Slides presented at IEEE DAPPS 2023.
Abstract:
Blockchains are decentralized; are they genuinely? We analyze blockchain decentralization's often-overlooked but quantifiable dimension: geospatial distribution of transaction processing. Blockchains bring with them the potential for geospatially distributed transaction processing. They enable validators from geospatially distant locations to partake in consensus protocols; we refer to them as minority validators. Based on our observations, in practice, most validators are often geographically concentrated in close proximity. Furthermore, we observed that minority validators tend not to meet the performance requirements, often misidentified as crash failures. Consequently, they are subject to punishment by jailing (removal from the validator set) and/or slashing (penalty in native tokens). Our emulations, under controlled conditions, demonstrate the same results, raising serious concerns about the potential for the geospatial centralization of validators. To address this, we developed a solution that easily integrates with consensus protocols, and we demonstrated its effectiveness.
Paper: https://arxiv.org/abs/2305.17771
OpenChain Webinar: AboutCode and Beyond - End-to-End SCA
Analyzing Geospatial Distribution in Blockchains
1. Analyzing Geospatial
Distribution in Blockchains
Shashank Motepalli and Hans-Arno Jacobsen
IEEE International Conference on Decentralized Applications and Infrastructures
(IEEE DAPPS 2023)
3. IEEE DAPPS 2023
The Pinnacle of Decentralization?
Often overlooked, but easily measurable dimension:
Geospatial distribution in consensus
3
Society
Crypto
Economics
Political
Sciences
Blockchain
Decentralization
Technology
4. IEEE DAPPS 2023
Geospatial Distribution of Validators in Ethereum
69.91% of total validators in two regions, namely the US and
Germany
Data: Accessed on November 23, 2022 https://etherscan.io/nodetracker 4
7. IEEE DAPPS 2023
The Challenge: Geospatial Centralization
01
Analyze
geospatial
distribution of
validators at
consensus layer
02
Understand if
consensus protocols
inhibit geospatial
distribution
03
Address the
factors restraining
geospatial
distribution
7
Hypothesis:
Geospatially distant validators (minority) are marked as crash failures
8. IEEE DAPPS 2023
Our Approach
GeoDec tool to emulate consensus protocol with arbitrary
locations
Prove geospatially distant (minority) validators are disadvantaged
Address concerns of minority validators with our solution
8
10. IEEE DAPPS 2023
Classical BFT Blockchains
block#123 block#125
block#126
10
Reach a quorum
Validators
Proposer
#126
Epoch 28
Validator rewards
Block is appended
11. IEEE DAPPS 2023
Measuring Validator Performance
To counter free-riding and nothing at stake problems,
validators with liveliness less than a threshold are penalized
Liveliness =
Number of blocks signed by a
validator
Total number of blocks
committed in an epoch
Motepalli, S., & Jacobsen, H. A. (2021, September). Reward mechanism for blockchains using evolutionary game theory. In 2021 3rd Conference on Blockchain Research & Applications for
Innovative Networks and Services (BRAINS) (pp. 217-224). IEEE. 11
15. IEEE DAPPS 2023
Measures how distant a validator is from rest of the validators
Only consider the quorum
Haversine distance between the validators is considered
GDI values depends on all the validators, so changes with epoch
Geospatial Diversity Index (GDI)
15
23. IEEE DAPPS 2023
To Conclude
Designed and developed GeoDec emulator
Minority validators are disadvantaged (punished) in consensus
protocols
We present GDI smart contract to preserve higher geospatial
distribution
23
Hello everyone, I am Shashank. PhD student in Computer Engineering at University of Toronto. I am here to present my work at MSRG with my advisor Prof. Arno Jacobsen.
Before we dive in, let us ask some important questions.
What are the first things that come to your mind when you hear the word blockchain.
We asked ChatGPT the same question, look at its response.
What catches my attention here is the term decentralization.
Decentralization is at heart of blockchains. If we do not care of decentralization, we have already solved most problems in centralized systems. Decentralization adds all the complexity.
Later: Maybe I need to cite ChatGPT
Decentralization is a complex concept involving social, economic and political dimensions on top of its technology. A lot of attempts are made lately to quantify decentralization but its difficult to measure all its dimensions.
In this work, we want to consider one often overlooks but easily measurable dimension - the geospatial distribution in the consensus protocols. I.e., how the participants are distributed in a blockchain.
Let us see validator distribution on Ethereum, a major blockchain. As we can see the validators are mainly concentrated in two regions, namely US and Germany. Distribution is similar when I checked it recently too.
Is it Ethereum problem? No it is similar distribution in even other blockchains, such as 0L.
As we can see, most validators are located in US and Finland, none outside North America and Europe. This makes us question if other validators are being neglected by the protocols.
We observed validators joined from Singapore and Australia but were dropped next epoch.
With that being said, let us understand if geospatial distribution is really important?
IEEE DAPPS slide number
Images with points
Why do we need more geospatial distribution?
Firstly robustness
Secondly fairness
Use some statements and examples from Notes below. Let us now move on to the problem in hand.
NOTES:
Firstly, boosting geospatial diversity contributes to the \textit{robustness} of the blockchain. In other words, the blockchain would be more resilient to downtimes caused by various factors, such as the lack of electricity required for computation, changing regulations, and even wars. Imagine most validators of a blockchain network located at a data center that does not adhere to fire safety or is prone to natural calamities. This situation is not far from reality; for instance, in September of 2021, an outage at an AWS data center could bring down the entire Solana blockchain~\cite{amazonSolana}. Even worse, a cloud provider chooses to bring down 40\% of all validators of a live blockchain for not adhering to its policies~\cite{HetznerSolana}.
Secondly, geospatial diversity facilitates \textit{fairness} of the blockchain resources, notably for time-critical operations, by lowering latencies and providing equitable access to users from all geographic regions. For illustration, consider the advantages of arbitrage traders on a decentralized exchange, such as UniSwap~\cite{adams2021uniswap}, due to geographic proximity to validators.
Thirdly, promoting geospatial diversity can reduce the burden to comply with regulations of judridictions the validators do not belong to. For instance, according to SEC, everyone who does transactions on Ethereum falls under the US Jurisdiction because the US has around half of the network's validators~\cite{SECEthereum}. This practice is worrisome as one might break the law without being aware of the blockchains' underlying infrastructure.
What are we trying to achieve in this work?
Firstly, there is little to no studies focusing on geospatial distribution of validators and dissection of consensus protocols.
Secondly, understand if there are factors leading to geospatial centralization.
Finally, we address the factors leading to more geospatial distribution
We look at consensus layer and only on geospatial locations, not data centre companies.
Blockchains can be classified into two types based on how they achieve finality.
We are interested in ones with deterministic finality, i.e., classical consensus protocols.
We have a set of validators participating in the consensus protocol, fixed for an epoch.
One initiates transaction by sending it through the client, one of the validator would propose the transaction to others.
We need a quorum. We are looking to latest ones such as HotStuff - where we have optimistic responsiveness - we do not wait for everyone.
PoS can have these as underlying protocol
Reward mechanism plays important role in consensus.
Quorum based consenus open up some attack vectors such as free riding problem and nothing at stake problem. To address these, we only reward validators who participate in consensus protocols.
We use liveliness as a measure to participate performance of validator.
Penalization either in terms of penalty or jailing.
Let us now look into our emulator
We run core consensus protocol, HotStuff, on compute canada cluster. To mimic the WAN, we use pairwise latencies given by WonderProxy
Netem tool helps us emulate these latencties between the nodes.
So given the input in cities. We use netem to emulate the WAN and run the performance metrics.
Provide the intuition for the proof.
Make minimal changes and easily deployable - we used a smart contract.
Provides same security as underlying protocol.