Towards a Monitoring Protocol Over Information-Centric Networks

Abstract

Managing data centers remains an important challenge despite the extensive research in this area. As the future prospective of networking is location-agnostic networks as in Information Centric Networking (ICN), some problems are solved but new challenges arise. When a data center runs over an ICN network, how can it discover failing replicas, monitor the load on running replicas, or decide when and where to instantiate new replicas? All these questions are hard to answer when the network is purely location-agnostic. In this paper we discuss the problem of pinpointing replicas and locations on ICN networks. We propose a draft of a protocol, Name-Centric Monitoring Protocol (NCMP), the use of which makes it possible to scan for replicas serving a named entity (content or service) and communicate with each one individually. A main application of NCMP is to monitor data centers and pull load information from service replicas.

Load Balancing in the Presence of Services in Named-Data Networking

“…There is a growing interest in the applicability of NDN architectures in data centers. Researchers acknowledge resource monitoring and load balancing as main challenges in this setup [96][97][98][99][100][101]. Although research in this area is at its preliminary state, we can still distinguish two main trends.…”

“…7 An example of data center over pure NDN. In NCMP [101], the controller sends management messages to gather information about the capabilities of service replica resources. During the session, the controller gets all replica's IDs and then uses them to get information about a specific metric from a specific replica…”

“…In an earlier work, we propose a Name-Centric Management Protocol (NCMP) [101] as a monitoring protocol that works in a pure NDN. NCMP aims to scan different replicas serving the same named content or service to gather information about the load and liveliness of these replicas.…”

“…It indicates to which extent an approach is compliant with the basic NDN architecture. The second aspect is the possibility to factor out the load balancing task to an external entity like the controller in IC_DCN [96] or the monitor in NCMP [101]. We elaborate further on these aspects in Sect.…”

“…Solutions such as SoCCeR [69] and PAF [71] constantly apply network probing in all routers, introducing a considerable traffic overhead and limiting network capacity. In terms of load balancing, this probing should be done only in the network components responsible for load balancing, in a similar fashion to NCMP [101]. This way, the network is not flooded with control messages and metrics are collected only where they are consumed.…”