Abstract
In recent years, as a result of the proliferation of non-elastic services and the adoption of novel paradigms, monitoring networks with high level of detail is becoming crucial to correctly identify and characterize situations related to faults, performance, and security. In-band Network Telemetry (INT) emerges in this context as a promising approach to meet this demand, enabling production packets to directly report their experience inside a network. This type of telemetry enables unprecedented monitoring accuracy and precision, but leads to performance degradation if applied indiscriminately using all network traffic. One alternative to avoid this situation is to orchestrate telemetry tasks and use only a portion of traffic to monitor the network via INT. The general problem, in this context, consists in assigning subsets of traffic to carry out INT and provide full monitoring coverage while minimizing the overhead. In this paper, we introduce and formalize two variations of the In-band Network Telemetry Orchestration (INTO) problem, prove that both are NP-Complete, and propose polynomial computing time heuristics to solve them. In our evaluation using real WAN topologies, we observe that the heuristics produce solutions close to optimal to any network in under one second, networks can be covered assigning a linear number of flows in relation to the number of interfaces in them, and that it is possible to minimize telemetry load to one interface per flow in most networks.
Highlights
Monitoring is an essential component of network operation and management tasks
We introduce the In-band Network Telemetry Orchestration (INTO) problem, which is focused on optimizing the use of network resources for INT
A procedure to solve any variation of the INTO problem should be able to provide high-quality solutions within short time intervals. This is necessary so that the monitoring campaign can quickly adapt to network policy changes and traffic fluctuation. In this first set of experiments, we evaluate the solution quality and processing time of the algorithms introduced in Section 4, namely Concentrate Heuristic (CH) and Balance Heuristic (BH), by comparing them to the mathematical programming models
Summary
Monitoring is an essential component of network operation and management tasks. In recent years, monitoring networks with high level of detail (e.g., per-packet hop-by-hop delays, instantaneous queue size) is becoming crucial to correctly identify and characterize network events related to faults, performance, and security. A new monitoring concept has been proposed: In-band Network Telemetry (INT) [10,11,12] It makes use of new capabilities of emerging programmable switches [7, 9, 13] to encapsulate processing metadata information (e.g., queue occupancy, processing time, policy rules) into “production traffic” packets (i.e., packets originated at the application layer). INT brings new opportunities regarding accuracy and level of detail, to be able to use it in practice, there is the need to understand the trade-offs between quality and costs involved in employing it Executing this type of telemetry involves modifying production packets traversing the network, which may significantly degrade the performance of end-user applications.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
