Radio Cloud Evolution Towards Memory-Driven Computing
Ylänne, Olli (2019)
Ylänne, Olli
2019
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2019120424662
https://urn.fi/URN:NBN:fi:amk-2019120424662
Tiivistelmä
This thesis was done for telecommunications equipment company Nokia. Thesis objective was to evaluate new technology for low latency data storage solution in cloud environment.
Mobile phone network cloudification aims to move mobile phone network computation into cloud data centres. Cloud based mobile phone network (Telco Cloud) products consists of Virtual Machines (VMs) running in data centre servers. Nokia Telco Cloud solution contains a data storage which can be accessed from different VMs. This data storage is known as Shared Data Layer (SDL). Currently SDL is accessed through a TCP/IP network. Certain 5G use cases require ultra-low latency. Data access over TCP/IP network is challenging from latency target point of view.
Memory-Driven Computing (MDC) is a new computer architecture. MDC-architecture contains a shared memory which can be directly accessed from different physical computing nodes. Objective in this thesis was to evaluate the possibility to implement SDL compatible, low latency, data storage in the MDC-architecture shared memory. Such data storage could help to reduce data-access latency in Nokia Telco Cloud products.
Evaluation was conducted following way. SDL and MDC-architecture compatible sharedmemory data storage solutions were designed and implemented. Performance (access latency) and reliability of the implemented solutions were evaluated in MDC-architecture type environment (Hewlett Packard Enterprise Superdome Flex product). Evaluation was performed using Case Study research method. Set of real-world SDL use cases from Nokia Cloud Base Station product were identified. Identified use cases were simulated in evaluation environment for the implemented and existing SDL data storage solutions. Simulation results were analysed, and conclusions were drawn.
SDL compatible data storage was successfully implemented to MDC-architecture shared memory. SDL implemented to MDC-architecture shared memory provided notably lower latency results than the existing SDL implementation in the evaluation environment. However, scope of the evaluated subject is broad, and this thesis studied the subject from a limited scope and further study is recommended to be conducted.
Mobile phone network cloudification aims to move mobile phone network computation into cloud data centres. Cloud based mobile phone network (Telco Cloud) products consists of Virtual Machines (VMs) running in data centre servers. Nokia Telco Cloud solution contains a data storage which can be accessed from different VMs. This data storage is known as Shared Data Layer (SDL). Currently SDL is accessed through a TCP/IP network. Certain 5G use cases require ultra-low latency. Data access over TCP/IP network is challenging from latency target point of view.
Memory-Driven Computing (MDC) is a new computer architecture. MDC-architecture contains a shared memory which can be directly accessed from different physical computing nodes. Objective in this thesis was to evaluate the possibility to implement SDL compatible, low latency, data storage in the MDC-architecture shared memory. Such data storage could help to reduce data-access latency in Nokia Telco Cloud products.
Evaluation was conducted following way. SDL and MDC-architecture compatible sharedmemory data storage solutions were designed and implemented. Performance (access latency) and reliability of the implemented solutions were evaluated in MDC-architecture type environment (Hewlett Packard Enterprise Superdome Flex product). Evaluation was performed using Case Study research method. Set of real-world SDL use cases from Nokia Cloud Base Station product were identified. Identified use cases were simulated in evaluation environment for the implemented and existing SDL data storage solutions. Simulation results were analysed, and conclusions were drawn.
SDL compatible data storage was successfully implemented to MDC-architecture shared memory. SDL implemented to MDC-architecture shared memory provided notably lower latency results than the existing SDL implementation in the evaluation environment. However, scope of the evaluated subject is broad, and this thesis studied the subject from a limited scope and further study is recommended to be conducted.