NoSQL stores for coreless mobile networks

Abstract

The goals of 5G networks-low latency, high bandwidth, and support for fast mobility-are non-trivial and they demand improvements across all involved technology fields. Researchers are therefore exploring approaches that leverage on network function virtualization and software-defined networking for meeting the demands of verticals expected to use 5G networks. One approach which appears promising is the concept of a coreless mobile network where the key network functions are placed at the edge of the network. In this article we focus on management of the user-context state in a coreless mobile network, and posit that these network functions can use a NoSQL data store for maintaining the user-context and other state variables. We first present an overview of promising NoSQL data stores and evaluate their suitability. We then present the results of benchmarking the Apache Geode data store as an example of a state management solution which could serve a coreless mobile network. During our tests we observe that the Apache Geode data store is, subject to its configuration, capable of delivering the data model, consistency, and high availability required by a coreless mobile network.

Nervion

“…The core is a key mobile network component that handles the essential control functionality such as mobility, session and security management, and also bridges data communication between end devices and external networks like the Internet via the radio access network (RAN). Driven by the limitations of the traditional mobile network architecture and the wider trends in the mobile network industry, such as embracing network softwarization and supporting diverse use cases, mobile core networks have received a great deal of attention from the standardization bodies [1][2][3][4] as well as research [5][6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21] and open source [22][23][24][25][26] communities.…”

“…RAN emulation has become the commonly used mobile core system evaluation methodology, as reflected by [6][7][8][9][10][11][12][13][14][15][16][17][18][19][20][21], compared to other alternatives like testbed-based experimental evaluation, simulation and mathematical analysis. The general idea behind RAN emulation is to emulate the RAN elements -base stations and end devices (UEs) -from the perspective of the core system under test.…”

Data Consistency in the 5G Specification

“…Along with the UDM we will introduce Brewer's CAP theorem and explain the need to circumvent it. Finally, this section reviews the work of Ojala et al [5] on managing the state of a 4G network in a distributed database, also subject to the CAP theorem.…”

“…al. [5] has discussed a distributed Core Network that relies on a distributed NoSQL database to provide the UE Context. Their work highlights the impossibility of providing strong consistency for a geo-distributed database and makes the trade-off of a weakly consistent distributed database with high consistency for geographical zones that are centered around a given UE.…”

“…Weak consistency vs. strong consistency hybrid databases, while correct, could be inefficient for some of the data involved in the 5G features, such as Network Slicing. In particular, Ojala et al [5] center high consistency zones around the geographical location of the UE, meaning that all information relevant to the user must be around him. But in the instance of network slicing a network slice treats a group of users as a single entity with geodistributed data.…”