Announcement: Changes to node payout rates as of December 1st 2023 (Open for comment)

[Toyoo]

@arrogantrabbit

Essentially you are cramming another, highly inefficient, filesystem into every chunk.
Because your ad-hoc filesystem cannot outperform host filesystem — end result will be worse performing node with extra steps.

Actually, the opposite is true. With regular file systems you pay with overhead for tons of features that are not used by storage nodes, starting from file ownership, all sorts of metadata like access time or custom attributes. Hard links come with a pretty big overhead, having to split directory indices from inodes, just like any other form of deduplication. And yet you still need to implement workarounds for common issues like bad performance of large directories on many file systems—in storage node’s case this would be the two-level directory structure in the blobs directory.

I have a rough draft of a design in which I estimate only about 48 bytes per piece would have to be stored in fast, random-access storage, as opposed ~280 bytes with a default-formatted ext4 or 1kB for NTFS, while significantly reducing other I/O necessary to perform uploads/downloads/file walker. This reduction would make SSDs redundant for many setups, as this data structure would fit in RAM easily. I’d love to have more time to refine it…

@arrogantrabbit

$10 gets you an enterprise SSD on eBay.

Note that you still have to have a free SATA/SAS/whatever connector to use it, and a bay/mounting point. Not all setups have one.

@arrogantrabbit

It does not matter. SSDs almost never fail, compared to HDD.

Can you cite data on that? I’d love to see actual studies on this topic, because there’s a lot of hearsay flowing around.

[arrogantrabbit]

@Toyoo

@arrogantrabbit

It does not matter. SSDs almost never fail, compared to HDD.

Can you cite data on that? I’d love to see actual studies on this topic, because there’s a lot of hearsay flowing around.

I don’t have data, but I have a little insight into SSD operation.

@BrightSilence

That’s simply not true. SSD annual failure rates are just below 1%. HDD just over 1%. Yes, they fail less, but not significantly so to not consider this.

Unlike HDDs, full spectrum of failure modes of which results in data loss, SSD are much more robust, and short of shooting through the controller it’s very hard to get it to lose data, as long as device stays powered:

• unlike with HDD, there are at least a dozen of various maintenance threads running on the controller all the time; curiously, the thread that serves host reads and writes is not even the highest priority.
• cells degrade gradually, and are constantly monitored, refreshed, and data is relocated long before retentions becomes a problem.
• [properly designed] firmware is designed to refuse writes if no more healthy cells are available.

This ensures that SSD fails in a controlled manner, by becoming readonly, and half decent filesystems handle failed writes by retrying them to the other storage device.

As a result, adding an SSD reduces reliability by negligible amount. Much smaller than if another HDD was to be added.

[arrogantrabbit]

@BrightSilence

There are 24TB HDD’s right now. If SSD caching or tiering was really such a good solution, it would make sense to still offer larger FusionDrive’s. But the performance just isn’t up to snuff for todays use cases anymore.

Hard drives don’t fit into Mac hardware anymore, physically. Except Mac Pro — but you can add third party controller, and have whatever solution you want.

With the rest of hardware — if you need so much data — you would use network storage or directly attached storage instead. Most users fit in half the terabyte; provided that the rest of the data lives in iCloud. I myself fall into that category. Just a different usecase.

@BrightSilence

Just because they do something doesn’t mean it’s the best choice.

They doing something has a higher chance to be the best choice than what I can come up with twigs abs sap in the closet — if nothing else but because stakes are much higher. It’s a good guideline to use. Not just Apple — but any large enterprise, that cannot afford to use subpar solutions.

[ACarneiro]

@arrogantrabbit

Not just Apple — but any large enterprise, that cannot afford to use subpar solutions.

They will all use “sub par” solutions if they can get away with it long enough.

[BrightSilence]

@arrogantrabbit

Unlike HDDs, full spectrum of failure modes of which results in data loss, SSD are much more robust

Oh absolutely. The flip side is that SSDs tend to fail hard and instantly when they do fail, with no chance to recover any data. HDD’s are better (though this is getting worse lately) at signaling impending failure ahead of time. But for sure things that fall under the bitrot category are far less likely to happen on SSD’s, as long as you don’t chuck them in cold storage.

@arrogantrabbit

This ensures that SSD fails in a controlled manner, by becoming readonly

This I have never encountered. Though I have had only 2 SSD failures, both were just gone and didn’t work at all anymore, with no chance to recover data.

@arrogantrabbit

They doing something has a higher chance to be the best choice than what I can come up with twigs abs sap in the closet — if nothing else but because stakes are much higher. It’s a good guideline to use. Not just Apple — but any large enterprise, that cannot afford to use subpar solutions.

Well, since not a single consumer solution uses tiering like that anymore and chipset implementations for it are either no longer promoted or phased out, I think the industry has spoken. The only devices still offering those mixed solutions are NAS devices, which require redundant SSD’s for tiering or r/w caching solutions, acknowledging the impact of SSD failure risk.

As for not using subpar solutions. I’ll just leave this image.

[arrogantrabbit]

@BrightSilence

This I have never encountered. Though I have had only 2 SSD failures, both were just gone and didn’t work at all anymore, with no chance to recover data.

Were these consumer SSD? Anecdotally, I have had quite a few SSD fail, (because I buy used stuff that is sometimes already worn out), and a few failures happened exactly as a described. It’s in fact part of the design requirements, the failure mode. I have never had an ssd (we are not talking about sd cards and other thumb drives) drop from this dimension abruptly and entirely. Does not mean it’s impossible, just means the probability is very small, because care was taken to design the failure mode.

This is not the case with consumer SSDs where price is the driving factor.

@BrightSilence

As for not using subpar solutions. I’ll just leave this image.

Everyone beating this horse, but I believe this is a brilliant design that works to solve a specific problem. Without discussing that problem it makes no sense to smirk on a solution.

There is a whole discipline called interaction design.

In this case the goal clearly was to ensure best possible user experience when using wireless mouse. Having cable attached while the mouse is in use ruins the experience — cable pulls inconsistently, aiming precision worsens. The standard lighting cable is pretty stiff. Old wired mice had a long flexible cable on purpose.

Apple did not want to worsen user experience. They could put a cop-out sticker “for best experience don’t use while charging” but this is bad. Who obeys stickers? So they designed the product in away that makes it impossible to use it while charging. Some people lol at “dumb designers” but the rest enjoy consistent experience.

Fwiw I don’t use Magic Mouse. I have large hands and it’s not comfortable to use for any nontrivial duration. I have used Logitech performance MX for a longest time. It uses AA rechargeable and can be charged while in use, but i avoid doing so, because the cable ruins the experience and annoys me, causing to misclick too often: when your workflow is optimized even slightest pull disturbs the aim. So I used to keep a box of spare batteries at my desk.

A year ago I migrated to their newer mouse (the old one worn out — physically; some rubber under the thumb is gone) that has built in lithium battery. You have to charge it. I would forget, and once every few weeks i had to work with cable attached. I hated it. I bought second mouse to not to have to endure that.

Maybe I’m too demanding, but I don’t think I am. I apply the same principles in my work. No compromises that affect user experience. If this is amusing for some people - great.

[Pentium100]

@arrogantrabbit

Having cable attached while the mouse is in use ruins the experience — cable pulls inconsistently, aiming precision worsens.

You know what’s worse experience than having the cable pull inconsistently, ruining the precision? Not being able to use the PC at all or charging the mouse for a couple of minutes and then hoping it lasts long enough to do something urgent.

That’s why it’s a stupid design IMO as it makes a small problem (some inconvenience using the mouse when it is charging) into a much bigger one (being unable to use the mouse at all). Though Apple users probably just buy a second mouse.

Bu what do I know, I don’t use Apple products and I use wired mouse and keyboard. In general, I prefer some inconvenience over not being able to do something at all.

[BrightSilence]

@arrogantrabbit

Were these consumer SSD?

To be fair. Yes.

I won’t quote your text about the magic mouse, but apple had only 2 reasons for that. Form over function and the condescending assumption that their users are stupid. Yes, it’s annoying to have a stiff wire hanging from your mouse if the battery runs out while you’re working. The only thing worse though… Not having a freaking mouse at all. This isn’t clever design. It’s babying your users because you assume they will keep it plugged in. At the cost of loss of functionality for those with more than a single brain cell. There are other major issues like it not being ergonomic because it’s so flat and way too small (though many suffer from that last one these days). I use a Logitech MX Master myself, not dissimilar to what you use and I’m glad I at least have an option if it runs out while I’m working. Replaceable batteries would be fine too, but it’ll be a cold day in hell before apple would choose that option.

[arrogantrabbit]

@BrightSilence

Replaceable batteries would be fine too, but it’ll be a cold day in hell before apple would choose that option.

But… but… the previous Magic Mouse: Change Magic Mouse batteries - Apple Support

They even sold rechargeable AAs and the charger

[BrightSilence]

…again. They will never do that again.
Did I get away with that?

Anyway, the rest of the point is still valid. It baffles me how some people twist themselves into a knot to defend Apples bad decisions.

Also, let’s try to get this back on track to the original announcement post.

[penfold]

@Pentium100

Bu what do I know, I don’t use Apple products and I use wired mouse and keyboard. In general, I prefer some inconvenience over not being able to do something at all.

I am same using wired - though I do like and use the full size Apple USB Aluminium keyboard. If not on an actual Apple device.

[daki82]

@BrightSilence

This I have never encountered.

I have, it was in 2010 or so my raid0 with 3x 256GB (i dont remember the brand) 1 drive going r-only. windows-xp did boot, but boy, with 460 error prompts. Got a Samsung as Replacement
Wich was in use until 2021 resting in my basement now.

[Alexey]

You was a little bit late, because of

Taking Payments to The Next Level with Raiden

or here

SNO Payment mega-FAQ

SNO Payment mega-FAQ What is Storj? We explain a lot more over on our website, but Storj is basically AirBnb for hard drives. Some people have free hard drive space (you?), and some people need hard drive space to store their data. Storj connects people who have space (storage node operators, or SNOs), with people who will pay to store data. Data is stored on the Storj network. How do I get paid as a storage node operator? Certain peers (called Satellites) on the Storj network manage details of…

or

Do I store data on someone else's computer?

It’s funny. That was often said about general cloud computing when it became mainstream 10-15 years ago. Here’s how it works: All over the world, there are zettabytes (billions of TB) of unused storage capacity. Rather than build massive, centralized data centers with proprietary equipment, Storj Labs acts as the Airbnb for disk drives. Storage node operators sign up to share their excess storage capacity and bandwidth with the network, committing to stay online at all times. Then, when data i…

We uses this comparison for a long time actually. I even was sure that you took this analogue from us.
If not - I’m very glad that you figured this out yourself! And your comparison is quite well describing the situation. Especially if we remove a sarcasm.