PhD Position: In-Band Network Telemetry Management in Fast Packet Processing Architecture

Team

The PhD position is proposed by the RESIST team of the Inria Nancy Grand Est research lab, the French national public institute dedicated to research in digital Science and technology. The team is one of the European research group in network management and is particularly focused on empowering scalability and security of networked systems through a strong coupling between monitoring, analytics and network orchestration. https://team.inria.fr/resist/

Contacts: Jérôme François (jerome.francois@inria.fr) and Isabelle Chrisment (isabelle.chrisment@loria.fr)

Context

Network monitoring is a critical network operations in order to get data about network performance, security, QoS, etc. Over the years, there have been multiple methods which have been proposed and developed. Although full capture of network packets is not scalable, a main protocol and architecture which have been promoted and highly deployed are NetFlow/IPFIX. In general flow-based monitoring provides higher scalability by aggregating data over multiple packets to create flow metrics. Recently, SDN (Software Defined Networking) empowers the programmability of network actions including monitoring. It is now possible to dynamically program fine-grained and even stateful monitoring.

Recently, In-Band Network Telemetry (INT) was proposed [1]. Rather than collecting single point observations, the objective of INT is to collect data along the path a packet is forwarded through. For example, it is possible to collect information such as latency, or queue sizes for each hop. On one hand, it enables advanced monitoring scenarios where end-user services and application and can take decisions based on their own context and network conditions, for example changing the video quality or using another CDN server during streaming. Although such kind of applications have been already considered thanks to end-to-end metrics like latency or jitter, INT is expected to provide a better view on network conditions. As an example, finding and locating bottleneck in a network is difficult today unless you have multiple vantage points.

On the other hand, INT can be considered as critical and can also serve attackers to better plan their attacks, for example by targeting the bottleneck nodes. Access to INT data must be protected. Lightweight encryption can be used [2] but adds computational overhead. In general, access to telemetry data need to be controlled because data might be shared by multiple actors (such as telcos and data center operators) assuming some agreement between them. With the flexibility of network and service today, such agreement can be dynamic and we propose in this thesis to add access control mechanism over telemetry information. However, in-band network telemetry information is produced and transferred along packets and so must be handled at the dataplane till a sink node that can report metrics possibly aggregated to a server. So, performance is a major requirement for doing in-network INT operations.

[1] Changhoon Kim, Anirudh Sivaraman, Naga Katta, Antonin Bas, Ad-vait Dixit, and Lawrence J Wobker. In-band network telemetry via programmable dataplanes. ACM SIGCOMM Symposium on SDN Research(SOSR), 2015

[2] X. Pan, S. Tang, S. Liu, J. Kong, X. Zhang, D. Hu, J. Qi, Z. Zhu, Privacy-Preserving Multilayer In-Band Network Telemetry and Data Analytics: For Safety, Please do not Report Plaintext Data, Journal of Lightwave Technology 38 (21) (2020) 5855–5866. doi:10.1109/JLT.2020.3007491.

Subject description

This thesis will define and develop a framework for INT which fulfill the following requirements: allow multiple stakeholders to share their telemetry information, enable a dynamic access control mechanism (to add, delete or protect information at network borders) and leverage hardware-accelerated fast packet processing techniques. So, the PhD candidate will have to investigate existing distributed access control mechanisms taking into account the limited capacities of network nodes (even with hardware acceleration) and considering either in-band or out-of-band control. In addition, we will also consider sink nodes whose the role is to aggregate and report data. For aggregation multiple, stateful information might be kept and modified although intermediate nodes usually add and transfer information. We propose that aggregation at sink nodes can be also programmed and shared by each INT users. In that case, access control over stateful data must be also set but other problems arise because programs consuming such data are more complex and isolation between them are necessary. Fast packet processing platforms provide such mechanisms by design but the PhD candidate will have to research on how to leverage such mechanism while still sharing stateful information, for example shared memory with also access control mechanism at sink level.

For developing the solution, the team has access to different hardware-accelrated platforms: P4 switches, SmartNIC programable with P4 or XDP and NetFPGA. Evaluating limitations and advantages of each platform from the state of the art will help to assess the feasibility of the proposed solutions and select one platform for real testing.

Required qualifications
• Required qualification: Master in computer science
• Required knowledge: networking, programming (C, C++, python)
• Knowledge and skills in the following fields will be appreciated: network programming (P4, NetFPGA, XDP)

Application deadline
May 31, 2021 (Midnight Paris time)

How to apply

Upload your file and send it as well by email to jerome.francois@inria.fr and isabelle.chrisment@loria.fr. Your file should contain the following documents:
• Your CV.
• A cover/motivation letter describing your interest in this topic.
• A short (max one page) description of your Master thesis (or equivalent) or of the work in progress if not yet completed.
• Your degree certificates and transcripts for Bachelor and Master (or the last 5 years).
• Master thesis (or equivalent) if it is already completed and publications if any (it is not expected that you have any). Only the web links to these documents are preferable, if possible.

In addition, one recommendation letter from the person who supervises(d) your Master thesis (or research project or  internship) should be sent directly by his/her author to jerome.francois@inria.fr and isabelle.chrisment@loria.fr.

Applications are to be sent as soon as possible.

 

Logo d'Inria