From a conversation on Discord on june 24, 2018. Thanks to Regnar for saving and editing it.
jl777 – http://cryptocartography.io/txscl_vis/ Right now a small test is running, only 64 chains and getting 7000+ tx/sec. Now 9000+, it might make it to 10,000 tx/sec using 64 chains. These are full bitcoin protocol crypto tx. 9892.
JohnWestbrook – So to out-perform visa we need another 47k/second…What’s the relationship between chain count and tx?
jl777 – More chains, more capacity, if you need double the capacity, double the number of chains. If you need 100x, then 100x the chains. KMD is now leading in dPoW security, JUMBLR privacy, scaling and atomic swaps.
JohnWestbrook – Nice, so this will theoretically tell us the exact number of chains to hit 60K/second. Is a chain defined as a full node?
jl777 – A chain is defined as a blockchain, you know, like bitcoin. Bitcoin seems to work pretty well, so we just replicate chains over and over.
JohnWestbrook – Right, but what do you mean when you say 64…ah gotcha. That’s what I was asking.
jl777 – 64 different chains, means 64 chains that are all different. It seems about 6400 chains will hit 1 mil tx/sec.
JohnWestbrook – Jeez
jl777 – Nobody ever said it was easy to do.
JohnWestbrook – haha
jl777 – But it can be done by replicating working tech, ie. the basic bitcoin protocol blockchain. No magic needed and unlike other scaling proposals, we secure each of these chains with dPoW.
JohnWestbrook – That would be 64000 instances running, correct? so 64K VMs could handle this? Right. So they are bitcoin-level secure out of the box. I want to write an article on this, thus my questions, gotta wrap my little mind around and see how it can easily be communicated to people who glaze over when they hear the word “chain”.
jl777 – The problem with having many chains is that there are many chains, so how does one chain interact with the other, let alone all 6400 of them. That is where our MoMoM tech comes in.
JohnWestbrook – Ah right, I was reading up on that yesterday.
jl777 – Audo has good materials on that, but the essence is that we add to the notarization data, additional levels of merkle data that allows any chain to validate any other chain’s tx, as long as KMD is around to create the proofs. This ability allows to create a burn protocol, so that coins can be burned on one chain and recreated on another.
JohnWestbrook – Could mpich be used to run multiple instances of a daemon and essentially individual chains?
jl777 – What is mpich?
JohnWestbrook – Quick and dirty super computer, allows you to tie together CPU threads from many servers on a network to run a process multiple times.
jl777 – No user has to run all the chains, they would just run the ones they care about, so however the usage is localized, users would use the few chains they need. We use komodo assetchains, so you launch a new komodod for each chain assuming a geographic allocation of chains, you would have citywide chains, countywide, statewide, nationwide. So most people just have their city’s chain locally, if they are travelling to a different city, they convert to the new city’s chain.
JohnWestbrook – Right was just thinking that through, so where is the bottleneck then…
jl777 – A bit of prep work but at least it is possible and it can all be automated. The burn protocol takes 2 to 3 notarizations, which automatically builds in security to the process, but it does take half hour or so.
JohnWestbrook – So in a sense the user experience is massive tx capacity and the notarization can continue on without impacting the performance users experience?
jl777 – Notarization only impacts a user when they are transferring coins from one chain to another then they need to wait 2 or 3 notarizations.
JohnWestbrook – okay, I see.
jl777 – All the rest of the time, the notarizations just happen, securing the chain.
JohnWestbrook – Very nice.
jl777 – No need to break things into unsecured shards, when the blockchains are already independent and dPow secured.
JohnWestbrook – Exactly.
jl777 – This is the first viable candidate to use for a nationwide crypto, all the other proposed solutions I see are either insecure or unproven or don’t exist.
JohnWestbrook – This is so much superior to ethereum proposed sharding.
jl777 – It is much simpler and simpler is much better, yes. It also costs much less per tx, a lot less, it ends up being as secure as BTC. Sort of a magic trick.
JohnWestbrookm – Yeah in every way. more secure, faster, scalable, etc.
jl777 – But this is just one use-case of our contracts tech, we continue to create useful solutions while most other projects have some sort of pie in the sky whitepaper, many describing what we already have.
JohnWestbrook – Very true. Meanwhile you are getting about 5X the current avg tx speed of Visa, and roughly 50X that of PayPal.
jl777 – Tx or payments?
JohnWestbrook – TX per second.
jl777 – One tx can have 1 payment or 100+.
JohnWestbrook – Visa avg about 1600 i think. oh right! Lol.
jl777 – At 100 payments tx, you lose 90% of tx/sec, but gain 100x on payments so a 10x increase in payments / sec. 10,000 tx/sec of 1 pay -: 1000 tx/sec of 100 pay = 100,000 payments per second. that is with “only” 64 chains.
JohnWestbrook – Ah. need to confirm the numbers on others…but for example i think ether is 20 or so, maybe that’s old info.
jl777 – What is the point for single blockchain throughputs when we can spawn an arbitrary number and we don’t need some magical maglev vacuum sealed hypersonic roadway. Just an old fashioned asphalt road suffices.
JohnWestbrook – haha yes.
jl777 – Whatever capacity the hypersonic maglev road can achieve, N asphalt roads will get much better overall traffic throughput and be much less expensive. There is only one complex/clever part about our scaling and that is the MoMoM, 99% is just brute force replication of working blockchain tech.
Grewalsatinder – I think train track example again suites here instead of roads.
jl777 – Granted, being able to spawn a chain from the command line is pretty cool and what makes it possible to automate.
Yorklab – If there is no limit (or if it’s crazy high), then Komodo should advertise as such. State the actual tested then state what’s theoretically possible, because the tested number proves the theory.
jl777 – The only limit is the number of blockchains the internet can support. More chains, more capacity and we are seeing a linear scaling of capacity per chain, now the bottlenecks have been fixed.
Yorklab – Then Komodo’s scaling capability is unlimited. Never heard anybody else state this before, much less prove it.
jl777 – It is limited by the internet’s capacity.
blackjok3r – @JohnWestbrook If you have any questions for me about the tests design let me know. It’s incredibly simple in its function. As James said we just automated a system that allows for spawning unlimited chains and then filling all of their mempools with TX at the same time.
jl777 – While it is being notarized.
Blackjok3r – yes.
jl777 – And that also allows the burn protocol to be done at the same time.
Yorklab – Speaking only of Komodo’s capability.
jl777 – KMD is needed to create the proofs, dPoW must go to KMD chain, it’s only a few hundred KMD per year per chain, so it won’t be that much of an effect, until we get 10,000+ chains notarizing. But considering that 10,000 chains can be spawned automatically it is just a matter of having the funding to launch a single cluster of 10,000 chains.
jl777 – On a separate topic I added: fill or kill, immediate or cancel and good till canceled modes for BarterDEX trades, from the Alice side. Bob automatically makes custom sized utxo for Alice so Alice just needs to have the utxo for the purchase and the dexfee. It also autosplits if there is only one utxo. Until these improvements it was running in a hybrid (and silly) config of immediate or cancel for Alice but Bob was treating it as a fill or kill, so I think order matching will be 10x better now.
Matt Watt – How on earth does one code all that stuff…? I have difficulties to just run a file on the linux terminal. Just learned to point the terminal to a directory by using cd <filename> and then execute the file by ./
Jl777 – It’s not just me coding.
Matt Watt – Gonna take course C++. Want to know more… build own mining prog and setting up FPGA cards.
Jl777 – Best way to learn is to do.
AndronicusBass – Get active and peer to peer human networking will help you advance your goal.
Jl777 – We recently got the burn protocol working so all the pieces are in place for a nationwide crypto.