Another exciting talk accepted.
In this talk titled "LSDN - Manage complex (virtual) networks in cloud
environment with Linux kernel facilities" Vojtech Aschenbrenner et al
present LSDN, a tool for effortless management of complex (virtual)
networks.
Often management of virtual networks has dependencies on complex
infrastructure orchestration tooling - encumbered with long
running agents like OVS etc. LSDN has no external dependency on
any such tools and leaves no agents running thus not only improving
usability but also robustness of the operations.
The tool relies entirely on the functionality provided by the Linux
kernel, mainly on the Traffic Control (TC) subsystem with its recent
features.
Vojtech will describe LSDN features and challenges faced in its
development (and how various kernel and iproute2 issues were addressed);
then delve into how to write applications centred around the offered
C-library and backends. Vojtech will conclude by talking about the
future plans for LSDN.
More info:
https://www.netdevconf.org/0x12/session.html?lsdn-manage-complex-virtual-ne…
cheers,
jamal
Another exciting talk from the program committee
Amen to the Whitebox! Whitebox! Whitebox! chants. But the emerging creed
of "Open" networking operating systems (NOS) is sadly still plagued with
ASIC vendor SDKs and the proprietary and legacy baggage they carry.
Vendor SDKs quarks impact control performance and dictate how control
and datapath APIs look like. The SDK glue creates undue burden on
operational interfaces design: config, debug-ability, and monitoring all
suffer just so the NOS can adapt to the many every vendor SDKs.
The switchdev and related hardware offload interfaces revolution in
Linux have matured over the last few years. They provide a singular,
cleaner interface to ASIC features. No different than other hardware
offloaded by Linux - the hardware vendor provides a driver and leaves
the innovation to the NOS and applications.
Shrijeet Mukherjee and David Ahern have been in the dragons den of
vendor SDKs and have come out scathed but wiser.
In this talk they will discuss typical software architectures for
a NOS and introduce a path for transitioning SDK based solutions to
Switchdev and related offloads.
More info:
https://www.netdevconf.org/0x12/session.html?building-a-better-nos-with-lin…
cheers,
jamal
PS: Another reminder, early bird registration (currently 20% off will
go up on June 1st).
Is TCP - at 44 - too old, slow and has bad habits that cant be fixed?
Does shaving a few milliseconds of latency on a web transaction matter
that much to call out for a new transport? The new kid on the block,
QUIC thinks so. QUIC has been globally deployed at Google on thousands
of servers. If you are running chrome as a browser and using YouTube,
you are likely using QUIC. Estimates are 7% of the internet traffic
is now QUIC.
In this talk Jana Iyengar and Ian Swett will provide an overview of
QUIC, its motivations and the performance improvements observed at Google.
They will discuss performance bottlenecks in Linux kernel UDP that
were seen at Google's servers, what was done to overcome them and
what work remains to be done in the kernel as QUIC becomes more
widely deployed.
More info:
https://www.netdevconf.org/0x12/session.html?developing-and-deploying-a-tcp…
cheers,
jamal
PS: Another reminder, early bird registration (currently 20% off will
go up on June 1st).
First Workshop announcement!
Jamal Hadi Salim will chair a working session with face to face
discussions on different Traffic Control topics.
https://www.netdevconf.org/0x12/session.html?tc-workshop
cheers,
jamal
PS: Speaking in the third person is allowed.
Reminder again: Early bird registration closes on June 1st.
The program committee has accepted another exciting talk!
Suricata is a widely deployed open source IDS/IPS/NSM.
Suricata 4.1 will include eBPF and XDP support. In this respect,
Suricata is amongst pioneers to deploy these technologies in a
production environment. The journey to get to eBPF and XDP support
and remove the cobwebs took about 2 years. Eric Leblond will describe
some of the challenges met and how they were addressed to get to
production level quality.
More info:
https://www.netdevconf.org/0x12/session.html?suricata-an-xdp-adventure
cheers,
jamal
PS:- Another reminder that:
1) Registration early bird deadline (before prices go up) is June 1st.
2) We do offer bursaries for attending the conference.
Cumulus Networks has been a solid supporter of the Netdev conference
since 0.1!
We appreciate and welcome them back as a Gold Sponsor for Netdev conf
0x12.
https://www.netdevconf.org/0x12/news.html?gold-sponsor-cumulus
cheers,
jamal
Again reminder:
1) June 1st is the deadline for early bird cheaper registration fees.
2) We do have bursaries for folks who need help getting to the
conference.
In this 1.5 hour instructor-led tutorial, Donald Sharp(FRRouting
maintainer) will start by introducing the FRRouting Suite to the
unitiated.
He will then guide us into a basic network configuration.
In the second half of the tutorial, Donald will delve deeper
into more advanced features on how FRR fits into the classic Data
Centre architecture.
https://www.netdevconf.org/0x12/session.html?introduction-to-frrouting
cheers,
jamal
PS:- Reminder: Early bird registration offer ends June 1st.
Folks,
Nathan Jay et al assert that there is no one-size-fits-all TCP
congestion control algorithm.
Different apps have different goals for their service lifetimes:
Streaming video apps are sensitive to bandwidth fluctuations,
voice chats desire low latency and bulk transfers only care about
completion time. Varying link characteristics add another dimension
to a congestion control algorithm: hardwired assumptions about the
cause of packet loss or measured latency inflation can lead to
reduced application performance.
PCC-Vivace was created to address these issues.
PCC-Vivace congestion control algorithm provides an explicit
utility function that allows developers to provide weights to
different performance metrics like throughput, latency, packet
loss and jitter. PCC-Vivace's online learning framework also
allows it to adapt to a variety of network conditions,
consistently delivering high performance.
The talk will discuss the challenges of implementing PCC-Vivace
in the Linux kernel. Nathan will present initial results comparing
the performance of Vivace with other existing Linux congestion
controllers(BBR etc). The talk will also demonstrate the
implementation's flexibility by creating and testing a variety of
utility functions.
More info at:
https://www.netdevconf.org/0x12/session.html?a-pcc-vivace-kernel-module-for…
cheers,
jamal
HAProxy is a well known high performance TCP/HTTP load balancer used
as a front end to applications by many very-large sites. Often
HAProxy is stationed at the frontiers of the wild wild ^Wwest
internet in the defense against DDOS - and at the heart of the
storm sits PacketShield. PacketShield is built using NDIV. NDIV
has been deployed very effectively todate and handles the intended
line-rate 10Gbps processing under a variety of hostile real-world
DDOS attacks.
Willy Tarreau and Emeric Brun have been contemplating migrating
NDIV to use XDP. In this talk they will describe what NDIV does
well and what is currently lacking with XDP to make it convincing
to make the jump.
More details:
https://www.netdevconf.org/0x12/session.html?challenges-migrating-from-ndiv…
Reminder: Early bird registration expires on June 1st.
cheers,
jamal