Description 

Following some major fiascos (like the classic DAO attack), the Ethereum/Solidity community has been painfully aware for some time now that smart contracts are largely custom software (modulo rampant copypaste) and thus, bugs get introduced during development - against the effects of which the security and dependability of the underlying blockchain platform does not really provide any defences. Thus, a whole industry of development-time verification and validation tools is emerging to be able to "get smart contracts right the first time" - at least Solidity ones. (The lack of such tools for other platforms, as Hyperledger Fabric, is a problem, but not targeted by this proposal.)

In contrast, "consortial" blockchain networks in general, and Hyperledger Fabric-based ones in particular, carry at least the potential to also use runtime dependability and security measures. Informally, the user does not have to "pay" for each execution step and the incentives of node operators are not cryptoeconomic, so either the platform (through customized consensus) or the smart contracts (through additional code) can have the luxury to deploy measures which protect against chaincode bugs. For non-decentralized software, the techniques created by decades of fault-tolerant computing have been codified as patterns and pattern languages; better-known examples include (state) rollback and rollforward, modular redundancy and n-version programming.

Some of these patterns are straightforward to apply to Fabric chaincode; if an implementation "fails" (e.g., throws an exception), let's try another one (n-version programming). Others may want to rely on Fabric capabilities: different peers running different chaincode implementations (software diversification) is possible in Fabric, but would not really be possible in other leading distributed ledger technologies. And some patterns are really not trivial to introduce; e.g., equitable resource allocation is a pattern to share resources among clients equitably - which, in Fabric, translates to giving equitable access to the constrained resource of block capacity to clients (an important and as-yet unsolved problem for all IoT/cyber-physical applications where Age of Information concerns about ledger data arise).

The proposed project includes the following key activities:

1. A systematic review of the applicability of fault-tolerant software patterns to Hyperledger Fabric chaincode as the potentially faulty software
2. Creation of programming idioms for applying the applicable patterns to Hyperledger Fabric chaincode
3. Prototyping "fault tolerance gadgets" (libraries) for easy application to chaincode

The book "Patterns for Fault Tolerant Software" from R. Hanmer (Wiley, 2013) is a good reference of the patterns that the project will target.

This is a research-focused project which is relatively light on programming. At the same time, a rather deep dive into the workings of Hyperledger Fabric is expected.

Additional Information

The project is expected to primarily focus on the patterns collected in the reference book Hanmer, Robert S.Patterns for fault tolerant software. John Wiley & Sons, 2013. It also serves as a good general introduction to fault-tolerant computing. The scope will be extended towards other academic papers (and existing fault-tolerant computing APIs, as, e.g., SA Forum) as and if justified by the progress.

Learning Objectives

• Introduction to open source culture and collaboration tools
• Deep knowledge of the operational principles of Hyperledger Fabric
• Software engineering techniques of fault tolerant computing

Expected Outcome

D1: Report on the applicability of classic fault-tolerant software patterns to Hyperledger Fabric chaincode

D2: Report and example set of applying fault tolerance patterns to Hyperledger Fabric chaincode

D3: A set of prototype fault tolerance gadgets

Timeline

Relation to Hyperledger 

Hyperledger Fabric

Education Level

Completed BSc level education in computer science or computer engineering (or the demonstration of equivalent competence in the core CS/CE topics) is a requirement; ongoing or finished MSc level studies are preferred.

Skills

Highly analytical thinking and the ability to effectively "tinker with" existing software and code bases (i.e., Hyperledger Fabric) is preferred.

Future plans

Depending on the results, proposal of a Hyperledger Labs project. Results are expected to be further utilized and improved on in academic research targeting Hyperledger Fabric.

Preferred Hours and Length of Internship

No preference.

Mentor(s) Names and Contact Info

Imre Kocsis, kocsis.imre@vik.bme.hu, assistant professor, Budapest University of Technology and Economics (BME), Budapest, Hungary