Network Performance and Vault Profitability


#1

Step-by-step: the road to Fleming, 2: Sybil resilience

The SAFE Network only accepts new Vaults when there is a capacity deficiency. This means a user can’t just create 10 million vaults and connect to the Network. Clearly, this needs to be balanced against the fact that we don’t want to prevent home users from joining.

It’s rare for Maidsafe to acknowledge, that the utilisation of data center architecture will form the basis of the Safenetwork…at least for x years.

Autonomous in operation, clustered in physical distribution.

Home vaults add resiliency, but do they contribute to throughput in any significant way (without local mesh/ peering)

Are home vaults effectively a way of supporting the idea, but at a disadvantage, if entering the lottery is the primary motive of running a vault.

When I run the executable for the first time, should I be presented with the option to create a local vault and/or a provisioned vault (for a fee) and have a reasonable understanding of rationale.

Or is the profitable (latency wins) farming aspect of the network, to be realised only by investors (pooled capital) and those in the know (like the priests of old, who knew the movements of the sun)


Big corporations running farming nodes
The SAFE Network Primer: An Introduction to the SAFE Network
#2

‘Provisioning API’ to facilitate a marketplace of paid instances? Network pushes images to infrastructure providers, who wrap, stack and deploy.


#3

Hey stranger, nice to hear from you. So debate (healthy and wise). We have not finalised the throttle here but recognise it’s a powerful weapon, but I am not seeing the data centre being advantaged here? My view (please do debate) is that if vaults are waiting to join then

  1. It’s a home node, not costing anything
  2. It’s a data centre node, costing something

So continual try to connect and see if the network wants you will be fine for home users, they have no cost and try later or just want on line if their computer is on line. Data centre dudes though are paying to try.

Yes, with p2p-quic and protocols like that they will provide very nice storage and delivery of chunks for sure.

The latency wins is not defined yet, I am still unconvinced it is the best measure and node age gives us options that are nice here (are you a good dude but a wee bit slow, but reliable etc.)

So please do keep poking any situations where you think data centres are benefited (BTW I am not saying I won this debate here :wink: plus it’s not my decision, but I do argue a lot in house to make sure people win, not corporate data centres)


#4

Haha, I was just trying to poke at this subject on the dev update thread, but decided it wasn’t the right place, and what I had to say wasn’t very relevant!

I think I disagree with that assessment of the relative cost to an individual user and a data centre. I’d love to be wrong, but my feeling is that economies of scale will always win, at least in an essentially free market. There would certainly be some poetic justice though if SAFE could be the thing to finally turn the tide that Adam Smith set in motion!

It would be great to find a way of actively weighting the odds more in the home users favour, but it does seem very hard within the constraints of maintaining privacy. In some ways I’m always inspired by low tech things like the current system of gaining trust, just things that are more costly at scale than for an individual.

My pessimism is coming from someone who’s computer is constantly reminding them they have less than 10 GB of storage left though!


#5

Thanks David,

we don’t want to prevent home users from joining

I mean really, doesn’t this just state a fact that data centers are dominant and if we need extra capacity, then home vaults should not be excluded. Maybe the author does not have the full picture as yet.

This is my main point, to embrace reality and state that yes…data centers will be dominant for a period until maturity. Currently DevOps; NoOPs programme on ‘cloud’ API’s and this is a major source of revenue for the data center/ app model. As Safe takes over and the API programming is done on Safe, the Data Center becomes only a utility, loses profitability, charges more for provisioning, home vaults become more viable etc…

But in the meantime, is it not right and just and…to offer the facility of ‘cloud’ provisioning’ to the average user, who might want to deploy a hundred vaults for a calculated gamble. Of course without leaving on the home computer, basically no chance of winning Safecoin and so thats not a ‘no cost’ proposition.

i.e the average user understands the farming concept, but wants a simple way to have a crack with an always on instance…why not make it super simple with a Provisioning API built in, rather than a whole new learning curve, maintenance etc.

By having built in provisioning, this virtually eliminates any advatage the big money/ priestly knowledge have, because 7 billion people know how to click a button and pay some Safecoin for the convience of a greater lottery chance.

Maybe 5G, if it doesn’t kill us first (and with the will of the people) could be a mesh that brings Safe to full potential.

Yes, with p2p-quic and protocols like that they will provide very nice storage and delivery of chunks for sure.

This is nice to know as local vaults will dominate…one day :slight_smile:

The latency wins is not defined yet, I am still unconvinced it is the best measure and node age gives us options that are nice here (are you a good dude but a wee bit slow, but reliable etc.)

Very important decision that will become clearer after Fleming results I’d guess…but if all nodes had close to the same chance of entering the lottery, thats ideal for sure.


#6

It just means that if the section is not wanting vaults then any user trying to join will not be able and prevented.

Then to add focus the home user is the preferred one

The home users have the economics of scale on their side. Its like the billions of fish contrasted against the much fewer whales. The fish need little to survive and eat what is in free abundance.


#7

But in the beginning there won’t be billions of fish. The whales could be dominant for some time. Extra protection to guard against unwelcome influence of the whales will be needed it seems to me during the early stages of the network.


#8

I am pretty sure that early on data centre vaults will be playing a role in the uncertain territory of home vaults. They will provide a level of stability, but may not be making their owners much safecoin compared to the stable home users.

I say uncertain territory early on because of the smaller numbers and widely varying conditions. But as numbers increase then there will be more certainty in that there will be enough users with the bandwidth/reliability to become those huge numbers of fish compared to the whales. If I were to take myself as an example, my connection is soon to be upgraded to 20 times the upload speeds and will enable me to use my home computer as a vault on the community network which has very high bandwidth cutoff for joining.

Also if you look at the community network, there is not much success in getting people to expend money to pay for data centre resources month after month. And this is from the core group of this forum who arguably would be willing to pay something to be a part of the SAFE network. Yet we don’t see much of that. What will happen with the global network, will we see people expending money for data centre resources for a much smaller return than using their own home resources?

Its an interesting social question, and I’d argue it is as much about people’s willingness to expend the effort to maintain (and pay for) data centre vault as opposed to a home vault.

  • for the technology minded and capable people they *may* use data centres
    • but quite possibly run a home vault too if running data centre vault(s)
    • otherwise definitely running a home vault if bandwidth allows
  • for the technology minded and not capable, they *may* use a collective farming vaults in data centres managed by some smarter people.
    • but would expect them to run a home vault too.
  • for the others they *may* set up a home vault if bandwidth allows. They would follow a 2 step instructions to download/install, and set wallet address.
  • and majority of people will just surf and use the network

#9

How can the Network know I am a home user?


#10

The person who wrote the article knew though.

It was saying they don’t want to prevent home users (and all users) from joining, the important focus is on the home user. Not that they want to exclude all others.

So on one one hand you cannot just accept any vault wanting to join otherwise you could get a huge glut of space. But then you don’t want to stop too many otherwise too many home users (and all others) will be stopped and discourage home users. Home users are more likely to be discourage than institutional. So they focus the talk on the home user as the most likely to be discouraged and the desire to have as many as possible home vaults

To answer your question. The network will discourage institutional data centre vaults because of the cost of running the data centre. There is the method. Design the rewards to encourage home vaults.

  • Have you worked out the cost to run a vault with 2mbits/sec of continuous bandwidth 24/7 in a data centre?
  • Cost to the home user on Optus or TPG (NBN) or other unlimited quota is the extra electricity for the activity (mere cents). (Yes I get no complaints from TB(s) per month)

While you bring up good points well worth discussing, i thought you took the original statement from the OP as saying something it did not say.


#11

Getting G.fast at my location, which has tremendous capacity for throughput with a software upgrade. The NBN CEO states the upgrade will not be enabled unless there is a demand for it. I asked how I can register my demand, they have no such database :roll_eyes:

Also if you look at the community network, there is not much success in getting people to expend money to pay for data centre resources month after month. And this is from the core group of this forum who arguably would be willing to pay something to be a part of the SAFE network. Yet we don’t see much of that. What will happen with the global network, will we see people expending money for data centre resources for a much smaller return than using their own home resources?

Wouldn’t put much weight on this myself, I dont think you have anywhere near the eyes on this forum after 5 years of v2 development.

Its an interesting social question, and I’d argue it is as much about people’s willingness to expend the effort to maintain (and pay for) data centre vault as opposed to a home vault.

That’s the key point for me, if you make it easy by:

  • (provisioning API)
  • give choice (market place of plans from API)
  • explain the odds (level playing feild)
  • compare the cost ($ kwh widget + additional disk space if required; etc)

Personally I would go with the provisioned option, which should be very keenly priced, given the scale, repeatability and vetted image and I would suggest this is the fastest way to growth.


#12

Cost-effective corporate vaults would not be running in these environments, they would be operating in areas where Verizon FIOS (or comparable) download and upload speeds of 900mbps are available. Those are the installations that warrant special consideration and pose potential problems for the home vs. data centre equation, at least in the early stages of Safe Network.


#13

But it still costs to run a machine doesn’t it? If I could get a data centra machine (or VM) with at least 2mb/s bandwidth for a few cents per week or month then I’d like to know for my own use. Never buy a home PC again, just a tablet and teamview into the VM


#14

Going rate in some areas for Verizon FIOS is about $215 per month for 940mbps down/880mbps up. That works out to about 50 cents per month for 2mb/s. Not free but could be very cost-effective for large-scale farming operation(s).

P.S. Birthday cake was yummy. Thanks, Discourse.


#15

Thats no good for calculations. I am after cost for the machine, not per 2mbit/s (wanted at least 2mb/s). So thats 215$ per month for the machine or a VM on a machine ??

For vaults then I estimate you can only have a limited number of vaults running on any one machine. Lets say the realistic number is 8 to 16 vaults on a high performance 4 core machine

Now for argument sake its 64 vaults, then that equates to $3.36 per month. Not as good as the 50 cents when just looking at bandwidth.

Of course that is unrealistic and 32 poorly performing vaults is more likely or 16 reasonably to good performing vaults, It looks like $6.50 to $13.00 per month per vault is more realistic for that machine if it is a machine and not a VM

Now what about storage? If your vaults are to be 2 to 10TB in size then you need the storage for all vaults. At 64 its a minimum of 128TB and if 16 vaults its 32TB. What is that going to cost?? Especially if you want to be big (in storage) vaults.


#16

I heard that 5G will be extremely fast both in terms of bandwidth and in terms of latency. And the 5G transceivers have to be installed very densely because of the short reach of the radio waves with short wavelength. That will potentially be excellent for the SAFE network.

The prediction is that it will be extremely costly to install the 5G infrastructure. But with the huge potential market for the Internet of Things (IoT) big corporations other than the usual telecom companies may be interested in launching 5G as early as possible in order to be able to launch their IoT platforms and services.

And if so, those giant companies, such as Google and Huawei will pay for much of the 5G infrastructure, and then they would also benefit greatly by running their services on the SAFE network.

This, then, means, that the big tech companies will not only gladly pay for 5G infrastructure but also invest a lot in running SAFE farming nodes! To ensure that the SAFE network is and will remain robust enough to support the IoT explosion.


#17

As a side note: Actually, even the 5G transceivers could be seen as nodes with proof of resource! That would eliminate all those ridiculously siloed telecom companies sitting like little totalitarian emperor middlemen on top of the network.


#18

BitTorrent a dominantly home “vault” p2p protocol is consistently clocked at around 20% of internet traffic year after year (according to Sandvine Global Internet Phenomena Report and other studies). This 2014 study “An Analysis of BitTorrent Cross-Swarm Peer Participation and Geolocational Distribution” shed some more light on network participation by home connections. Despite the problems associated with home internet asymmetric connections (which that study suggests solving with more local mesh/peering), BitTorrent works exceptionally well. If Safe Vaults accounting for some percentage of dynamic content + static content caching approach even some fraction of BitTorrent in performance then data centers may not get much of a look-in?

My offtopic question: Will it be possible to financially reward caching to better encourage BitTorrent level of performance?


#19

I would say forever, due to

  1. the Pareto principle
  2. if there is a reliable storage network, that can be used to store your data, then there is less need to buy local drives and pay more for your PC, there is no “free storage” for home users (even if @neo is saying otherwise).

#20

For the foreseeable future the need to have a drive in your PC will remain. The OS needs to be stored somewhere and SAFE is simply too slow for the OS. And the small drive sizes are growing way faster than the OS needs.

So all that spare drive space on your primary drive might as well be used for the network.

And yes you could have a 16GB core OS stored in a SSD chip and pull the rest in some form. But that is not going to fly in a practical sense. And with SSD drives exploding in avail space I’d say that your PC will nearly always come with plenty of spare drive space to share on the network. In fact it may make it easier and more desirable when you consider such things as 5G

My thoughts is that there will always be a mix of data centre and home vaults. But as the costs of datacentre vaults will be higher than home vault using spare storage on the primary drive, then I’d say the mix will be healthy as a gut feeling