A Blockchain that Handles Millions of Transactions per Second?
Anyone who tried to send some bitcoin over the past few days and found themselves paying a fee of $20 or more—only to wait hours for the transaction to confirm—knows about blockchains' scalability issues. Bitcoin cash was forked from the main chain to try to speed things up a bit, but the 30-something transactions per second the new cryptocurrency could reasonably process pale in comparison to Visa's 50-something thousand.
Which is why, if Kochava founder and CEO Charles Manning is right, the product his company is building could attract a lot of attention in blockchainland. He expects XCHNG, a smart contract platform Kochava hopes will disrupt the digital advertising industry, to process millions of transactions per second.
"Yeah, millions, exactly," Manning told me on a phone call last week, in response to an incredulous request that he repeat the number. "That's what's amazing to me, is that financial services folks often point to their transaction volumes and say, well it really needs to beef up. If you want an example of a very difficult, high-volume ecosystem, it's adtech. Think of every impression as a trade."
Processing a transaction in 45 or 50 milliseconds is nothing new for the industry or for Kochava, a six-year old company that, until around two-and-a-half years ago, had nothing to do with blockchain. Maintaining that kind of speed when moving from a centralized model to a decentralized one, though, has so far defied solution.
To be sure, XCHNG—which Manning expects to go live at the beginning of 2019—does not rely on proof of work, which would probably preclude even speeds a few orders of magnitude slower than Kochava aims for. XCHNG is its own animal: "It's a ground-up blockchain," Manning says, "it's our own consensus engine, it's our own protocol for peer-to-peer, it's our own persistence layer. It's not just some offshoot of ethereum." If it achieves its goal, that will be due to a healthy dose of originality.
XCHNG is designed to replace insertion orders (agreements detailing how a publisher will serve a buyer's ads) with smart contracts. These are blockchain-based protocols, first proposed 20 years ago but popularized in 2015 by ethereum, that enable self-executing bits of code to be written onto a distributed ledger.
As noted above, however, XCHNG does not take the common route of building on top of ethereum. This is by necessity: "It can't be on ethereum," Manning says, "based on the model and the volumes and the dynamic. It would actually also be potentially really expensive on a per-smart contract basis if you use the gas model" (gas is the ether-denominated fuel, so to speak, for ethereum's proof-of-work consensus protocol).
In proof-of-work blockchains like bitcoin's and ethereum's, miners compete to find an acceptable hash in an extremely expensive process: bitcoin, by some estimates, burns as much electricity as a small country. XCHNG will run on a proof-of-stake model, meaning that the node that mines the next block is selected partly at random and partly based on the size of its stake in the blockchain's token. The nodes on XCHNG's network are not simply writing transactions to the ledger, however, but serving the ads themselves. "It's like a proof-of-ad-serve model," Manning says.
Proof of stake enables scalability that proof of work (currently) cannot, but it's hardly new. What is new is a form of radical pruning, the "Daily Rolling Chain," which means that, as XCHNG's whitepaper explains, "each node on the network only persists and can only validate a daily transaction history for those open contracts for a given day's date and moving forward in time."
Where full bitcoin nodes must download the ledger going back to the 2009 genesis block, XCHNG nodes' memory will be considerably shorter: one day. This is bound to rankle purists and does raise concerns about the network's security, but Kochava aims to allay these fears through a ranking system that tasks reliable, reputable nodes with checking that new transactions are consistent with the ones that went before it.
Some nodes would have an incentive to store all previous transactions, the whitepaper argues; the reason has to do with the next question.
Why a blockchain?
Since the cover of Bloomberg Markets proclaimed "It's All About the Blockchain" in 2015, there's been a rush to commit anything and everything to an immutable ledger, whether it makes much sense for a given set of data or not. And failing that, it's best to at least say you're building a blockchain. So I asked Manning why Kochava thinks it needs one.
In his view, there are very good reasons to decentralize insertion orders. The ability to transact with third parties without trusting them was bitcoin's great contribution, and trust is severely lacking in the adtech industry. Manning walked me through the history.
Chapter 1: "Publishers would have an ad server, advertisers would buy their inventory, publishers would serve that inventory, and they would provide a report back to the advertiser. The advertiser fundamentally didn't trust the publisher, because most of the time they were lying. And they had no way to prove that, without necessarily having an advertiser ad server."
Chapter 2: "You had an ad server with the publisher and an ad server with the advertiser, and the ad request would get tossed between the two of those things. And anything with greater than a 10% discrepancy would be in conflict, but anything with less than 10%, they would just consider it okay. It's amazing that they would consider 10% discrepancy okay, but it's because of that complication."
Chapter 3: Google and Facebook capture nearly all of the industry's growth and move in on incumbents to create a duopoly.
Chapters 1 and 2 illustrate the case for a blockchain as a single source of truth in an industry that has been rife with inconsistent reporting and outright fraud. Chapter 3 illustrates the case for a distributed network, and XCHNG is designed to be truly public and distributed. It is not a "private blockchain."
The issue is, in order to be fast and distributed at the same time, XCHNG has opted for an amnesiac ledger that appears at first glance to skimp on security, consistency and reliability. Manning believes that sufficient nodes can be persuaded to store the complete transaction history, however, due to another potential consequence of introducing a blockchain to adtech: turning digital ads into a true asset class.
If ads are traded on a single, transparent platform, all sorts of ancillary industries can sprout up, Manning says: tracking and measurement (Kochava's core business), rating (à la Moody's), and industry reporting (comScore's reports could be deterministic, not panel-based). These business models would require nodes to keep much more extensive ledgers than just day-of transactions.
You could be forgiven for wondering who this team is that claims to be creating probably the world's most scalable blockchain. "We're in a world of real-time systems," says Manning, and Kochava's engineers are used to "transacting a million-plus transactions per second." None of them, though, are "blockchain technerati … blockerati"—warmer—"chainerati?"
There's been minimal Medium and no GitHub hype. Even the whitepaper is by request only. XCHNG has kept a low profile in order to avoid what Manning sees as fintech's folly: "Every single company built their own blockchain implementation, and they either forked something else or they built something from scratch. And we learned from that and said, in adtech, there is value in having this be open, but there is even more value if everyone's on the same system."
Kochava's ambitions may not end at adtech, though. XCHNG has been designed so that its peer-to-peer framework can be decoupled from its inventory hash, can be decoupled from its consensus mechanism. "The combination of our consensus and peer-to-peer system will end up finding its way in lots of other toolkit blockchains," says Manning.
First, though, the public will want some evidence that this millions-of-transactions-per-second claim is real. Manning says we can expect to see some numbers early next year.
The views and opinions expressed herein are the views and opinions of the author and do not necessarily reflect those of Nasdaq, Inc.