Publication date: 21/10/2022

Namesort descending Role

Examining board:

Namesort descending Role

Summary: Despite hardware middle-boxes being an integral part of modern enterprise and telecommunication networks, they
are usually proprietary, have little or no programmability and, when they exist, this programmability is restricted to
traditional network functions, like firewall, NAT and others. Furthermore, they are hard to vertically integrate with
other packet processing elements, such a closed and inflexible ecosystem partly explains the high capital and
operational expenditures incurred by network operators. This led to the Network Function Virtualisation movement
that proposes to disaggregate the tightly coupled Network Functions and hardware middle-boxes and deploy Virtual
Network Functions on commodity servers, promising to reduce CAPEX by consolidating multiple Network Functions
on the same hardware, and reduce OPEX by enabling on-demand flexible service provisioning.
Thus, there is room for exploiting this market by using state-of-the-art network programmability. In-Network
Computing (INC) is a promising field that aims at exploiting the capabilities of programmable network devices, such
as programmable switch ASICs and programmable network interface cards (SmartNICs), to offload computing from
embedded hardware to the network. With this, new enabling functionality can be achieved using a coordinated and
well-defined framework to deliver to applications the liberty to control not only network-related behaviours but also
using the network as an offloading facility, increasing the expressiveness and the use of the resources present on
programmable network devices.
The thesis introduces a novel framework proposal for prototyping and implementing in-network computing using
multiple techniques for selecting and steering data flow between software and hardware applications. A framework
(PIaFFE) that uses P4 language for decomposing and deploying Virtual Network Functions (VNFs) into small
embedded Network Functions (eNFs) on in-network processors is created, developed, deployed and evaluated to
allow the flexible embedding of network applications. PIaFFE expressiveness is demonstrated through multi-domain
use-cases, from micro-applications and network function virtualisation approaches to passive optical networks and
cloud robotics technologies, shifting the network traffic offloading and processing between the software and the
network devices in an automated fashion, reducing network latency and CPU usage on commodityservers, and also
increasing throughput.

Access to document

Acesso à informação
Transparência Pública

© 2013 Universidade Federal do Espírito Santo. Todos os direitos reservados.
Av. Fernando Ferrari, 514 - Goiabeiras, Vitória - ES | CEP 29075-910