Topic: Sia - Siafund Redemption Deadline: June 1st, 2015 - page 18. (Read 68779 times)

I'm happy to announce that we are starting to implement one of my favorite features: wallet-based approval of transactions instead of signature based approval of transactions. This sort of works like a script.

Wallets will have balances, scripts, and a way to interact with a script. This script will have the power to submit updates and transactions (such as allocating storage or editing a file, or of course sending money to another wallet). The rules of the output depend entirely upon what script is loaded into the wallet.

The typical signature-based-transaction behaviour can be simulated by having a public key in the data field (or several public keys in the data field. Or pointers to files that have whole lists of public keys... you get the point) and verifying an input+signature against the public key.

The script is going to be a full turing machine, and Luke (nemo) has started working on the details of the script. It's going to be a bytecode where each instruction has costs some volume of money (to discourage long scripts) and a timeout, in case an infinite loop occurs or more computation is allocated than the quorum has the ability to compute against. The script will also have access to higher level functions such as 'verify signature', that way verifications don't need to be written by-hand using the bytecoin.

This is all a ways away from being implemented but we've been making good progress.

That's a good suggestion. It should be relatively straightforward to implement scheduled rate-limiting; any half-decent torrent client does. Of course, you'll have to maintain some minimum bandwidth at all times.

I guess that the participants can also back up their own file by themselves to make an extra layer of safety.

Yes, one server can host many participants, each at 64GB (revised from the original 8GB). It's expected that most servers will have on the order of dozens of participants, and we're trying to write the program such that a modest machine (maybe $300) can handle all of the computational power required to operate ~20-50 TB. These are all vague estimates at this point.


So, this has a lot of answers. Quorum-important data like error-detection and wallet info, as well as the block history is either stored at full redundancy (everyone in the quorum has a full copy) or at a redundancy of 8. So wallet data and other equally important data will only be lost if 112 out of 128 participants go offline at the same time. This gives us extremely high reliability, even when the assumption is that 50% of the network is dishonest.

Files on the other hand get to choose their redundancy. If you are storing extremely important documents (like your private keys, for example), you'll want to choose the maximum grade redundancy. This comes at a cost, instead of paying $5 per TB per month, you'll be paying $50 per TB per month. But if you're only storing 100kb at that price, who cares? For everyday files, you'll want to pick a more reasonable redundancy, like 100 out of 128, which means that your file will only be lost if 28 participants disappear at the same time. Assuming no malice and reasonable uptime (like 98%) per participant, it's extremely unlikely that you'd lose the file. We let users pick their redundancy because everyone is going to have different levels of paranoia/trust in the system, and we don't want to force one person to pay what they consider to be 10% too much and force another person to have redundancy that they worry is too little.

When one or more participants go offline, the network immediately replaces them with new participants. This involves having the new participants download the files and error-correct them and consumes a lot of bandwidth. This can be optimized in a few ways and we're exploring the best way to handle this. Overall, replacing a lost participant should take something in the 6-8hr range, and even less if the new guy has lots of bandwidth. That's why we set the default at 100 non-redunant participants. Assuming 98% uptime, it's going to be extremely rare that 28 out of 128 go offline in the same 8 hour window.

If the network is malicious though, and 50% of participants are cooperating to take down a file, they can all go offline together and a 99.9% uptime wouldn't save the file. If you're paranoid about your file being attacked, you should use the 8 redundancy that the network as a whole uses. (16 non-redundant participants)

Can one computer/server have two hosts/participants? For a quorum, is it true that the only time for a quorum to be down is the time that the quorum loses all 28 redundancies and one or more of the other 100 participants are down? How does a quorum or host get recovered from catastrophes?

Thanks a lot. I already have in mind what I want to do with the NAS. As I've mentioned before (or perhaps I haven't) I'll be taking advantage of my offices 100/100 Mbps fibre optics. Unethical I know but I'm the network admin there so I do what I want.

With the client, would we be able control how much bandwidth you are providing to the network? I was thinking it would be a good idea if you are able to set a schedule on the times you are able to provide maximum bandwidth to the SIA network. For example me setting a cap of 100/100 Mbps between 5pm and 8am (Non work hours) and a cap of 20/20 Mbps between 8am and 5pm (work hours).

So, we don't yet have any hardware recommendations, but when the time is closer to buy bulk storage for Sia we'll be doing some research and testing and sharing some guides. At the very least I think you'd want RAID 5, though ZFS may end up being more efficient. When one of your hard drives fail on Sia, you can repair it using Sia but that's going to require downloading an enormous amount of data (for each TB lost you'll need to pay for 4-8TB of network traffic. You'll only end up downloading about 1TB but that's because other nodes on the network are helping you move the process along. The numbers aren't fully worked out but they're not friendly). Because you can expect hard drives to fail frequently, you should have a local backup to save yourself the bandwidth. But because Sia is also doing redundancy, you'll only need 99% uptime instead of 99.999...% uptime.

You shouldn't need a ton of RAM. Instead you'll end up storing most of what would be in RAM on-disk. We'll be optimizing rigs so that the most expensive component is by far the storage and not the processor or case or anything else.

There will be a 10mpbs connection speed requirement for all participants. Each participant will need 10mpbs in bursts, so one participant requires a 10mbps connection (even if you only contribute 64GB). Because the network randomizes these bursts though (and they are infrequent), a single 10mbps connection should be able to support storage of several TB.

But again we'll write many pages of documentation on this so a newbie to Sia can catch up relatively easy and feel like they are optimizing the amount of money they make.

Thanks for the reply.

Do you have any hardware and raid specifications that would best suited for the sia network? I was thinking of using freeNAS and using raid-z (ZFS),  raid 5 or raid 0 + 1 http://doc.freenas.org/index.php/Hardware_Recommendations

It's not highly sophisticated but it's easy enough to set up for a beginner and from what I've researched most FREENAS users use the zfs raid configuration that would require minimum 8gb ECC ram + 1gb for every TB

Yes, early adopters will likely see greater returns. Because the cost of storage is managed by a supply-and-demand algorithm, storage becomes cheaper as more people join the network. Over time, providers' "mining rigs" will become more sophisticated (just as we've seen with Bitcoin), so you'll want to increase your margin as much as possible by using a dedicated NAS, RAID array, etc.

Additionally, the value of a siacoin will increase alongside adoption, so you stand to benefit by not cashing out immediately. Long-term, the inflationary nature of the currency will deter people from holding on to siacoins.

As someone who wants to start hosting from the beginning. Is it likely that my beginning storage machines/servers would have to increase storage capacity to be able keep up with increasing adoption of Sia? Would I be able to be compensated for my beginning investment of hardware as a host early enough so that I can start increasing storage capacity through the sia coins mined and compensated for being a host? I know it sounds greedy, but it'd be great to know how much of a benefit it would be for being an early host.

We have a lot of ambition and believe that Sia has the potential to be the standard for all cloud-based storage. It's very competitive, reliable, cheap, fast, and it's decentralized. Various estimates place the total cloud storage market around $20 billion annually by 2018, and I think that Sia can be a huge chunk of that. 2018 I think is the earliest that we'd start to see mass-adoption, and then by 2020 we could probably be the biggest player on the field.

It's a long timeline.


That's true, but I would be very disappointed if Sia only managed a $30m valuation at peak. That would mean that the storage never really took off. If people actually started using Sia for their high-priority files, $250 million is the lowest value that I could imagine. We're here to play the long game, and we want to become a vital component of everyday internet infrastructure. If we could pull that off, there would be a lot of money involved.


I don't think undercutting is going to be an issue, but the Siastock that's already been sold will be available to anyone who's willing to pay whatever the sellers are selling at. We'll have Sia up and running before we raise our next round, which means we'll be converting Sianotes into Siastock and then selling Siastock directly. We'll have a better grasp of the mechanics of that though when we're closer to that point.


Sia isn't meant to be used directly as your primary tool for transferring money. But you'll still need Siacoin (a currency) to access storage, and the Sia network will need a simple and trust-free way to distribute fees among the holders of Siastock. I don't think at that point the Nxt AE makes sense anymore. We'll know a lot better when we're further along.

I think using the Nxt AE for Siastock even after the IPO would increase exposure for both Nxt and Sia .. Sia isn't going to be used as a currency, It's going to be a much needed storage service for people who care about encryption and privacy. There's already a big community for Nxt, might as well take advantage of  that.

The market will decide the Siastock and Sianote price. In the next run of fundraising, if Sia shows enough progress, the price of Sianotes will increase. It would be stupid for the existing investors to sell low. For instance, during the first run of the Sia pre-IPO, the individual sales prices are always higher than the issuer's sales price. Sia's progress will be reflected on the price of Sianotes on the Nxt AE. Let's watch.

I would assume that they will not use Nxt AE as Siastock will be ready to be issued directly to investors at that point.

When you are ready to raise money for the real ipo which you have mentioned should be at least 10x more expensive are you going to continue

using the NXT asset exchange under the same asset? If you do so only now asking for a much higher price than the pre ipo you will have pre ipo asset

holders undercutting your price to make their profit and you will not raise as much. Can you explain what your plan for this is?

This would be possible if Sia becomes the standard for decentralized crypto storage.  It will take years.

even if sia coin just managed to become a 30m $ USD market cap coin .... it is still going to be 30x from ipo price.

Wow, that is a hefty estimate. Then again I really don't have the competence to judge it wrong or right. But what kind of timeframe do you think this could be accomplished in, potentially?

The network does not have to be decentralized, though I think that there are substantial advantages to it being decentralized, from a societal perspective even if that's not what's best for the company. Imagine if the internet were fully centralized, where you had to go through some central server and pay some fee in order to host your own website. It would have killed a lot of random quirky things that ended up being very beautiful.

Siastock gives us the potential to have tens of millions and even hundreds of millions of dollars in annual revenue. For a single product, this is plenty for me.

@Taek - does Sia network has to be decentralized? It seems the services that Sia provide do not require Sia network to be decentralized as long as the network is encrypted. The network is encrypted and distributed but the company can be an organized legal entity