Impact of lossless compression algorithms on in-memory database synchronization
| dc.contributor.author | ION, ANDREI | |
| dc.contributor.department | Chalmers tekniska högskola / Institutionen för data och informationsteknik | sv |
| dc.contributor.department | Chalmers University of Technology / Department of Computer Science and Engineering | en |
| dc.contributor.examiner | Tsigas, Philippas | |
| dc.contributor.supervisor | Duvignau, Romaric | |
| dc.date.accessioned | 2022-11-30T14:28:06Z | |
| dc.date.available | 2022-11-30T14:28:06Z | |
| dc.date.issued | 2022 | |
| dc.date.submitted | 2020 | |
| dc.description.abstract | In-memory databases have gained popularity in the last decades due to the increased demand for high-speed access to data. Redis is one such database that provides a sub-millisecond response time for incoming requests. These speedups are of particular interest in the telecommunication industry, where 5G technology needs to provide multi-Gbps network speeds. Ericsson is a global leader in 5G network equipment that benefits from in-memory database improvements. When a fault occurs in the system, methods to prevent data loss are needed. One such method is the data replication on a secondary node. Synchronization after a fault in the replica node puts pressure on the primary node to withstand the incoming requests from clients. Requests are buffered until the replica is ready. The buffering can demand a lot of main memory space, and if the system runs out of memory, the synchronization restarts. The novelty of the thesis work focuses on minimizing the impact over main memory size when synchronization between primary and replica nodes is taking place. Practically, adding multiple types of compression over received data in the Redis network layer. We gather performance metrics related to memory size reduction, requests per second, CPU utilization, relative time spent on CPU, and Maximum Main Memory used. We show that compression over the random data set, an extreme case, does not provide any memory size reduction, and it has a significant negative impact on performance. Another extreme data set is a single character generated multiple times. Intuitively, this data set is highly compressible and provided unrealistic compression ratios of 71.1. Lastly, we showcase the real data set with a 3.582 compression ratio when using the ZSTD algorithm; furthermore, the above data-set showed a higher maximum transfer rate when using compression. The maximum transfer rate shows how much bandwidth can the system support when synchronization is ongoing. Given these data sets, we showcase the positive impact of adopting compression in a 5G network. | |
| dc.identifier.coursecode | DATX05 | |
| dc.identifier.uri | https://odr.chalmers.se/handle/20.500.12380/305854 | |
| dc.language.iso | eng | |
| dc.setspec.uppsok | Technology | |
| dc.subject | in-memory database | |
| dc.subject | redis | |
| dc.subject | replication | |
| dc.subject | compression | |
| dc.subject | performance | |
| dc.subject | synchronization | |
| dc.title | Impact of lossless compression algorithms on in-memory database synchronization | |
| dc.type.degree | Examensarbete för masterexamen | sv |
| dc.type.degree | Master's Thesis | en |
| dc.type.uppsok | H | |
| local.programme | Computer systems and networks (MPCSN), MSc |