In Chia’s consensus, the plot filter is a constant, which each plot has a 1/plot filter probability of being eligible for a challenge for a given signage point. The plot filter’s primary benefit is a reduction in disk I/O. Its primary tradeoff is an amplification of the effect of plot grinding. Plot grinding will be feasible with multiple GPUs, and if nothing is changed with Chia’s consensus or plot structure, it may become economical with future changes in xch price and advances in computing technology. This proposal will mitigate the cost-effectiveness of plot grinding by reducing the size of the plot filter every two years, to 256, 128, 64, and finally 32.
This is a bit over my head. The absolute catastrophe for Chia will be any widespread reports of drive failure. So when those OG famers with drives getting into 3 years of farming, should those start failing? For sure this will become a widespread concern. Chia folks say nope. Drives will be fine. To that? I say drives will be fine when you know what you’re doing. When you run them cool and you aren’t running consumer grade, used hard drives 24/7. But what you should do, and what best practices are? Two different things. The ENTIRE proof of work concept of Chia hinges on storage longevity. It might be why they are giving some bogus timeline when the world will instantly switch over to SSD technology and will save the day. This is about as logical as saying you are building a 10 million dollar solar farm so that you can power a massive Chia farm rather than powering something slightly more useful than powering a blockchain. I’m not sure if people want to remain oblivious to the fact that Chia is a new idea. A new system. An untested, unproven system. In theory drives should be fine and should run long enough to ROI. Theories don’t pay bills.
Any spec is under idea/proper conditions. A lot of people buying used. You don’t know heat or how the drives were handled. Right, and Mr. Personality is the expert all right. He’s predicting the global SSD takeover right? He is a futurist too? Crystal ball reader? Chia is going to put all this to the test. Maybe it passes, maybe it doesn’t. The entire success depends on this, but hey, farming is so light those drives will last long enough for ROI. Sure, sure. Chia farming is not “normal” usage of a hard drive. I’ve already asked previously if anyone could name anything out there were people are pinging consumer grade hard drives 24/7 for years.
If Chia is starting to talk like, we won’t worry about added hard drive workload because everything will be using SSD? That sounds to me like covering their ass future tense. Our plan is to increase demands on the storage for the sake of the blockchain, but don’t worry, most people will be using SSDs so it won’t matter? Pretty arrogant to assume farmers want to invest in SSDs for the sake of securing a blockchain.
The reason I’m sounding my personal alarm right now is because it seems to me the direction with compression and GPUs, is that the workload on drives is going up. Not the same. Not less. But more. Of course compression is optional. Chia is becoming less power efficient, not more efficient. As things progress, it makes sense that faster, more power, and more workload on drives is going to be required in order for the blockchain to keep up and keep safe.
Trending in the wrong direction. Everything starts out small. Small cracks.
currently you really do not need to worry that hard drives are loaded by the profs.
On the contrary, the load is so low that you have to make sure that the HDDs do not constantly change in the parkstate and are activated again shortly afterwards. this can then really lead to an increased failure because the hard disks are not designed for this.
If I remember correctly for a current test on a prof of a K32 plot when it passes the filter 6 or 7 I/Os are needed.
This is absolutely nothing and totally negligible even over 10 years. If this was ever a problem anywhere, you would have had a more or less defective HDD from the start.
Even with a filter of 32 it would be no problem with uncompressed plots that would be 16x as many I/Os as before for the same farm size. Of course, this is problematic with the compressed plots because here also the 16x computing power is necessary to calculate the additional profs and you need more GPU power or a lower compression level.
I’m willing to, but at a substantially decreased amount. If this thing totally implodes and starts shredding HDDs, I’m good for a few TiBs of SSD space (one of the perks of being a hobbyist ). But point taken, banking on future tech to solve a today issue isn’t trying hard enough. We’ll know in the next year or two how many bargain bin white label drives ate s#$! instead of surviving the 4 +/- years to ROI.
Hdd’s, especially the enterprise drives that most farmers are using are made for 24/7 heavy workloads.
Chia’s hdd usage is super, super light, so going from almost nothing to 2x almost nothing is still nothing.
Even hdd’s in datacenters last easily for 5 years or more.
JM’s calculations always assume 5 year lifespan because that is the common warranty of an enterprise drive. So, very conservative estimate on lifespan there.
Yes if you use old consumer drives they may fail earlier, no matter what you do with them. They are cheaper, less robust and already been used before.
Really nonsense to start talking about hard disk “shredding” because of a negligible increase in disk I/O two years from now.
The reason Chia wanted to keep operations as low as possible is because they wanted the farmer to run with minimal hardware requirement.
This Chip is relevant to consider in combination with plot compression as it will increase the workload on the CPU (more proofs need to be looked up and decompressed/completed).
JM is who exactly? A self-professed expert who says you need to spend up your A on hard drives getting the maximum energy savings? 18TB or bust!? If you spent any time in this forum you would be reading about failing Seagate enterprise hard drives. Yes, qty of failure, not just the odd failure. A lot of people went with external hard drives because of the price. I think it’s BS to say “enterprise drives that most farmers are using”. Talk about conjecture. This is exactly my point. Sure, let’s say every farmer is running drives cool and using enterprise grade SAS even though I have ZERO data to support that. My only point is that MORE work on drives is coming. Not less. THAT IS FACT. It’s trending in the wrong direction. “Time will tell”. Right. And the entire concept of being a farmer is to make money. You can’t make money if your equipment fails around the time or before you make any sort of ROI. Chia hopes that’s the case but it’s based on zero evidence or history. Unproven is the only word for it. Please prove me wrong about these new innovations. Chia shouldn’t be saying “it’s only a little bit more workload on the drives”. They should be able to come up with figures. Real data and tell people what the increases are. Saying, don’t worry. It’s still nothing. Your drives are fine. ROI can be had. Mad Max comes up with all this amazing stuff but not once did he answer questions about hard drive usage increases. If it’s nothing, then state what you know. Like the data. If he doesn’t know? That may be more the truth of the matter. Why ignore an aspect of your technology? Maybe some people are sheep and make assumptions. Or people don’t care about ROI. Maybe people don’t care about 5 years trying to make ROI, then 1 year of so-called profit, then drive failure. When in fact they could have made 2x the money by selling the drive a year ago. But who would know? Just conjecture, no facts. Except for what was stated above, which is fact. Workload increase is coming.
It’s not the compressed plots that add more work. It will be the reduction in filter time that will increase the already low I/O to another higher LOW I/O…
Thanks for the clarification. None of this is predictable. They say this is the timeline for filter time, but is it? What’s behind door #2? Bram said plot compression is impossible. Bram the expert and founder. Hmm. This timeline and change, so minimal as you say, is set in stone? Never will need to be altered due to unforseen issues or threats? The fact that the founder Bram said that compression or further compression of plots was impossible, yet here we are, that should say to everyone that they’re no guarantees of anything. The work on the hard drives is not going down. I think we can state that as a fact.
Bram’s statement about plot compression is still true , the original plots have always been compressed. NoSSD and madmax are calling them “compressed” plots but in reality the new plot format simply creates plots that are incomplete, some tables are missing, hence the need for extra CPU/GPU power when the farmer needs to provide a proof as these tables are now calculated on the fly during farming.
Change is going to happen in new tech. If you’re not ready for change when it comes to Chia, than maybe sit it out until things become more stable.
Not really surprised that all this change is happening. It’s all part of Chia being battle tested in the real world.
It’s interesting to see people’s innovations to Chia Farming and how Chia is adjusting things to maintain as close as they are able to their original concept.
No doubt there will be more change ahead, but thus far I’ve been impressed with how Chia has handled it all.
I believe in what they are doing. I wouldn’t be here otherwise. I didn’t fill my drives because this development was coming. I don’t knee jerk react. Stable is good. Chia is less stable than its even been. Yes, being battle tested. And yes, same reason Windows has the target of hackers. It’s called payoff. Worth their while. Chia hasn’t been on many radars. Let the battle testing begin. It’s still nothingburger though. I’m a farmer. A service provider. Nothing more, nothing less. They owe me nothing. I’m sure my storage will be of use for a future blockchain technology that learns from what Chia attempted to do should they trip up on unforeseen challenges.
It will be if there is blockchain consensus, until another consensus is made.
This CHIP is a proposal that requires a hard fork, and therefor a consensus of the majority of farmers(netspace) It will then be hardcoded into the chain unless another hard fork occurs. These are not CNI decisions that they can make.
A. He is a well-know industry expert on storage, hence his nickname “Storage Jim” And now Chia’s VP of storage.
B. He has never said to spend your money on HDD’s, he says to not spend your money on them (new harddrives), just like the rest of the Chia team. All he does sometimes is offer refurbished disk to those who want them.
C. The only thing I said is that he uses the warranty period to calculate figures. So if the drives fails before that you can send it in for RMA.
[quote=“MisterSavage, post:9, topic:18387”]
If you spent any time in this forum you would be reading about failing Seagate enterprise hard drives. Yes, qty of failure, not just the odd failure. [/quote]
I have spend a lot of time on this forum, and yes I have seen the few people who complain about Seagate drives, like people have done for the last 20 years. Not that much difference in the failure rates in reality. Although Some other drives do have better reliability numbers, differences are not that big or Seagate would be out of business a long time ago.
A lot of people went with external hard drives because of the price. I think it’s BS to say “enterprise drives that most farmers are using”. Talk about conjecture.’
Yes, like I did. Why? Because a lot of them had Seagate Exos drives or Ironwolfs inside.
Well there’s the rub. Right? You should you these sized drives and I’m going to tell you why and show you why. Oh, you need to buy some? Well, let me sell you some as well.
I’m sure he is smarter than any person on the planet when it comes to storage. Does that make him a bright business man? A futurist? A predictor of human behavior? Insider knowledge that the world will cease to make or use mechanical hard drives? Full of wisdom is he? He knows how we should best spend money. Like hardware cost vs. operational cost. Maybe it’s smarter to go with smaller drives at much less cost than it is to go balls deep into debt to save a few bucks on hydro. Not according to the drive god. The VP of storage after all says only go balls deep into debt on hard drives so big that no sane person would ever buy them off of you should you want to bail on this Chia project in the future.
Regarding failed drives? No this wasn’t a “Seagate” thing. It was a Seagate enterprise grade drive failure story. Not anecdotal. An experienced, long time farmer. Along with others.
People can do whatever they want. It’s only a matter of time before we find out about hard drives and Chia. The do’s and don’ts aren’t written fully yet. Maybe people will start disclosing their drive failures or warranty replacement nightmares. Or, as it’s suggested, Chia is so light on drives, we won’t hear about widespread failures until the 10 year mark.
That “self professed expert” as you put it, spent 10 years at Intel as a product expert focused exclusively on data storage devices. I’ll take his expert opinion over your armchair-FUD any day.
On a more seriousness, I don’t understand why you put so much vitriol and aggression behind your words. You would get a much more reasonable discussion out of folks here if your interactions were less inflammatory out the gate.
I’m not here to make friends. If this expert of yours is selling hard drives on the side, you don’t call that a conflict of interest? I never claim to be a expert about hard drives. I was asking for factual information, not conjecture about where Chia is headed in terms of hard drive usage. Hard drive failure. The use of consumer grade used hard drives. If you have an expert on storage as you say, then engaging with the community, being friendly, engaging with the community that might help. But who am I to say anything.
). But point taken, banking on future tech to solve a today issue isn’t trying hard enough. We’ll know in the next year or two how many bargain bin white label drives ate s#$! instead of surviving the 4 +/- years to ROI.
