Huge amount of trash data?

[BrightSilence]

@Mitsos

Adding that capacity on the “faith” that it will be utilized, I’m sorry, but it’s a hard pass at least by me.

I suspect it will be for most node operators.

@Mitsos

This mysterious client appears to want to upload hundreds of PBs of data.

Do you have info that I don’t have?

@Roxor

When SNOs were recently asked about spare/hidden capacity we could bring online…

Right, we were asked. Not told. And the majority response seems to have been: “I’ll add capacity when it makes sense for me”.
I’m not in it for charity either. I want to see results before I invest, so Storj Labs would have to navigate within those constraints. Which is exactly what they seem to be doing with the Saltlake manoeuvres.

@Mitsos

If they see their drives getting filled, adding more drives, seeing them filled, then finding out that “yea, it’s not actually data, we need to delete that”, then see their drives lose 10TB per week and gain back 1TB for the next few months, then “oops, turns out the client doesn’t want to sign up”, then they’ll be unhappy. I don’t think there is any unreasonable assumption there.

I agree, that would suck. I also know they were talking about multiple potential clients and I know they know the risk of doing that. More likely they will keep that test data until some of those clients do sign up to prevent risking node exodus. Which would be really bad for them as well. But yes, it is a matter of trust that they will handle this well.

[Mitsos]

@Alexey

Unfortunately direct deletions are too slow for the customers (they need to contact thousands nodes over the world!), so we decided to get the request and delete the metadata on the satellites and then send BF to nodes… Sorry, it’s a business… So your node would collect that garbage and remove it when they could, not immediately as before…

I didn’t say direct deletions. I said bump the BFs up as high as they can be.

@Alexey

That’s not what’s happening. It’s not faith, we uploads data with a TTL, so it will be removed without a slow garbage collection, but on timely manner.
This is perfectly fits what our prospective customers wants to.

So you agree that all the required capacity is already available.

@Alexey

Nobody asks you to add a capacity right now, but you are informed. “He who is forewarned is forearmed.”

He who knows that the needs to add drives, will order new drives. He who hopes to fill those drives without seeing data flowing in, is doomed to go in debt.

[Alexey]

@Mitsos

I didn’t say direct deletions. I said bump the BFs up as high as they can be.

We are working on it! That’s not easy as we thought…

@Mitsos

So you agree that all the required capacity is already available.

Perhaps. I do not know for sure, but seems they need several PBs more, than we have free. At least I can imagine that because we removing unpaid data at this speed.

[Mitsos]

@BrightSilence

Do you have info that I don’t have?

No, I’m just going by the data available. Free capacity is 30PBs. If deletes and reservations need to happen, then the required capacity is higher than 30PBs. If storj is requesting all offline drives added immediately, then the required capacity if far higher.

What I’m trying to say is:

1. Add everything you have right now! = Go go go
2. Hey, heads up that we may need more capacity to come online soon = buy a couple more drives and get them online. No need for overnight shipping, but do it soon(ish).

[Alexey]

I would say, if you would bring more capacity to the network, we would get a signal, how much storage we may have. This is useful to attract even more huge customers. You always can shutdown the extra nodes which do not get enough data. After a month please.

[Roxor]

I dream of a time when I… as a humble SNO… am buying a 20TB HDD from Amazon every month… completely prepaid for by my Storj payouts… and all I ever do is slide it into a free slot in a JBOD and give it a fresh Identity. Other than that minimal work… the cash just rolls in.

For now… I’m still dreaming…

[Alexey]

This is not exactly a dream. That’s the plan. See a difference?

[littleskunk]

@Alexey

I do not know for sure, but seems they need several PBs more, than we have free. At least I can imagine that because we removing unpaid data at this speed.

Free space isn’t the problem. Last time I checked we had about 50 PB of free space and I think we are cleaning up additional 20 PB or so (I don’t know the actual numbers. Don’t quote me on this one.). So there should be enough capacity. The bigger question is how much bandwidth do we have at the end? How many nodes will still have free space and a fast enough internet connect? More nodes = less bandwidth consumption per node.

And I would prefer to end the upload cylce with a decent number of nodes so that we can tell all customers to keep increase their demand. We have to factor that future increase up front as well.

[Mitsos]

That’s not a dream or a plan, that’s a reality

Well at current (and projected) payouts every other month to be honest.

20TBs Ultrastars.

[Alexey]

We will see in the future.

[Toyoo]

@unrealSpeedy

My Boss would Kick me in the Ass if i would Invest Hours of working before the contract is signed.

Please consider that we don’t know almost any details about these prospective customers. In the places I worked in, there could be 2-3 contracts staged: one for preliminary work/feasibility study, another for onboarding, and the third one for long-term planning. What we observe might just be this first one—at least I hope so.

And nobody will share details before all stages get signed in fear the customer will go away.

@BrightSilence

In my experience, ALL providers lie through their teeth when responding to RFP’s and then rush to get close to what you require in the mean time.

Can confirm. But don’t worry, customers also lie, so it’s balanced.

@Alexey

So, you did not read ToS too and blindly accepted them?

Is it so surprising? Wars wouldn’t happen if people read more.

@BrightSilence

@Mitsos

This mysterious client appears to want to upload hundreds of PBs of data.

Do you have info that I don’t have?

They’re exagerrating for dramatism, like a true provider. They learn quickly.

This post has already suggested it’s less.

[BrightSilence]

@Toyoo

@BrightSilence

In my experience, ALL providers lie through their teeth when responding to RFP’s and then rush to get close to what you require in the mean time.

Can confirm. But don’t worry, customers also lie, so it’s balanced.

Guilty as charged.

[snorkel]

If you read ToS and ignore them, why waste time?
And yeah, we all have ignore them by running more than one node per IP.
(Spoiler: I just read part of ToS for the first time)

[pangolin]

The only purpose of TOS is to make it harder for other people to sue you. I don’t plan to sue storj, so why reading their TOS?

[nerdatwork]

@pangolin

The only purpose of TOS is to make it harder for other people to sue you.

That’s not the only purpose. Users could be banned for breaking TOS. Hackers could be prosecuted for breaking TOS combined with relevant local cyber laws.

[arrogantrabbit]

You must read and accept terms and conditions from Terms and conditions and if you continue, it will be treated as you accepted them. Please read carefuly before continue.

I would rephrase it and fix typos:

Proceeding constitutes acceptance of our Supplier Terms and Conditions. Please read carefully before continuing.

[ACarneiro]

I could swear they were going to remove the 1-Node-per-IP item from the ToS (this has been discussed a while ago, if memory serves right).

I admit I’m falling foul of that particular item as well…

[arrogantrabbit]

@pangolin

I don’t plan to sue storj, so why reading their TOS?

Why learn traffic rules, I’m not going to sue the driver I’ll crash into.

TOS is an agreement and terms of engagement between you and the company. Why would you choose to engage without knowing what you are getting into?

[Mitsos]

@littleskunk

Free space isn’t the problem. Last time I checked we had about 50 PB of free space and I think we are cleaning up additional 20 PB or so (I don’t know the actual numbers. Don’t quote me on this one.). So there should be enough capacity. The bigger question is how much bandwidth do we have at the end? How many nodes will still have free space and a fast enough internet connect? More nodes = less bandwidth consumption per node.

I’d like to add: more nodes = less bandwidth consumption per node, but still equal overall (/per SNO). If the required capacity is there, then the bandwidth benchmarks should show a pretty accurate network picture already.

But I can’t say that without backing it up, so here we go: I’ll use EU1 since that gives data currently and is a satellite most shouldn’t have GE’d from. There are currently ~4500 active wallets. I am assuming that nobody broke ToS by using multiple wallets (even though you can use a node for storj payments if you use storj as a client as well, I’ll attribute this to the “one node per IP” dogma. It’s there, but its not actually there).

Those ~4500 active wallets have ~22000 active nodes (not counting suspended, disqualified and so on). It doesn’t take a PhD to figure out that most SNOs are running multiple nodes. If they have the know-how to run those nodes, they have the know-how to add more nodes as fast as they can afford buying drives.

There are currently ~4400 full nodes. So 20% of the nodes are currently full. 4 out of 5 nodes can pretty accurately predict the overall network bandwidth. If the current bandwidth isn’t suitable for the prospective clients, then adding that 20% back is pretty doubtful to be suitable as well. Call me crazy, but I don’t personally think that all 20% of that is on 10Gbps lines.

Which brings us to the “use what you have”. We’ve established that the capacity is there. We’ve established that some expansion may be needed (IF the demand is there, not before). We are just missing the bandwidth. So was this just a “get a second internet connection, spin a node up and we’ll use it someday?”

[Roxor]

@Mitsos

That’s not a dream or a plan, that’s a reality

Well at current (and projected) payouts every other month to be honest.

20TBs Ultrastars.

So jealous! It would be so nice to have a few full drives that finish paying for themselves… then start to help fund expansion.

Would it be a reasonable guess that a new 20TB (starting with a pre-vetted/post-holdback-period Identity) would take about 2 years to fill? I’ve heard of 8TBs getting filled in a year…