Apr 19, 2019 · 52m · a16z
Towards a Decentralized Storage Network for the Blockchain Computer
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the a16z crypto series, host Ali Yahya interviews Stanford PhD researcher Ben Fish to explore the technical, economic, and cryptographic mechanics behind decentralized storage networks. They discuss how open protocols, cryptographic trust models, and Proof of Replication enable secure, censorship-resistant alternatives to centralized cloud storage giants.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →
The host as informed peer 4.2 Guest teaching 5.4 Guest disagreement 0.2 The host pushing back 0.7
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 24:40 Clarifying Proof of Replication Terminology
The guest gently pushes back on the host's framing by noting that the term proof of replication is somewhat misleading and clarifying what it actually proves.
Hardest push from the host ▶ 40:26 Commodity Resource Attack Vector ChallengeThe host directly challenges the network security model by asking if commoditized storage allows dominant actors like Amazon to easily launch 51 percent attacks.
Biggest teaching moment ▶ 32:04 Tight Proofs of Space ExplanationThe guest educates the host on the necessity of tight proofs of space, explaining why older proof of space schemes that allow 50 percent space savings fail for storage networks.
The host holds their own ▶ 11:27 Bruce Schneier's Trust FrameworkThe host demonstrates deep subject knowledge by bringing up security researcher Bruce Schneier's four trust mechanisms to contextualize the protocol paradigm shift.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Opening Title & Legal Disclosures | 3 | 3 | 0 | 0 | The host introduces the episode topic and guest Ben Fish before engaging in introductory banter about unused phone storage capacity. | |
| Marketplace Creation & Censorship Resistance | 4 | 4 | 0 | 0 | The host outlines key economic and functional reasons for decentralized storage, and the guest expands by comparing decentralized storage networks to Bitcoin's consensus ledger. | |
| The Challenge of Peer-to-Peer Trust & Cryptographic Frameworks | 6 | 4 | 0 | 1 | The host demonstrates domain expertise by citing Bruce Schneier's four-part trust framework to categorize the trust shift from institutions to cryptography. | |
| Proofs of Retrievability vs. Proofs of Replication | 3 | 6 | 0 | 0 | The guest explains the distinctions between proofs of retrievability and proofs of replication, detailing how verifier incentives work. | |
| Converting Computational Work into Useful File Storage | 5 | 5 | 0 | 1 | The host synthesizes how cryptographic storage replaces wasted Bitcoin computational cycles with useful file storage and raises environmental critiques. | |
| Fault Tolerance, DoS Protection, and Sybil Attacks | 5 | 5 | 0 | 0 | The host proactively identifies potential Sybil attack vectors in multi-replica storage, prompting the guest to explain fault tolerance. | |
| How Proof of Replication Guarantees Honest Space Allocation | 3 | 7 | 1 | 0 | The guest gently reframes the terminology around proofs of replication versus proofs of space to correct common misconceptions about space dedication. | |
| Technical Explanation of Tight Proofs of Space | 2 | 7 | 0 | 0 | The guest delivers a deep technical explanation of tight proofs of space, detailing why high space dedication prevents rational actors from cheating. | |
| Cryptographic Challenges and Asymmetric Encoding | 4 | 7 | 0 | 0 | The guest details asymmetric encoding and time-bounded challenge responses, after which the host accurately summarizes the mechanism. | |
| Commodity Storage vs. Specialized Hardware and Attack Vectors | 7 | 5 | 1 | 6 | The host presents a strong counterargument questioning whether relying on commoditized hardware like storage exposes the network to cloud giant attacks compared to ASIC networks. | |
| Economic Overhead and the Farmers Market Analogy | 4 | 6 | 0 | 0 | The guest uses a farmers market analogy to explain operational friction for enterprise providers, and the host prompts for a teaser on cryptographic accumulators. |
Statements from this episode (20)
Yahya: Half of average computer hard drive capacity goes unutilized
“You could say on average, ah, maybe half of the hard drive of any one particular computer is likely to be unutilized.”
Fish: Governments can easily force centralized cloud providers to censor files
“If you have Amazon in charge of all the storage, then maybe Amazon is not intending to censor Individuals, but say a government is, it's much easier for a government to prevent Amazon from serving files to their country than, you know, a amorphous internet of,…”
Yahya: Decentralized storage protocols guarantee fair rules and censorship resistance
“Then you end up having some of these additional properties that it's censorship resistant you can trust that, that that sort of the rules that make it a platform will stay the same or will stay fair because it's the control sort of by a, Body of representative…”
Yahya: Centralized storage is the weakest link for decentralized applications
“All of those will require storage, and if that storage were centralized, Then the whole point would be lost. That would be the weakest link. So you need decentralized storage for those applications to come along.”
Fish: Bitcoin is a decentralized storage network for a single file
“You can think of a, you know, decentralized accounting system, Bitcoin being the first example of such as a example of a decentralized you know, storage network. It's just that the one file that the entire system has decentralized the storage of is the ledger …”
Fish: Decentralized networks give anonymous devices Amazon-level cryptographic trust
“You don't have to spend, you know, millions of dollars on PR to you know, to garner a name like Amazon. Right. You just have a device. You join the network. Nobody knows who you are. But yet everybody should be able to have the same trust in you that they have…”
Yahya: Cryptographic security scales trust globally by removing human intermediaries
“There's, there are fewer humans involved and therefore this could be more scalable. You could scale trust to a larger number of people. You could imagine this being a truly global, ah, decentralized network for storage where everyone can participate because th…”
Fish: Standard proofs of retrievability are impractical for decentralized storage networks
“But then when you start to talk about, you know, the economics of an entire decentralized storage network, it seems quite impractical to say, okay, well, you're a user. If you want to store your files in this network, then you have to constantly ping the netwo…”
Fish: Bitcoin's proof-of-work energy is largely wasted computational work
“The thing about Bitcoin is that all that work is just sort of computational energy that is, ah, that is, say, somewhat wasted, although it maintains the system in balance.”
Yahya: The Bitcoin network consumes more electricity than Denmark
“The Bitcoin blockchain alone, and it consumes like more power than like Denmark or something like this.”
Fish: Proof of storage turns environmental footprint into practical file storage
“In a decentralized storage network that is based on proof of storage, the footprint is something useful. It's file storage itself.”
Fish: Data replication provides fault tolerance and censorship resistance
“Replication is what gives you fault tolerance and say censorship resistance. If there's only one node that's storing your file, then if that node goes down, you could lose your data. And from another perspective, if there's only one node storing your file, the…”
Fish: Cryptography cannot force nodes to store unencrypted file duplicates
“There really is no way to force someone to store multiple copies of a file. The only thing that you can cryptographically guarantee Is that somebody is using as much space as they claim to storing files.”
Fish: Storage networks must prevent nodes from deduplicating for profit
“And so if you take the perspective that nodes in a decentralized storage network are going to be, you know, acting not necessarily maliciously, but Rationally, and that if they could earn more dollars by saving a little bit of storage, and therefore destroying…”
Fish: Proof of Replication enables storing useful data on disk
“And so that's really what proof of replication is. It's a really tight proof of space that you can use to store useful data.”
Fish: Proof of Replication uses asymmetric encoding for fast file recovery
“Proof of replication designs this encoding process in an asymmetric way so that it takes a long time to derive a encoding of a file. In a very short period of time to recover it, or at least, ah, the decoding step could be parallelized.”
Fish: Proof of Replication prevents Amazon from easily attacking decentralized networks
“When you do something like proof of replication, it's an extremely inconvenient way of storing files, right? So if Amazon were to convert all of its storage to you know, proof of replication for formatted storage it would be quite inconvenient if not if not co…”
Fish: Proof of Replication balances ASIC barriers with commoditized storage access
“So I think that's sort of where, Where proof of replication falls. Somewhere in between something which has high, you know, barriers of entry, such as something which requires ASICs you know, versus something that is just so convenient for Amazon to do and par…”
Fish: A 1% storage overhead hurts Amazon but empowers individual users
“I mean, when you're talking about a huge business like Amazon where the potential impact of, you know, making a small difference in the latency that they have for their customers or a small overhead in the storage that they're using. It could be, it could real…”
Fish: Accumulators separate blockchain consensus from storage for mobile validation
“Yeah, I recently gave a talk on this in a way of looking at this as a separation of consensus and state but really it's a separation of the consensus layer of a blockchain accounting system from the storage layer you know, where the, you don't have to have the…”