AMA recap: March 29th, 2019

Friday saw us host another full on AMA, with over a 50 questions submitted. We got through around two thirds of them in the in the two hour session. Those not covered this time will be included in the next session.

Lead article image

🔗Question 1

🔗Why does the decentralised website running on dadi always resolve to the same IP Address? If that ip address would be bombarded/attacked/go down, the website would go down. Could you provide a technical explanation of how the decentralised websites work?

Joseph Denne: It doesn’t. It runs from multiple IPs. See: https://geopeeker.com And if one node goes down, another takes over, a process that is almost instant. We use BGP routing at the top level of the network to provide DNS (via Stargates). Stargates then route traffic to the most appropriate Gateway, which are the entrance point for requests to apps in the network. Gateways maintain a job queue of requests. Hosts connect in to their nearest Gateway and dequeue requests, returning results that are then handed back to the end user.

Arthur Mingard: Requests are load balanced to Gateways based on load and location. If you’re seeing a single IP return for all requests, this is because that is the closest Gateway to you. As the TTL expires on the DNS request, you may or may not receive a different IP based on how many Gateways are operating in near proximity and whether you are reassigned to the same Gateway.

🔗Question 2

🔗Is DADI aspiring to obtain certain ISO or other quality certifications in the (near) future? Which ones?

Joseph Denne: We have none in plan for the near term, but are investigating how edge networks could fit with established cloud certifications. ISO/IEC 27017:2015 is the most relevant from a security perspective.

🔗Question 3

🔗When self-onboarding hosts are live, will there be a benchmark/compatibility test as part of the configuration? Example: I connect my tv decoder to the network, it appears in dashboard, its status is turning orange (testmode), DADI sends traffic to device to generate a (internal) device-score. (read/write speed, network bandwidth…)

Arthur Mingard: Benchmarking will be carried out during periods of low load, and performance stats will be used to measure real performance under peak load. There are a number of measures of performance that relate to bandwidth, computational load and peer review of similar or identical tasks.

🔗Question 4

🔗What does Dadi define as a successful exit?

Joseph Denne: For the business? We’re working towards full decentralisation. Ultimately that means cutting ourselves out of the day to day running of the business as a corporate entity. Even if we can get 80% of the way there (the rest being held back by ongoing expectations of industry to be able to interface with a business for example), we’d be very happy, and would consider that a success.
Traditionally an exit looks like a buy out. We’re not targeting this, but any third party that wants to help us on the road to meeting our goals would be considered on the basis of its own merits: what would their involvement bring to the ecosystem and how would it help to drive it forward?

🔗Question 5

🔗From looking in to Consul (used for service discovery in the network), I learned that Consul has a consensus algo ‘based on raft’. Is this the version of the consensus byzantine fault tolerant? From StackOverflow: ‘Image a node that votes twice in a given term, or votes for another node that has a log which is not up-to-date like its own and that node becomes leader. Such behaviour could cause split-brains (case where two nodes believing themselves to be leader) or inconsistencies in the log.’ Wouldn’t this give potential problems to keep the network fair?

Arthur Mingard: The log and leadership election require consensus and each election process is does not allow double voting. As Hashicorp say, “There is no risk of a split-brain since the old leader will be unable to commit new logs”.

🔗Question 6

🔗What are web development trends for 2019 according to DADI?

Adam K Dean: We’re seeing a big movement towards the JAM stack, and to headless platforms in general. As websites tend to become web applications, with more work happening on the client, we’re interested in an architecture whereby we deliver static payloads on the edge (media assets + application shell) and then handle subsequent updates via API endpoints, which plays nicely with the microservices suite we’ve been developing and advocating for years.

🔗Question 7

🔗How do I know the traffic going to nodes is actual traffic from paying customers and not some generated traffic simulation?

Joseph Denne: You don’t. And that’s by design. Traffic in the network is anonymised as part of the approach to security. We are however working on adding header responses to requests which will expose the nodes that contributed to the response. This is already in place for Hosts for non-cached requests.

Adam K Dean: On top of that, nodes don’t directly expose an HTTP server, meaning it is not possible to directly query them. All traffic handled by your device comes from a real HTTP request that the Host reads from a secure queue.

🔗Question 8

🔗It might be a bit early to know, but since article 13 passed in the EU yesterday, forcing website owners to check content for copyright violations before it is posted, how will this affect DADI? Could DADI be a way to run websites without upload filters, or would DADI make themselves liable to EU-fines by allowing this?

Joseph Denne: It’s absolutely too early to know for sure, as we’ve yet to receive a full legal opinion on this front but our broad understanding is that individuals and businesses using the network will need to take responsibility for article 13. violations themselves, as we will not have any access to their deployed containers. Once we receive the full opinion, we’ll share it with the community.

🔗Question 9

🔗Is DADI planning on having another community for support once self-onboarding (B2C) start? I see more and more projects using a Discours or SpectrumChat type solution for community engagement platform.

Adam K Dean: 100%. We’ll open a new channel for this on Discord.

🔗Question 10

🔗Any plans for a behind the scenes deep-dive series on the network/or any learning on decentralized network development for us techies out there?

Joseph Denne: Yes! A lot coming on this front. Honestly there has been so much that we’ve wanted to share, but we’ve been so focused on developing that we’ve had no time to stop and write about what we’ve been up to. This is changing – we’re carving out a bit of time each week to write about the networking tech in detail. And the first results of this have been released this week:
https://dadi.cloud/en/knowledge/network/scaling-the-edge/
https://dadi.cloud/en/knowledge/network/edge-storage-deep-dive/

I’ve also been taking a cut of our internal weekly roundups and sharing them directly with the community here on Telegram, and in Discord. These are now also being published on site. You can see copies of the first few here:

https://dadi.cloud/en/updates/announcements/weekly-update-08th-march-2019/
+
https://dadi.cloud/en/updates/announcements/weekly-update-15th-march-2019/
+
https://dadi.cloud/en/updates/announcements/weekly-update-22nd-march-2019/

🔗Question 11

🔗Do you have an idea how to combat bad bandwidth nodes for storage? Is the minimum bandwidth enforced, and dropping below not accepted for a longer time? Or will it function like a CDN, that nodes closer to your location will be the hosts/servers of your data?

Arthur Mingard: Nodes are subject to regular health checks and any that fail are removed from rotation until their stability improves. There are TTLs in place for reservation in the queue which are effective in throttling Hosts with poor response times. Gateway does prioritise based on network topology so nearby Hosts will get elevated dequeuing privileges.

🔗Question 12

🔗I wonder why DADI is focused on AI tools for GPU usage – there are so many more use cases for them. For example, I see a lot of projects also support rendering. Can I ask why DADI is not also having a part focus on rendering?

Adam K Dean: Well, we’re focused on the provision of edge compute units—units of computational power that map back to different scales and types of capacity (RAM focused, GPU focused, CPU focused, etc)—rather than on the provision of tools on top of the network. What this means is that if you want to create a rendering service on top of the network, you can, and you won’t face direct competition from us.

🔗Question 13

🔗Is the networking code written in Node.JS?

Arthur Mingard: The first prototype of Gateway and Host were however, shortly after the crowd sale ended, we retooled to use Go. This with due to a trifecta benefits: performance, stability and footprint. These qualities were particularly significant for Host which is designed to perform well on a wide variety of hardware specs, down to low powered smart devices where those factors are magnified.

🔗Question 14

🔗You had a very descriptive article about edge storage this week, but when can we expect to see some actual code? This ‘patenting’ phase an 0 code in over a year is concerning.

Arthur Mingard: A lot of decisions around open sourcing relate to either the sensitive nature of the code, or the frequency in which R&D lead us to an alternate approach. An example would be the post-crowdsale switch from using NodeJS to using Go, where the source repository ceased to be maintained. We are looking at rolling out some of the core packages we’ve been working on shortly and will confirm dates in due course.

Joseph Denne: And we’ll go full Open Source as soon as we’re happy that the core IP is protected. (Note that this is not a process with a clearly defined end date.) In the meantime, we’re looking at engaging a third party to provide a code quality review, specifically to be shared with the community in response to this question. I expect this to happen in the next couple of weeks.

🔗Question 15

🔗Do you have an estimation when you will work on gpu implementation to the network, and is there already some work done for this?

Arthur Mingard: GPU container types will be explored and made available after the first iteration of self on-boarding has been rolled out.

🔗Question 16

🔗Why is DADI a complete loner in the crypto space, why are you not putting more effort in integrating into the crypto, but especially the ethereum community?

Adam K Dean: I wouldn’t say we’re a complete loner. But as with any business, our focus is first and foremost on building out our own proposition. In time we’ll be in a position to contribute more back to those projects that we lean on (such as Ethereum), but we’re not there yet.

🔗Question 17

🔗What can the DADI team do to increase transparency towards its community and be more active, not just every three weeks or so, otherwise complete darkness?

Joseph Denne: We’re one of the most open and transparent teams in this space. Regular AMAs, team members participating in key channels, a packed publishing schedule and an open door to the CEO if you want to reach out. Seriously, my partners and I are across pretty much all of our channels on a daily basis.

🔗Question 18

🔗Any update on international trademark and patent applications and what still needs to be done before going open-source?

Joseph Denne: International trademarking is pretty much complete. (For our main brand, DADI and the logo mark.) The patent application process is nearing completion, but won’t be closed out for a while. As soon as it is, we’ll move to Open Source networking repositories in full. As @arthurmingard noted above, there are components that we can Open Source sooner, which we’ll look to do in the coming couple of weeks.

🔗Question 19

🔗How can I trust DADI for taking custody over my staked amount?

Arthur Mingard: Right now custody is with us. We’re a registered business with a good track record, which is about as good as you can get in the traditional environment. As we move forward we will be moving to a trustless setup for staking, which will remove the need for you to trust us.

🔗Question 20

🔗Since Raspberry pi’s storage is on a Micro SD card doesn’t it make the read/write significantly slower than server-grade-SSD storage? A) How does that impact the network performance? B) Does it limit the possible use-cases for the edge network?

Arthur Mingard: A machine running on the same ISP with a topological proximity unmatched by standard cloud warehouse infrastructure far outweighs the potential difference between SSD and Micro SD reads. The Host application regularly runs I/O checks on its hardware and uses this to determine whether to store some frequently requested assets in-memory.

🔗Question 21

🔗Assuming the price of node stabilizes such that annual ROI is 10% in $DADI, do you expect that by the end of the year Dadi price increases or decreases? Basically if all possible nodes on boarded by EOY, would the network hold 10% ROI or higher, or expected income would not be enough?

Joseph Denne: There are a couple of questions being asked here:

  1. What impact does network scale have on the price on the token?

Broadly speaking: the larger the network, the more tokens are locked in POS = the less tokens are available to purchase = the higher the price.

  1. What does ROI look like at scale?

ROI is driven by capacity utilised on Hosts. A minimum specification Host running at capacity will return c.30% against its proof of stake, while a minimum specification Host running at 20% capacity will return c.6% against its proof of stake. We’ve modelled the tokenomics to be attractive even at the lower end of utilisation. Usage is driven by customer demand and by audience location. Customer demand is primarily driven by sales & marketing at DADI, but there is a very strong role for node owners to play as well. Want more revenue in the network in your area? Help to drive adoption!

🔗Question 22

🔗What is the maximum file size Dadi CDN will cache?

Arthur Mingard: DADI CDN is a Just-in-time asset manipulation service designed to optimise assets before delivery. This includes resizing, cropping and compression. It’ll cache very large files, but it’s purpose is to take large origin files and return a much smaller payload. The question really should be “How much can CDN reduce a file before it caches it?” and the answer is a lot. Here is an original image (893kb) and a 65% quality version (45kb) to compare.

🔗Question 23:

🔗What is the company’s biggest challenge product-wise over the next quarter?

Arthur Mingard: Security. Hardening the network ahead of self-onboarding is a huge amount of work, and it has thrown up additional requirements at various layers in the network that we are working through.

Adam K Dean: Yep, basically that!

🔗Question 24

🔗What incentives you can offer for those holding the token over one year despite being down -90%?

Joseph Denne: Hah. If only we were in control of the token value. Alas the whole market is down highly significantly, and in a cruel twist of fate, this happened pretty much immediately post our Crowdsale. If you look at what’s been achieved over the past year, and what continues to be delivered, I think you’ll agree with me that it represents a huge achievement. We’ve defined an entirely new approach to networking.

As early adopters you picked up the token at a price that reflected the valuation of the business at that moment in time. Since then the valuation of the business has only increased. However in crypto valuation and market cap are almost entirely divorced. Alas there is little that we can do about this apart from what we have been doing: building, proving the technology, partnering and driving customer adoption. The incentive for being on this journey remains the potential for upside through growth in token value and through returns from contributing to the network. And the potential today is higher than it has ever been.

🔗Question 25

🔗Which companies are using Dadi services? And where can I track usage data?

Arthur Mingard: Usage data is being made explorable, but it will not be tied back to specific customers. The reasons for this should be pretty obvious (tl;dr: it’s commercially sensitive). We report on customer wins from time to time (where a customer is happy for us to do so) and will be building customer stories on site as we move forward.

🔗Question 26

🔗Are you thinking about launching the website in several languages in addition to English to reach out to a larger group of potential customers?

Joseph Denne: Yes. Work on a site refresh is ongoing – it is being retooled to more clearly present the network’s underlying services (see: https://dadi.cloud/en/knowledge/network/network-level-services/) and should be live in April. Once this is live we’ll look to roll out translations. (Note that DADI Publish has gull multilingual support already in place.)https://dadi.cloud/en/knowledge/network/network-level-services/

🔗Question 27

🔗Will there be a client similar to dropbox and google cloud, which allows normie-friendly distributed file syncing?

Arthur Mingard: We’ve spoken about this but haven’t made a decision yet. The network isn’t looking to compete directly in this space, but we are considering the release of an open source app or a new brand specifically for distributed file storage.

🔗Question 28

🔗Plans for writing such content for manufacturers… setup/join an open R&D initiative maybe (5G)?

Joseph Denne: Yes, this is something that we are working on.

🔗Question 29

🔗How do I know all the payouts to staking devices are made correctly?

Joseph Denne: You will have sight of your earnings and your payouts will be visible and traceable within the Eth network. Individual payout txs will be linked to directly.

🔗Question 30

🔗When will I be able to actually experience DADI edge network hands on?

Joseph Denne: What do you mean by hands on? You can use sites that utilise network services today. See versus.com for a good example. If you have a property that you’d like to run on the network, reach out to us and we can chat about onboarding you to CDN right away. Failing that, we’ll be opening up Edge Storage post Mainnet testing for public testing in the coming couple of months.

🔗Question 31

🔗What is the projected use case for the Store GUI? I understand that distributed storage tech will be used in the DADI apps such as CDN & Co, but what is the benefit of having a GUI where one can upload files on the dadi account, when there are much more convenient solutions such as dropbox available?

Arthur Mingard: I’m not sure it’s fair to say that Dropbox is “much more convenient” having not seen or used Edge storage 🙂However we’re not competing directly in this space. Edge storage is designed for digital products to build on in the main – this means that a competitor to Dropbox could be built on top of Edge network. The GUI is designed to make usage easier for some projects. For example, say you want to host a static website on the network, you can upload your files directly via the site, and then map and end point to your HTML to make them publicly available. It’s all about lowering the barriers to use.

Related articles

More knowledge articles
Mail icon

Like what you are reading? Want to earn tokens by becoming a DADI Node? Save money on cloud computing services? Build amazing digital product with DADI Web Services? Join our mailing list.

To hear about our news, events and products or services subscribe now. You can also indicate which services you are interested in, which we use for research and to inform the content that we send generally.

* You can unsubscribe at any time by emailing us at data@dadi.cloud or by clicking on the unsubscribe link which can be found in our emails to you. Read our Privacy Policy.