Data Center Load Balancing Method Based on Nonlinear Network Traffic Analysis
To improve the quality indicators of the distribution system and load balancing of Data Processing Center clusters (DPC), it is necessary to apply the method of dynamic load distribution across servers based on the use of an algorithm for accounting for the states of input traffic and its prediction...
Spremljeno u:
| Glavni autori: | , , , , , |
|---|---|
| Format: | Статья |
| Jezik: | English |
| Izdano: |
Springer Science and Business Media Deutschland GmbH
2024
|
| Teme: | |
| Online pristup: | https://dspace.ncfu.ru/handle/123456789/29188 |
| Oznake: |
Dodaj oznaku
Bez oznaka, Budi prvi tko označuje ovaj zapis!
|
| Sažetak: | To improve the quality indicators of the distribution system and load balancing of Data Processing Center clusters (DPC), it is necessary to apply the method of dynamic load distribution across servers based on the use of an algorithm for accounting for the states of input traffic and its prediction. In accordance with this, the purpose of this article is to develop a practically feasible method of load balancing, characterized by taking into account its spikes, self-similarity, long-term dependence. According to the Takens-Manet theorem, the state of a process described by nonlinear dynamics can be represented for any moment of time by temporarily implementing one of its parameters using the embedding procedure and providing autoregression of its parameters as a trajectory in phase space. The restoration of the phase space of the process by the method of delays makes it possible to determine the properties of the process by individual parameters of the time series describing it. Therefore, the method of restoring the phase space, which provides an assessment of the dynamics of the states of the process under consideration, can be the basis for constructing a predictive model and then solving the problem of load distribution and balancing. This article also shows the features of modeling the traffic of computer networks by fractal Brownian motion. |
|---|