Two-Layer Load Balancing for Onedata System

keywords: Load balancing, geographically distributed systems, DNS, web clusters
The recent years have significantly changed the perception of web services and data storages, as clouds became a big part of IT market. New challenges appear in the field of scalable web systems, which become bigger and more complex. One of them is designing load balancing algorithms that could allow for optimal utilization of servers' resources in large, distributed systems. This paper presents an algorithm called Two-Level Load Balancing, which has been implemented and evaluated in onedata -- a global data access system. A study of onedata architecture, request types and use cases has been performed to determine the requirements of load balancing set by similar, highly scalable distributed systems. The algorithm was designed to match these requirements, and it was achieved by using a synergy of DNS and internal dispatcher load balancing. Test results show that the algorithm does not introduce considerable overheads and maintains the performance of the system on high level, even in cases when its servers are not equally loaded.
mathematics subject classification 2000: 68M11, 68M14, 68W99
reference: Vol. 37, 2018, No. 1, pp. 1–22