Over the last few years, NIC speeds have gone up
from 1G to 10G to 25G to 40G and now 100G and 200G
and even 400G NICS are going to be showing up in
the commodity market soon.
Every time these rates go up, new challenges come
to the surface in the kernel. It is imperative to
understand those issues and resolve them.
In this talk Tariq Toukan will review the main bottlenecks
that were/are being addressed to support NIC speeds of
100Gbps, 200Gbps, and beyond.
He will summarize several ideas and solutions ranging
from traditional tricks of the trade like page recycling
to innovative HW capabilities that helped to recently
achieve milestone packet rate of 100 Million PPS in
XDP_TX and 135Mpps in XDP_DROP.
More info:
https://netdevconf.org/0x13/session.html?talk-bottlenecks
A reminder to all:
Early bird registration is still open until Feb 20th.
https://netdevconf.org/0x13/registration.html
cheers,
jamal
The Linux kernel has successfully made its way across
datacenter top-of-rack and spine switches.
David Lamparter says the data centre is a friendly
sane environment. You aint seen nothing yet.
Campus distribution and access switches need
to deal with a much more hostile environment.
In this talk he intends to collate the currently relevant
features for deploying campus and access environments.
He will give some operational context, and suggest
implementation strategies.
https://netdevconf.org/0x13/session.html?talk-linux-switch-req-feat
A note to all:
Early bird registration is still open until Feb 20th.
https://netdevconf.org/0x13/registration.html
cheers,
jamal
Nutsnbolts talk.
Olga Albisser et al will describe the DUALPI2 AQM qdisc.
DualPI2 AQM is part of the IETF L4S infrastructure
standardization.
An initial patch was posted on the list.
The primary goal of DUALPI2 AQM is to provide ability
to deploy congestion controls like DCTCP (which is
traditionaly deployed in data centres) on the internet.
In this talk, Olga will present the details of DUALPI2
implementation and explain problems it solves.
More info:
https://netdevconf.org/0x13/session.html?talk-DUALPI2-AQM
A note to all:
Early bird registration is still open until Feb 20th.
https://netdevconf.org/0x13/registration.html
cheers,
jamal
Nutsnbolts talk.
Multi-access edge computing (MEC) was intended
as a network architecture that offers cloud-computing
capabilities within the RAN or core network in the
cellular world to allow processing of tasks closer to
the cellular customer. It has, however, evolved to be
generic to apply to any network for deploying
applications and services as well as to store and
process content in close proximity to mobile users.
Current MEC approaches are too invasive.
In this talk, Feng et al propose an
alternative light weight mobile edge computing
architecture which utilizes existing Linux
kernel mechanisms, namely: Traffic Control (TC)
utilities, and network Namespaces.
The talk will go into details on how the Linux facilities
are used and challenges encountered.
https://netdevconf.org/0x13/session.html?talk-viscum
cheers,
jamal
Accepted talk.
Aaron Conole and Marcelo Leitner want to make
it easier to integrate TC offload with conntracking.
This talk will describe their efforts to integrate
ConnTrack with TC in software datapath. They will
describe their implementation approach, the challenges
that they had to overcome, testing approach taken and
finally some preliminary performance numbers.
In the second phase of the talk they will cover
eBPF extensions for integrating with ConnTrack
and present performance numbers.
https://netdevconf.org/0x13/session.html?talk-conntrack-tales
cheers,
jamal
Nuts and bolts Talk.
The FRR (Free-range Routing) software suite is
going through a revamp. There is a move to an
async threaded model which is planned to utilize
new netlink features to improve scalability. This
model will also allow programming FIBs and NHs
on remote systems.
In this talk Mark Stapp will describe the current
state of the work and challenges involved in achieving
this goal. Some initial results and future plans
will be discussed.
https://netdevconf.org/0x13/news.html?talk-FRR-async
cheers,
jamal
Cloudlflare is back to describe their latest
DDoS mitigation system. Many lessons learnt.
In this talk, Arthur Fabre will focus on the
implementation of the Cloudflare XDP solution.
Perf-based packet sampling is used to detect
the attack. The followup mitigation subsystem
automatically generates eBPF code in response
to attacks.
https://netdevconf.org/0x13/session.html?talk-XDP-based-DDoS-mitigation
cheers,
jamal
Nuts-n-bolts Talk.
A set of changes to DCTCP that would make it
deployable over the public Internet are
dubbed the 'TCP Prague requirements'.
Bob Briscoe et al will introduce the important
components of the first 'TCP Prague' implementation
which runs on Linux.
They will provide rationale and explanation on
how the parts integrate together, both within each
system and at Internet scale; they will outline
the maturity of each component implementation, plus
some points of interest and issues where decisions
are still needed.
https://netdevconf.org/0x13/session.html?talk-tcp-prague-l4s
cheers,
jamal
New accepted Nuts-n-bolts talk.
Many network middleboxes either muck or block
all UDP traffic; this includes IKE and IPsec.
They are, however, happy to allow TCP connections
through because they appear to be web traffic.
Sabrina will describe to get the middle boxes happy
with TCP by introducing encapsulation over standard
TCP connections based on RFC 8229. She will further
describe the implementation approach which utilizes
existing kernel infrastructure (TCP upper layer protocol
mechanism (ULP) and stream parser), and finaly how it
can be used by userspace IKE daemons.
https://netdevconf.org/0x13/session.html?talk-ipsec-encap
cheers,
jamal
The PC has accepted another nuts-n-bolts talk.
With the addition of NVMe over Fabrics (NVMe-oF)
on Linux (using TCP) - there is that elephant in
the room question that needs to be answered.
Is there a performance impact when using TCP
instead of an RDMA transport such as RoCE?
Ariel Cohen has some answers. In his talk
he will describe a performance study comparing
NVMe/TCP to NVMe/RoCE on Linux. Throughput, latency
distribution, and CPU load are measured and compared. A
performance comparison of these kernel protocols to
user-level protocols using SPDK will also be described.
https://netdevconf.org/0x13/session.html?talk-nvme-tcp-roce
cheers,
jamal