An investigation of consensus and performance in distributed systems

Within a tiered or zoned architecture, new business constraints for regional data residency, imply a need for new architectural patterns. These constraints are not arbitrary: they arise directly from customers demand to ensure their confidential data stays within their defined borders unless required otherwise (e.g. data originating in the U.S. is intended for an E.U. entity). An architecturally clean approach is to negotiate the movement of data only at the persistence layer. The objective is to review the industry landscape for suitable replacements or to otherwise alter the existing design to facilitate the replication of data based on these requirements. Specific targets for consistency latency (with defined upper bounds) and parity performance for read and write rates must be met. Parity performance is measured via msg/s and is achieved if the modified/replacement system has at least the same performance profile (average, median, and 99% percentile) as the existing system.

Faculty Supervisor:

Matt Medland

Student:

Duc Truong

Partner:

Ethoca Technologies

Discipline:

Computer science

Sector:

University:

University of Toronto

Program:

Accelerate

Current openings

Find the perfect opportunity to put your academic skills and knowledge into practice!

Find Projects