise datacenter SSDs... once upon a time...

I know that I DONT want iLO/IPMI/Redfish on a laptop though, the out of band stuff eats up a ton of power.

[embedding-shape]

I feel like the only responsible thing is to have a home network with at least one persistent device running 24/7 with large storage for local backups of things, and ideally a E2E cloud backup of the same thing. With this, comes the possibility of using things like Grafana+Prometheus to monitor your personal computer/workstation as well. This is what I've done, together with alerts from various things, and since the recent "excessive writes by codex" issue, also alerts if there is obsessive writing over hours long period from any process on my workstation. Luckily I wasn't hit by the Codex issue myself, but certainly made me consider the issue more generally so, happy accidents :)

[tosti]

Just configure wear leveling in your storage layer. FDE may be the easiest way ti do it. Writes will be done linearly in circular fashion. That makes it difficult to delete stuff, but with encryption it's moot.

[markhahn]

devices already do wear management.

at most, you should configure your system and OS to give them a chance. for instance, being 100% full all the time is just bad. use TRIM and log SMART metrics.

[justsomehnguy]

If you are not using bottom of a barrel and you are not writing 4K HD 3H mkv every day then you wouldn't have a problem even with a 0.3 DPWD drive of a sufficient capacity.

[KomoD]

> 0.3 DPWD

What does this mean?

[justsomehnguy]

Drive Write Per Day (typo in my previous comment)

From the horse mouth: https://www.kingston.com/en/blog/servers-and-data-centers/un...

0.3 DWPD is for the cheapest ones but even for them you would never exhaust the flash resource in a common home usage.

To give you an idea:

a 120GB drive with 0.3 DWPD "allows" you to rewrite 120 x 0.3 = 40GBytes a day for 3 years straigth, which is 280GB/week or 1200GB/month.

for a 1TB drive it would be 300GB/day, 2100GB/week, 9000GB/month.

[markhahn]

why do you think this is a function of betas? or even of OS or app versions?

I guess you've been burned, or are worried about being burned, by some software that generated excessive writes? I think you need to describe that first, since this is a rare phenomenon.

which is probably why there isn't some kind of downdetector infra tracking it.

you know you can access smart counters on any OS, right?