International Journal of Computer Applications |
Foundation of Computer Science (FCS), NY, USA |
Volume 175 - Number 19 |
Year of Publication: 2020 |
Authors: Jay Parekh, Apurv Moroney, Lavina Golani, Radha Shankarmani |
10.5120/ijca2020920718 |
Jay Parekh, Apurv Moroney, Lavina Golani, Radha Shankarmani . A Timestamp based Novel Caching Mechanism for Distributed Web Systems. International Journal of Computer Applications. 175, 19 ( Sep 2020), 36-46. DOI=10.5120/ijca2020920718
Microservices are a loosely coupled distributed systems architecture. With the uncertainty in prediction of the size of the application, microservices play an important role in development and scaling. As they are independently functioning applications, difficulties in caching data becomes manifold. Caching in microservices is achieved either by maintaining a local cache with peer to peer communication, or a global cache with single store communication. However, multiple local caches come with communication overheads and data consistency issues, while a global cache has data management issues. This project attempts to find a combination of both to reduce the communication overheads and data size, while solving the problem of data consistency. Focus is to create a mechanism which uses both a global cache and a local cache. The global cache would act as a verification cache and the local cache would act as a data cache. This will minimize the size of the global cache and the communication call size. In comparison to the existing cache management techniques, this system will act as a middle ground. It inherits the low communication overheads from the global caching systems and also manages to keep the global cache size minimum by storing only verification data. The impact of this research topic is multi-faceted, not only in scaling web applications to a global scale but also in maintaining modular, data-consistent caches in a cluster of microservices. Another advantage of the proposed solution is that it can ameliorate the problem of bandwidth always falling short in high load applications