# TOKEN2049 - Boson Consensus: A Scalable Blockchain Sharding Consensus - Qi Zhou (QuarkChain)

- Channel: [TOKEN2049](https://streameth.org/token2049)
- Date: 2019-10-09
- Watch: https://streameth.org/watch/yt-9KDtK6jVEhE
- YouTube: https://www.youtube.com/watch?v=9KDtK6jVEhE

## Transcript

hello everyone my name is Chi Joel I'm the CEO and founder of Kwok chain so today I will present personam sensors are scalable approaching shouting consensus so here's a brief introduction of me I used to work at Google Facebook for the infrastructure working distributive system especially high-performance system then we found blood chain which we believe a lot of decentralized solutions can be applied to the opportunity area so that's why I moved to this kind of shoddy area okay so probably everybody knows that rock change right now suffering from low transaction per second for example like Bitcoin roughly delivers a probably seven transaction per second and a theory is roughly deliver fifteen transaction per second so this low transaction per second greater limits the wider applications are blockchain so there are a couple of solutions there aims to address this degree program and one solution is public - just to increase the block size of the of the blockchain for example beacon cash increase the block size of Bitcoin from one megabyte stusertwo mikvahs so that you can claim through science performance improvement which is about 200 transaction per second how error increase in the block size we also increase the block propagation time which means that the probability of still block of fox will be increased resulting in wasting of hash power are also further security concern in addition if there's a lot of computational transactions such as my counters are put into two large blocks if we want to perform this transaction in the middle of time then we need considerable computational resources another solution which basically employs multiple blockchains however immediately faced another issues called hash power pollution where suppose all these n chains employ the same people work hash algorithms then there exists list one blockchain has a most one over m has power I suppose N equals 100 that is 1 over 100 has power this means that an attacker can easily provide a double spending attack to revert a transaction on one of the blockchain and greatly compromised the security level of the blockchain a further issue is interoperability issue for example a user has a token in chain 0-1 to perform a culture on chain 1 and this kind of cost chain transactions will be extremely difficult to perform so to address this problem of both large block size and model buck chains we propose both long-term sensors well it is a two layer block chain consensus where it consists of a short seller which contains a list of homogeneous change or become shorts and it process the transactions in this in the blocks similar to multiple park chains and to adjust the hospital Lucian issue we introduced a root chain which just does not process any concessions but it described the canonical chains of all the shorts by including the the headers of the shorts into the blocks so so here actually is an example of the latter produced by Poisson consensus where we have two shorts which have the hash point turning to the previous block and for the root chain besides the hash pointer link to his blog it also has a hash pointer that points to the our headers are the sharp rocks so now we have n plus 1 chains and how we can produce the plug in these chains and which means that we have n plus one consensus that can be run independently for example route chain can run its own consensus for example proof of work with total difficulty as the is fortress rule to put his block and can also run some Depot's opposed as long as its body is valid which means that including the canonical chains or the sharp rocks and it sure can also run it sensors such as proof of work mistake or depose plus a consensus called routine first consensus so that all these blocks in the straw can be protected by the root chain so what is routine first consensus basically suppose we have two short forms before comparing these two shocks using their local consensus the first reverse lookup their longest route chain include this shock and to find which one should be selected based on the routing consensus and suppose a routes a shock has longer route chain fork then it will be so lucky no matter how how long an inertia for it is and if they have to say the ABS have the same shaft roof chain fork then the local consensus will be applied so with this routine first consensus we can derive a very important property often person consensus which says that any double spending attack on a sharp rock that is included but every chain must also attack the root chain this means that that security level of the sharp rock is also guaranteed by the routine so for example here attacker tried to revert the Salah to chain saws ill of these three blocks by creating a longer fork however this double spending attack will never succeed because it's corresponding root chain block which is this one it's shorter insulated chain which is this one so in order to perform a successful double spin attack the attacker must also create attacking fork on the root chain where thief his consists of three blocks blocks versus these two blocks include four corresponding attacking Shafak in order to perform this attack so that all this block will be reverted however this kind of double spending attack will be much more expensive just proponent of spending attack on Sherlock so but some consensus actually share a lot of similarities and differences with existing solutions so one famous solution is called a merge Miami where a miner mines the main chain and also auxilary chains simultaneously and if they reach the difficulty of any chains then just broadcast block of a corresponding change however it has two requirements first is that the auxilary chain must use in the same hash algorithms and second is that once the rich multiple difficulties of different chains all the blocks must be produced synchronously which gives a lot of pressures to network penneth well but some cam sensors first and in Chhabra can use any hash algorithms and all the blocks are produced asynchronously which significantly increase the utilization of never penneth another famous solution is composed which select a canonical chain by ticker can the total the difficulties they called uncle's are Fox a increased block rates however since those transactions can be overlap in the canonical study chain and also the Fox the transaction rate is also still be limited and the ledger actually produced by Poisson consensus is actually essentially a tag or a well structured tag however it tried to harvest the benefits of both blockchain and dad so one great possibilities they're supposed to director produced by Poisson consensus we can easily tell which one to be selected so that the security model is well-defined unlike it sitting some of existing that solutions that it depends the secure level is not well-defined and highly depends on node configurations and it also achieved higher to pro compared to single blockchain because all the plants sharp large rubrics are produce asynchronously and so that it can utilize never been with more efficiently and so previous sorry previous time we focus our most directly cosigner for some consensus so we also have now removed to end more engineering sign so with implementing for some consensus as in the quad chain which is the first element ation and using the proof of work as a major consensus the root chain used in the proof of work with is a hash insurance can use improver work with all kinds of hash algorithms which we have this test name has been running for having more than three or four four months and we also enable clustering which allow different processes to run different shorts in Indiana mean and communicate with each other with our PC this greatly allows you to enhance the capacity of the full node but just just add more machines into the network with this so the article here open sourced and we use in the six thousand nodes to run this quad chain and we are and demonstrate to achieve more than 14,000 transaction per second using a steering title of chunk parents transfer transaction and our community members using the open source code and the highest number they can get is more than 50000 transaction so besides the scalability another are they also are the major topics are important topics of Poisson come sensors why is that how to design a token economies so that the rubrics is sanitized she include the player to retain the sharp lock headers as much as possible and also suppose the sister network its riches is capacity how they are able to act more Shores in flight without stopping the network and we also have several topic in terms of in security so that any shards can be emitted when join a network can immediately be protected by the hash power of the root chain suppose Rajendra's proof-of-work which occur iike hash power reuse techniques which is inspired by the code reuse techniques that commonly use in software developments in addition with Poisson cam sensors we can demonstrate that it can achieve we're a highly efficient cost or transaction and the security level of the closure transaction can be the same as the in transactions in terms of the post building attack and the further topics in turnstile is the homogeneous chart where the shots can run different condensers can also run different for example virtual machines lectures or even talking economies so there are a lot of space can explore in this error okay thank you very much for attending my representation any questions yes um I'm sorry in thing just like they don't get so you mean well when there is a device join a network we how to decide which are it will runs so right now currently we are only in neighborhood like a fool a footnote that is joining Network which means that this no can this photo master process all the full adder however it can separate this fernell into different sub nodes however in choir this some notes to be to trust with each other so they are able to process all this sad sad data are also retained data within record cluster [Applause] are you developing any DFS along with your chain the ETF's decentralized applications yeah good question so right now the system is more like a steering style so that you can support like if your own virtual machines and for long user perspective it will similar to use East Iran for example I can do any chance for balance Chancellor no matter is in charge transaction across transaction or smart counter cause that no matter it is just my country reside in a local sure or automotive store so this is our first step and future stopped as I said like we are able to enable heater genius charts and so we are able to example incorporate latest example virtual machines and also consensus rectangle like a deep holes from user experience wise so there are a lot of space we can explore here so I don't think there's a limit there okay yeah yes pecking watch you