The Blocks in Blockchain are a Problem
By Robert Lawson Vaughn on ALTCOIN MAGAZINE
Blockchain is, well, a chain of blocks. That blockchain has blocks is a major problem.
But, first, what exactly is a block and why does blockchain have them? A block, in blockchain parlance, is a collection of transactions grouped together and processed together. Blockchain has blocks for no really good reason. I imagine that blocks of transactions were included in the emerging architectural model of blockchain to minimize processing overhead. Perhaps, it was reasoned that the overhead for Proof of Work was too much of a burden to also have to deal with individual transactions.
But, is there really an issue with transactions being logically grouped together? Yes. And it is a big problem. Requiring transactions to be batched together forces an unnecessary latency in transaction processing.
Some might argue that transactions come in so fast that the latency introduced in batching transactions is no big deal. One might argue that it’s like arguing over packet size in TCP/IP. Unfortunately, those arguments are incorrect. Whereas TCP/IP breaks large messages into smaller packets, blockchain actually does the inverse. Blockchain takes small messages and clumps them together. Batching of transactions does cause impactful latency. While I have not crunched the numbers one can logically work out the relative impact. As transactions enter the network they are queued into blocks until the block fills up. And, given the volume of traffic at the particular time of day, day of the week, etc. then the degree of latency or increase in processing time increases inversely with demand. The slower the transactions come in the slower the queue fills up. The faster the transactions come in the faster the queue fills up. One might consider this to be a self-optimizing protocol feature of blockchain but that is not the case. Blockchain, because of the frequency or cadence of block processing forces congestion on the network as transaction volume increases.
The end result is that conventional blockchain systems will always be terrible at transaction throughput.
The solution is to not necessarily throw out the concept of a block but to redefine it. A block should always be limited to one transaction. Hmmm, so why blocks at all? One might view a block as the administrative and cryptographic function that connects records serially for immutability. If we go with that definition, then we can logically structure a blockchain such that the transaction stays a transaction and the block is a type of wrapper to the transaction. The result would be that a major source of latency, transaction batching, would be removed from the blockchain latency equation.
