Search the transcripts of our AMAs. We hold fortnightly sessions with senior DADI engineers and partners and take questions from the community.
What made you choose NSQ as opposed to RabbitNQ or some of the other well established messaging queues?
There were a whole host (excuse the pun) of performance factors that we evaluated during the benchmark phase, prior to selecting NSQ as the primary candidate for the messaging queue.
The network applications are built with Golang, and the fact that NSQ was one of the initial reasons it made the shortlist. RabbitNQ relies on the Erlang runtime, so another moving part outside of the core codebase language.
It also persists data when there are no connected hosts, which allows us to take the core monitoring services offline without losing valuable logs.
The topology is, by design, horizontally scalable and will cluster and scale much better than RabbitMQ. So far, it’s been exceptionally performant, and the results of our benchmarking are promising.
How far down the track do you see devices for the 'connected living room' being done and is there a focus on one in particular right now?
Joseph Denne Quite a bit to say here! The quick summary is that we're fairly advanced, but it's important to note that we're addressing this through a multi-channel strategy that will see is working with various OEMs to access existing and future hardware. The Founding Nodes are conceived as the first step towards realising the vision – a plug and play device to be easily set up in the home.
And of course, DADI technology will also make use of spare capacity in the machines already there (such as laptops or desktops). We're also working to produce more dedicated devices. (More on that in the coming months.) Plus, there's the potential to leverage capacity in businesses and offices.
You are giving a payout per hosts. Depending on the app deployed on the host, cpu & memory & bandwidth requirements per hosts will vary a lot. Are we able to set a host rich of bandwidth and low on cpu? Do you allow customisations like that and do you have plans to reward more for more intensive hosts?
The network is aware of the capacity available and the capacity committed at node level. This information is used to inform node selection per-request, as well as to structure payouts on the basis on contribution.
Will GPUs be used in nodes? If so, only in hosts or other nodes as well?
Arthur Mingard: Yes, depending on the applications that have been deployed to the Host machine. dApps that require more GPU power will prioritize distribution to machines that meet their requirements.
Do we know if setting up nodes will require high-end rigs? Or will regular VPS machines suffice?
DADI Host runs extremely well on low-power machines, and is specifically designed to do so. We’re targeting support for a wide array of devices, including mobile phones and laptops.
Gateway and Stargates require exponentially more power, as they handle far more in the network than individual Hosts.
So in answer to your question, yes, regular VPSs will be able to be used to contribute to the network.
In the event a bad actor got administrator access (sudo) to a Stargate running system, what is the worst that could happen? Will the bad actor be able to manipulate DNS request/replies? Any measures to detect this bad behaving and downrate reputation? Where can I find more about Stargates engineering?
Adam Dean Stargates monitor each other, so a single machine attack would result in the SG in question being off-lined. We take a very strict approach to security in this layer of the platform, with access only available via our secure VPN channel, and within this access is tied to named individuals by machine. In addition to this, we monitor all activity on Stargate machines, so there is no chance of anything happening without us being alerted.
Does DADI Store spread out files to hosts? Does that mean that it would be beneficial to have a lot of disk space on my hosts?
James Lambie: In the first iteration of Store, when files are added they will be spread out amongst a selection of trusted nodes and then when requested by a consumer they will be transferred from the closest of those nodes that has the data, potentially passing through a series of closer nodes where the data will be cached along the way. The amount of storage space on a node given over to the capacity requirements of Store will be configurable.
What type of hardware would be needed to support AI computation in the network in the future when a partner like Agorai launches its marketplace on the network? I assume a Raspberry Pi wouldn’t be the most optimal choice.
Arthur Mingard: It depends on the specific use case. In general, deep learning requires high specification GPU resource. As more complex and resource intensive applications are deployed into the network we’ll be introducing benchmarking tools to identify the optimum purpose of service hardware. As requirements evolve, so too will the incentive to contribute computational power in support of them.
Can I replace my Dropbox subscription with DADI soon?
DADI Store will move to Beta in the next month or so, and will be released to the network in the next few quarters. So yes.
I live in one country for 6 months and another for 6 months. I would need to move my host(s) between the two. As much as this is possible, would it be penalised to a point it would be better to have it hosted? Can this also be answered for the founding node too?
Adam Dean There is no penalty to nodes switching networks. Upon joining, nodes will connect to the most topologically and geographically relevant gateway, and will host applications that the network requires.
In a country like China where access to some webpages (e.g., crypto exchanges) or services (e.g., Telegram) is routinely blocked, do you think it’s technically possible for the government to block the function of a DADI host? I understand that DADI network is not a sensitive website or service and there is no political reason or incentive to block a host. Nonetheless, as a layman I would like to know the possibility of some scenarios.
Adam Dean While it would be possible to block the entire network, this would mean blocking all sites and services, and would require the blocking of all Stargates and Gateways, which as the network grows, will become increasingly difficult to do. It wouldn’t necessarily be possible to block individual domains though. In addition to this, as the network will exist within China as well as outside, blocking it with a national firewall wouldn’t be an option.
Isn't a distributed network more vulnerable than a single 'centralized network'? More nodes means more doors to knock on, which in turn means increasing chances to gain unauthorized access and mess things up?
Joseph Denne It's a matter of perspective: a bunch of servers in a single location represents a single building/IP subset target to break in to/attack, whereas hundreds of thousands of machines in distributed locations would require:
Knowing where those machines are
Gaining physical access to the property
Hacking into the machine
And if you manage of all this, you would find yourself with access to a fragment of a build, in an encrypted area of RAM. In short then, no. Distributed networks are inherently more secure - insanely more secure! - than centralized networks.
Elaborate on the possible second round of founding node lottery. What is the reason behind it? Is there a need for a quick controlled growth of the DADI network due to high load? Or simply to satisfy the unlucky ones from first round?
Joseph Denne: This is mainly about the large volume of applicants versus the initial number of devices we had allocated to give away. At sign off of the pre-production units we had an opportunity to extend the production run a little, so we took it – and that gives us more devices to give to our (hugely supportive and very lovely) community. Plus, of course, the more devices connected to the network, the better.
Is it likely that the minimum specs for Hosts/Gateway/Stargates might become higher in the future? Context: for those of us looking to acquire hardware.
Arthur Mingard: In the long term, specs will naturally increase in line with more demanding dApps. If you’re looking to support the network with spare computational capacity with a personal laptop or desktop, it’s more likely that you’ll upgrade the machine long before there’s a significant reduction in earning potential.
Has the increase for POS been determined and if so what is it?
James Lambie We're still working on the basis of our original PoS requirement, but that doesn't mean there won't be an increase. It comes down to supply and demand.
Can I use DADI Store as a Dropbox alternative? Will my data be safe and private? What about pricing?
James Lambie: For the first part, see previous question! Pricing information for using Store will be released in due course, but will vary based on whether it is used as a standalone service or in conjunction with other dApps.
Why do we still need computing resources based on DLT, since there is already IPFS, bigchaindb etc? How will DADI compete with these?
It’s difficult to compare. IPFS is a protocol (which we’ll actually use in some of our products), bigchaindb is a DB storage engine. DADI is a whole suite of web services or, if you like, a set of modular pieces that allow you to assemble websites and applications as complex as you like.
Building the various pieces as well as the network they run on is quite a powerful combination, and that’s what sets us apart from any existing proposition.
To your point, your examples are part of the solution, not a solution in and of itself.
Please also explain your plan for security of data on DADI Store. Thanks.
James Lambie: Store's first iteration is much like a Dropbox alternative. Your data is chunked and stored and put back together when you request it. A future iteration may allow you to give others access to view the data. In terms of security of the data on the Hosts themselves, the same is true for Store as it is for the other dApps running in the network - data is encrypted, with relevant security keys stored in allocate private regions, or secure enclaves.
What would be the best way for a contributor to contribute more than the typical if running a gateway is not possible due to hardware requirements? Will there be other ways to utilize excess tokens, say for those who have 100k+?
Joseph Denne Totally depends on your line capacity for example, but there's nothing to stop you running more Hosts. And gateways should be possible in this context - we're looking at a possible rental model for backbone devices for example.
I have just been reading through the whitepaper and looking at the API service. I have a question – how is data in the database replicated between hosts to create redundancy?
James Lambie We are using DADI Store and our FileStore data connector for API within the network. Store handles replication and edge caching. We're also playing with an alternate API data connector which leverages OrbitDB, which is a distributed database solution built on top of IPFS.
What will node payouts be?
Node payouts will vary by node type, with Stargates earning more than Gateways, which in turn more than Hosts.
There’s a balance to be struck between the on-boarding of nodes and the onboarding of customers. Overreach on the one hand and payouts will be diminished. Overreach on the other and we may not be able to meet capacity requirements.
To manage this we will be controlling rollout in the first instance — opening the gates to node contributors slowly at first and speeding up over time. Customers will be on-boarded directly as well, with online signup not due until the end of Q4.
Specifics relating to expected return on POS will be announced in early June.
Assuming the worst, how difficult would it be to migrate the DADI project to a new chain (on the spectrum of ‘relatively easy’ to ‘extremely difficult’)?
James Lambie The work required for this leans more toward the 'relatively easy' end of the spectrum. While it wouldn't be a walk in the park, we do build our products with the future in mind - and this relates not only to the chosen blockchain platform but every technology we choose. Assuming, of course, that by 'the worst' you don't mean Armageddon.
If DADI changes the staking requirements in the future due to token price increase or supply issues, will excess POS be unlocked for withdraw by existing node holders?
James Lambie: Yes, this is a likely scenario.
Will DADI have any tools to help migrate corporate websites running on legacy IIS example IIS 6?
In short, yes. we’ve been building tools that can help with this.
There are multiple aspects to a migration from an existing platform to the DADI suite of web services.
The main one is migrating the data into DADI API, which depends a lot on what type of data store the existing project uses. we built a series of tools to automate migration of data from various systems, so changes are we can provide you the tooling to ease the process.
Then there’s the template layer, which again depends on what the existing technology looks like. because DADI Web supports virtually any template engine, it may well already support your language of choice, meaning that changes required are minimal.
If it doesn’t, it’s pretty simple to build support for it