Consumer IoT (CIoT) needs edge computing
Liu Tao, the founder of ION chian: consumer IoT shall have edge computing in the future
Fang Jun: Let’s start from the fundamental questions: After going through your white paper, you put emphasize on “Edge Computing”, could you please explain it in a simple way? What is the requirement for IoT nodes? What is the percentage of all the IoT network nodes that can work out edge computing?
Liu Tao: Edge computing is one kind of technology which can give computation skills to the data produce source devices. And at most times, the so-called data produce source is the IoT device terminal itself. The edge computing can mainly solve data quantity, privacy protection and instant reaction.
IoT nodes can be realised by introduction of edge computing chips or modules and together with certain embedded programming system. Regarding to the IoT devices with or without installation of edge computing chips, we can classify them with whether it has source and computing power or not. If it has both source and computing power, it is possible to conduct edge computing after some modifications. The application range of IoT is very wide, so ION chain will focus on the field of consumable level IoT.
Fang Jun: So, in the field of IoT with consumable level, what is the percentage for them to use the “edge computing”?
Liu Tao: It is estimated that by 2020, there will be 35 billion IoT devices connected to the network. If all of them would be process with central cloud, there would be no network which can handle this work load. My understanding is that the IoT devices manufactured in the future will be equipped with edge computing skills except pure data acquisition sensors
Fang Jun: Yes, when I was reading your whitepaper, I found that the ION chain node you talk about looks like much “bigger” than what we normally understand? For example, the locker in the share bicycle and the garage ground lock mentioned in your scenario? Can the IoT nodes undertake that high cost? The IPFS question will come later.
Liu Tao: As long as the device have the built-in ION ID and configurable ION chain application layer protocol, it could be a ION chain node. Normally, the ION chain node means the IoT node. Before, the IoT nodes transmit data to the cloud, but now the data is transferred to the ION chain and communicated peer-to-peer.
Fang Jun: How do you add the built-in ID and protocol to the nodes? Like, Add one more chip? What is the estimated additional cost for one IoT node connection?
Liu Tao: It is necessary to discuss with module supplier about adding ID and protocol layer. So far, we are under negotiation with Ai-link and MXCHIP. If we can get support directly from the module level, there is almost no additional cost for IoT device manufacturers. In addtion, it can give users one more choice, whether they want to give the cloud data free of charge, or upload data to the ION chain for mining.
Fang Jun: According to what you’ve said, 1) Adopt B2B2C mode, and collaborate with device manufacturers directly, 2) Users could choose whether they want to upload data to cloud or ION chain. Then, where does your chain run？The chain consists of all the nodes? Do you have any other node besides the user node?
Liu Tao: There would be one DAPP store on top of ION chain, and the underlying data and instructions are all open sourced, so people can develop DAPP based on that. This design is also the process to consume the ION coins, which means the devices produce coins with data and liveness, and the DAPP will use that. For sure the ION Foundation will invest on the initial ION node, afterwards large device manufacturers can also join and work as nodes to share committer and storage awards.
Fang Jun: The IoT notes for users are the same with the nodes for your device manufacturers? Are they two categories or more?
Liu Tao: The nodes for device manufacturers are committers, IoT device nodes are like digital wallets and data storage nodes. That is all.
Elder Ryan: But RFID shall be the most universal IoT application scenario? If you restrict on gateway level, a gateway will be isomorphic to a small computer. Then, it will become isomorphic with Filecoin.
Fang Jun: We understood your node classifications. Let’s continue with IPFS, so the IPFS should be committer? But there will be another question, committer is still the edge of edge computing? Like nodes in the digital wallet, is there a IPFS storage? My question is actually very related with Elder Ryan’s, you can answer that in a comprehensive way.
Liu Tao: We hope the scenario would be like this, image it, I have a smart refrigerator at home, and through AI algorithm I know that I need to drink one cup of juice every night. When the juice is nearly over, the refrigerator will place an order on JD.com and the JD UVC can bring the juice to my window. Then, the robot will help me to take it, put back to the refrigerator. This is communication between three items, and we are not the medium between them anymore. RFID mainly works on source traceability and data digitization, and there are some other blockchain projects especially designed for this application.
IPFS mainly solves the storage issue of massive IoT data. So far, it is forward mode, in which have the data on chain first, and then the chain will be stored in IPFS. So it is currently not in the edge computing layer. If we use chips, it is possible to run the proesses in a reverse mode. The data is stored in IPFS first and then read through the chain. IPFS is a storage network which is centralized logically but decentralized physically.
Fang Jun: I am going to ask some detail questions about the design philosophy of ION chain. You’ve mentioned that most of the blockchain projects put “value creation” and “value transfer” together, like Bitcoin or Etherum. So why do you want to split them up ?
Liu Tao: We hope the devices which are connected to ION chain are “alive”, and that is why we have proof of time lapse. The integrated devices become ions, and their feature is to combine freely with other ions. Regarding to how the combination could be, we leave it to the DAPP developer.
The reason why value creation and value transfer are together is because of the consensus mechanism. In PoW, value creation is a special value transfer, but in ION chain, the value creation comes from data quality and online activity of the devices. So, value transfer is the reflection of data value and transfer of device control authentication. Thus, they themselves are two different processes.
Fang Jun: So The value creation and value transfer are in one node finally? Or the nodes are divided into two levels? The logic is clear, but it is not that obvious who will implement it?
Liu Tao: Value creation is mining, and it is what we called “every device is a mining machine.” They are not in the same node, but two chains anchored to each other. Actually, we are trying to apply chain in data mining, and DAG in data transmission.
Fang Jun: You just mentioned two chains? Are those for value creation and transmission? How they will collaborate with each other？
Liu Tao: Their relationship is mutual anchored, or you can consider as two shards in one chain because they have shared data.
Fang Jun: I will find another time to ask more questions about the two chains. I would like to ask one general question: from your perspective, what is the biggest value it can bring if combining blockchain and IoT? Compared with other projects, does ION chain have any special design, or provide some particular application value? I am not going to ask too much about the consensus mechanism because it’s already very detailed in your white paper. But If you have anything would like to share with us, please go ahead.
Liu Tao: IoT devices themselves are decentralized, but what the industry is doing now is to add them cloud servers, and they don’t share with each other, which causes that the items cannot interact with each other. The value of ION chain is to make all the related parties involved in via “every device is a mining machine” and design of token economics, and then define P2P communication protocol between devices. Finally, it can reduce data transmission volume with edge computing to make the network a sustainable development.
Attach: Open discussion between Liu Tao and some group members
Wang Kejie: May I ask you what kind of information will be recorded on ION chain?
Liu Tao: We will maintain the pedigree of one IoT device to define which data is valuable (it means the data could be used by other devices or individuals). Data is stored on IPFS, and there is only index on the chain.
Wang Kejie: What kind of content is the index stored on chain for? IoT devices interact with each other? You mentioned that 35 billion IoT devices will be connected to the network. Even though they only keep very basic information, the data volume will still be tremendously huge, so how does the IoT nodes store the blockchain data?
Liu Tao: Index that contains valuable data generated by IoT devices. Interaction with IoT devices is to store the information itself, and one of our most important tasks is to define the application layer protocol for device interaction. For example, all the devices will have turn on,turn off, sleep and other functions.
vion williams：It seems that ION chain will integrate all the IoT devices in the future, but IoT involves a wide range of fields. Also, as mentioned in the whitepaper, there will be different consensus algorithm for different fields. But different algorithm and consensus mechanism may have deviation on the value creation, how to balance that?
Liu Tao: The mining mechanism is proof of data quality and proof of time lapse.
Fang Jun: Is it “proof of time lapse” ?
Liu tao: Yes, we need to use different smart contracts to match the devices in various fields.
vion williams：what is the qualification standard for data quality？
Liu Tao: How to compare the value of data generated by different devices horizontally is a core issue. Our idea is to hand this over to our community and market. Proof of data quality is designed through a weight mechanism. The ION chain mode is B2B2C, at the beginning, we only collaborate with IoT device manufacturers, and do not accept individual users.
Elder Ryan：I think the IoT you mentioned here is more about IoT’s Gateway. In terms of traditional IoT based on RFID, like, I have 100,000 books, and each book has one RFID label, if I want to put the book storage record onto your chain, how should I do？Do I need to purchase one gateway from you? Also, if there are too many labels, how to solve the data throughput issue?
Liu Tao: It is necessary for RFID to go through gateway when uploading data. However, RIFD is not the application which ION chain wants to integrate because it is one way.
Elder Ryan： I have read one paper about Iot On Edge-Host with Blockchain from Saskatchewan Unversity. Their conclusion is that there will be a 1–1.2 second latency in every 10 resources.
Liu Tao: This paper mainly talks about comparison and latency about access and transfer layer.
Elder Ryan：I think this performance is much better than i expected, so I am also wondering what could be your expected performance.
Liu Tao: COAP is a HTTP protocol which is simplified by binary instruction, and the existing test data is 5000TPS.
Elder Ryan：It is very high? Is it fine to disclose your testing benchmark?
Liu Tao: We will release the mainnet in this Aug. Still in test now. We are useing internal network and there are 9 nodes.
Elder Ryan：Ok, I understand.
Elder Ryan：I notice that in your whitepaper, you talked about IONIZATION algorithm, and the separation of value creation and value transfer. It seems that it is used to solve the value differentiation based on various consensus algorithm. But the paper lacks a functional description about the value transfer layer, what is the implementation logic?
Liu Tao: Value transfer is accomplished via modified PoS, or we call it PoS mechanism with deposit. There are different strategies for the three situations in value transfer, which are between devices, device and human and between humans. Nothing special other than that.
Mai Niu: I don’t think it is necessary to have multi nodes in the public chain. It could be the fastest way if we have only one node. By the way, we don’t hear about other voting election for nodes except EOS, I don’t know whether there is anybody who understands why BM chose 21 main nodes.
Liu Tao: If there is one node, then we don’t need any blockchain structure, even no DAG, because the traditional infrastructure is enough. E.g. Alipay can accomplish 300,000 TPS.