Network v15 OhSnap

[jennijuju]

If you have any questions wrt the Network v15 OhSnap Upgrade and Lotus Releases Timeline, leave a comment here!

[Eliovp]

@jennijuju , I have a question.

Wouldn't it be much fairer if this only applied to new CC sectors?
Sectors generated after the upgrade.

Because right now, this is a big finger to the people that were doing real deal sealing from the start.

So again...

Wouldn't it be much fairer if this only applied to new CC sectors?
Sectors generated after the upgrade.

[stuberman]

@RobQuistNL The calculation I used is based on (130 PiB x 10 collateral) where 1PiB costs 4928 FIL per PiB = 6,406,400 FIL x $20/Fil = $128 million (I miscalculated the first time with an extra decimal point.

[jennijuju]

. I was under the impression that its just 1 single layer of PC1 and a single C2 - which is doable with a single EPYC machine with 1 GPU

@RobQuistNL the whole new proof is new thus the trusted setup, you can find more info on that spec here https://github.com/filecoin-project/FIPs/blob/master/FIPS/fip-0019.md#specification

[RobQuistNL]

Sealing 1 32 GiB sector with verified data requires 1,5 FIL though. Looks like the one I linked is indeed a 64 GB setup and not 32 so 6m fil makes sense :)

I And yeah I read the PDF too @jennijuju :)

[dkkapur]

One more thing I'd like to flag -> SPs that are just here to abuse Fil+ are unlikely to leverage it for SnapDeals for older sectors, since the max lifetime of that sector is less than the max lifetime of a net new sector, thereby having a lower expected rewards value over time. It's much more likely that they will opt for a new sector that can earn for longer for the same collateral/effort to abuse the system. We're working on getting a lot more transparency around on-chain client and SP behavior on the Fil+ side and using this to push back against clients likely to be abusing the system. @RobQuistNL @Eliovp if you're particularly interested in this -> would love to chat at an upcoming governance call about some of the efforts and get your suggestions/feedback.

[stuberman]

@RobQuistNL let me try the sector math for verified deals:
130 PiB = 32GiB x 4,259,840 sectors.
4,259,840 x 1.5 FIL each = 6,390,760 FIL x $21.20 (today's price) = $135M

Frankly, I have always seen the collateral cost as the major hurdle in sealing, the hardware and OPEX pales in comparison, especially when storing verified deals.

[Eliovp]

Unbelievable.
The naivety...

Hey, @jennijuju I bet you like this new upgrade don't you?

99% of the top 100 storage providers(garbage CC sectors) are in Asia.
Hey, look at this program here --> https://slingshot.filecoin.io/recovery
"An incentive to restore lost data (not even going into detail about how the data was lost.......)"
How many of those (from a specific part of the world) are helping out to restore the data? (You can see the countries btw, just look......)

I know I'm just wasting time here by writing this, even though I'm right, just as I was right during the space-race period, how I was also right about the Slingshot program that was being cheated as well.

Woke, Unicorns, Rainbows, the world has gone mad.

[Eliovp]

Simple solution, everyone happy.

• CC Sectors before the upgrade are not eligible for verified deals.

There, problem solved.

[momack2]

Your personal, demeaning tone here is absolutely unacceptable, @Eliovp. I don't care if you are frustrated or have differing opinions, no one should have to deal with your rudeness and harassment. Shame on you for personally attacking @jennijuju when she is just trying to answer your questions about a FIP designed and approved by the community, not her. If you can't be polite and offer opinions respectfully, get out. We don't tolerate this sort of rudeness & disrespect in the Filecoin community.

Please review the Filecoin Code of Conduct and consider this an official warning.

[Eliovp]

@momack2 , oh damn, an official warning.
Is speaking out loud, telling the truth wrong these days?

Now I'm the one that feels attacked and offended, no seriously, isn't this a decentralized open-source project?

[dkkapur]

As a sidenote, @Eliovp - participation in the Restore effort requires manual validation and selection of individuals/entities, so using this as an example of who is participating/how they are participating/what people are interested in doing on the network is a red herring.

[Eliovp]

@dkkapur Hey, don't worry about it.

Today, I officially stopped caring (again).
"Offical warning, thumbs down emoji's, telling me what I can or can't do"
Seriously?

If you read in between the lines, of everything I say, it's always because of concerns towards the project.

Anyway, I will be back again when things go sideways, just to say "Told you so" (again).

[cryptowhizzard]

Still, there is a legit worry.

A lot of CC sectors (Where ever they reside on the planet) who were created in the SR can be upgraded this way in a very short timeframe without a lot of hardware to be a verified sector. With a lot of LDN requests incoming to make verified deals this might be worth consideration how we intend to handle that because the incentive to cheat only get's larger this way.

@dkkapur

[jennijuju]

re SR sectors in question, please see #7935 (reply in thread)

[Angelo-gh3990]

I agree there are a lot of cc sectors out there, but they have a short useable life span... To short to make it worth to be used for snap deals I believe, it would be to much of a hassle to manage that over the potential benefit it will give.

I think most new cc sectors with a normal lifespan will be used for snap deals making this a very good solution to seal a deal for a customer. I'm just looking at the benefit from a customers perspective as they will have there data sealed very quick when you have cc sectors at the ready.

Some concern is around people that would do self dealing amongst themselves or with a small group of people, but they have to come up with a 10x collateral at that point to. I don't know if they will really go do that.

[dkkapur]

Thanks @cryptowhizzard - agree that it is a risk, but its lessened by sector max lifetimes + in general, I think people have been a lot more opportunistic in the last couple of months with LDNs and hoping to make the most of them. I think the FIP introducing DataCap removal + using a dealbot or something similar to start tracking retrieval attempts is a good way to start minimizing how much DataCap ends up in the wrong hands. IMO, LDNs likely need to continue moving in the direction of:

• more up front KYC from a client
• less friction to first allocation combined with better tracking of on-chain behavior post allocation
• audit tools that flag risk based on on-chain behavior to prevent future allocation
• and subsequently, removal of DataCap + banning of a client address + associated addresses in the case of abuse

cc @galen-mcandrew @Kevin-FF-USA

[NSC-FIL]

I do think Snap deals functionality will be an improvement to Filecoin and appreciate the hard work the devs and others have put into getting this launched.
But the concern that "bad actors" will self-deal and increase the network storage power and unfairly benefit from increased block rewards, while reducing the chance for other SP's is a real possibility. Is there a way we could have the the Fil+ program estimate the anticipated increase in our network storage power from verified deals before the upgrade happens? And maybe explain the precautions Fil+ has taken in anticipation of the upgrade and how penalizing "bad actors" would be enforced? It might help alleviate some concerns.

[jennijuju]

Thanks for ack the work and effort being done here!

re your concerns, they seems mostly related to FIL+ governance rather than the upgrade itself (which is what this discussion for).

This upgrade also introduce a method to remove unused DataCap from bad clients.

I think you will get better responses and discussions if you repost your questions there https://github.com/filecoin-project/notary-governance/discussions.

[rjan90]

First of all, for those voicing their wanted changes / issues about this upgrade. Those comments / thoughts should have been posted during the FIP discussion, and/or draft-process: https://github.com/filecoin-project/FIPs/discussions and not after the FIP has been accepted and on its way to be implemented.

That said, I want to weigh in on the positive side. I think the OhSnap-upgrade is a big improvement for the Filecoin-network as a whole.

• Clients data will be sealed way faster (15-30 minutes, compared to 4-6 hours in a fast sealing pipeline)
• Storage Providers can upgrade CC-sectors more easily then todays time consuming mark-for-upgrade process.

About some of the concerns:

• V1-sectors will not be eligible for SnapUpgrades
• Storage Providers has to put up 10x collateral for verified deals, so its not like a Storage Provider can get a lot of power without putting up a lot of collateral to back it.

For the concerns about malicious actors filing for FIL+ datacaps, the OhSnap-upgrade also includes the ability to revoke a verified clients datacap, which should ease some of those concerns. filecoin-project/FIPs#226.
That said, I personally think the governance process can be improved, and recommend that if you have any ideas/thoughts about how to improve, you start a FIP-discussion outlining the idea.

[Reiers]

We at TechHedge is really excited about this upgrade.
CC has served its purpose (at least for us) and this is the next step.
Cant wait to turn the random data into something useful!!

I can see that this discussion derailed a bit from general questions. So I will just add a few answers/highlights to clear things up.

• Like a few people have already mention; v1-sectors will NOT be eligible for snap-up.
• You will still need the production line of workers to do snap-up. Here are a little peek at resource table from my testing:

32GiB Sector

Sealing Pipeline	SDR	CUDA	HW	RAM usage
AP	~3 minutes	-	TR Pro 3975wx	-
RU	-	~5 minutes	FE 3090	-
PRU2	-	~10 minutes	FE 3090	~192GB

• 10x collateral for FIL+ Deals is still the case.
• DataCap removal will also be introduced in this upgrade.

[RobQuistNL]

I agree.

I'm just afraid that it will be abused.