ELTE IK fejléc
Navigációs hivatkozások átugrása
About us
EIT ICT Labs Master School
International application
Navigációs hivatkozások átugrása
BSc programmes
MSc programmes
PhD School
Postgraduate specialist training courses
Higher-level vocational training
International Studies of ELTE
5G Edge Computing Programming Assessment
A Novel Toolset of Complex Machine Learning Solutions for segmented and fully distributed computing environments
Vendor neutral data-plane solutions for software-defined networking
Intelligent Test Traffic Generator for Telecommunications Networks

Vendor neutral data-plane solutions for software-defined networkin

The EIT Digital Doctoral School offers an Industrial Doctorate position in the field of software-defined networks. The doctoral candidate will seek to develop vendor neutral data-plane solutions for software-defined networking. The work will be carried out with the support of EIT Digital Partners Ericsson Hungary and the Eötvös Loránd University (ELTE), in Budapest.

The thesis will create generic, efficient and verifiable solutions applicable to hardware independent programming of diverse networking device architectures.

This will be based on novel models of current industrial use-cases for software defined networking (SDN) data-plane programming. With a specific focus on domain-specific programming languages (e.g., the P4 language), the novel models will be derived from the analysis of high-level solutions applied in SDN. The work will be aligned with the emerging standards and test suites, with specific reference to those produced by ETSI to guarantee the application to the market.


As of now, most effort in SDN is still narrowly focused on control-plane programmability. Therefore, there is a general lack of know-how (in both industry and academia) on programming the data-plane. Industry trends seem to slowly recognize that freely programmable packet transmission shifts network life-cycle management from hardware to software level, enabling enterprises to automatize, optimize and adapt their packet forwarding systems more efficiently and reliably. 

Early stage data-plane programming solutions are available and must be mentioned, but each of these solutions are centered around one narrow use-case each. Examples are Network Function Virtualization, Openflow extensions, and OpenFastPath.

Obviously, the more use-cases a generic domain specific language (DSL) for SDN can cover, the more tasks efficient networking devices can be used for, without extra costs and development efforts.

Without a generic DSL on other hand, producers/developers struggle with patching together solutions pushed by various vendors. This implies a niche for data-plane solutions that harmonize the various potential industrial use-cases of next generation SDNs to leverage and serve the accelerating demand for a generic, data-plane language.

One use-case pool to drive the development of such a solution includes the use cases of Data Center Gateway, Mobile Gateway, Broadband Network Gateway, the already mentioned NFV (see eg. NetBricks) and OpenFastPath, service oriented 5G networks, security-aware switches, multi-domain DSLs, IOT possibilities, IaaS Cloud (eg. OpenStack, Kubernetes), and possibly more.


One of the goals is to investigate how high-level solutions can be realized efficiently — compared to the performance of low-level, hardware-specific solutions —, while preserving their high degree of expressivity and usability. A subproblem is the evaluation of optimization possibilities of task scheduling using static and/or dynamic analysis methods.

One DSL for SDN, P4, promises liberation of SDN data-plane programming from hardware-specific limitations. P4 is backed by the industrial giants of communications (AT&T, Cisco, Google, VMware, and many more), a growing open-source community, and several large universities (eg. Stanford, Princeton).

It is still unclear, what exact industrial use-cases P4 is capable of addressing, how much effort is needed to adapt P4 to effective use-cases, and whether this pool is large and diverse enough to make P4 the DSL for generic data-plane programming. Since P4 already has some use cases, there's a lower bar obviously, the question is how high we can raise it.

Another closely related question is whether a more generic language is needed, and how it can be adequately defined.

Research focus/topics

In stage one we assess and define concrete use-cases, based both on industrial input and on existing literature. Questions of scope, feasibility, importance, relevancy must be considered.

Next level is use-case modelling, e.g., creating a minimal prototype by emulating use-cases (e.g., Data Center Gateway) using x86 processor, and measure, e.g., lower and upper boundaries for execution time, delay and jitter. We then evaluate the accuracy of the model by comparing our results to current research literature, and then refine the model in iteration, until sound accuracy is reached. After successful modelling, we implement use-cases using a selected DSL, e.g., P4. Based on efforts and concrete conclusions we can then propose well-founded DSL improvements if needed.

We expect that the acquired know-how can be effectively converted into pure software code, evolving into a prototype decision support system. Ultimately, such a system links complex use-case definitions together with existing knowledge about network HW specs, yielding key metrics and qualitative data (hardware acceleration, P4 extern actions, etc.) on efficient HW/SW separation. Using the data, users can take advantage of generic P4 expressiveness, and at the same time deploy expensive high-performance HW drivers where it is needed most.

Expected outcome

A toolset of effective vendor neutral data-plane solutions for software defined networking. The toolset will address a set of SDN data-plane programming use-cases of industrial applications.

The solutions provided will lift data-plane programmability constraints by enabling synchronous modular deployment of reusable, portable, scalable software solutions, and high-performing HW-specific configurations.

The toolset will be supported by a semi-automated solution selector (i.e., a sort of decision support system) helping the application of the various tools. This will be based on an extensive dataset collected about the model runs of the addressed industrial use-cases on specific hardware architectures.


The position is based in the Doctoral School Training centre in Budapest where a strong ecosystem for Digital Infrastructure exists.


  • Industrial partner: Ericsson Hungary
  • Academic/research partner: Eötvös Loránd University (ELTE)
  • DTC location: Budapest
  • Number of available PhD positions: 1
  • Duration: 4 years
  • This PhD will be funded by EIT Digital, Ericsson, and the Eötvös Loránd University (ELTE).

The person applying for this position should be already enrolled in the Doctoral School of ELTE, and should have started his PhD studies in 2017.


Those interested in applying should send a letter of interest to Zoltán Istenes (zoltan.istenes@eitdigital.eu EIT Digital – Budapest Doctoral Training Centre Lead).

Please apply before October 15, 2017.