- people - netdevconf.info

0x19: secunet, Bronze Sponsor!
by Jamal Hadi Salim 17 Feb '25

17 Feb '25

We are pleased to welcome back friends of the conference, secunet, as a bronze sponsor for Netdevconf 0x19! secunet is a long time supporter of the conference. We are grateful for their continued support. So who is Secunet? secunet(https://www.secunet.com/) is Germany’s leading cybersecurity company. In an increasingly connected world, the company’s combination of products and consulting assures resilient digital infrastructures and the utmost protection for data, applications and digital identities. secunet specialises in areas with unique security requirements – such as cloud, IIoT, eGovernment and eHealth. With security solutions from secunet, companies can maintain the highest security standards in digitisation projects and advance their digital transformation. secunet is an IT security partner to the Federal Republic of Germany and a partner of the German Alliance for Cyber Security. cheers, jamal

1 0

0x19: Bronze Sponsor, RELIANOID
by Jamal Hadi Salim 13 Feb '25

13 Feb '25

We welcome back long time sponsor RELIANOID, as a bronze sponsor. Thank you for your continued support RELIANOID! RELIANOID simplifies the complexity of the site reliability through network and application load balancing, security, high availability, and identity protection. RELIANOID services are open core and are delivered by a dedicated team of experts with experience built over 15 years. Discover the Site Reliability Experience with RELIANOIDhttps://www.relianoid.com/ cheers, jamal

1 0

0x19: Tentative Schedule Published!
by Jamal Hadi Salim 11 Feb '25

11 Feb '25

Subject says it all: https://netdevconf.info/0x19/pages/schedule.html cheers, jamal

1 0

Zagreb ticket - 11th Feb
by Vincent JARDIN 11 Feb '25

11 Feb '25

Hi Jamal, I was about to book 3 entry pass today since we are still the 11th of February but the early bird entry pass is not available anymore, please can you check ? From the website, they should be available till today : Early Bird Rate Jan 19th - Feb 11th Physical - $250 USD Thank you, Vincent

3 2

0x19: Registration open
by Jamal Hadi Salim 10 Feb '25

10 Feb '25

Folks, Registration for 0x19 is now open! Register early to take advantage of early bird rate. https://netdevconf.info/0x19/registration cheers, jamal

5 6

0x19: Talk, Host-based TCP Connection Migration
by Jamal Hadi Salim 09 Feb '25

09 Feb '25

Current approaches to migrate TCP connections (using TCP_REPAIR) across machines require middle box support for seamless transparency. Most existing proposals use the Intel/Tofino programmable P4 switch for this purpose, an approach which is operationally burdensome. Shuo Li, Steven Chien, Tianyi Gao and Michio Honda describe a proposal that obviates the need for programmable switches. In their approach, the host that has established the connection is responsible for the traffic redirection. They use a combination of TC hardware offload and eBPF to achieve their goals. The authors will go over the implementation details and challenges and finally illustrate how they integrate their connection migration method with Ceph, a widely-used object storage system to illustrate a specific use case. Details: https://netdevconf.info/0x19/10 cheers, jamal

1 0

0x19: Talk, Reservoir to the tap: with fairness and humility
by Jamal Hadi Salim 09 Feb '25

09 Feb '25

Fintech workloads and few other modern work loads require very fine grained control (down to a couple of milliseconds) over packet pacing and at the same time be able to do this for millions of flows; all requiring different granularity of flow pacing. In this talk, Anjali S Jain, Vadivel Kannappa, Joshua Hay, Robert Hathway, Magnus Karlsson, Chihjen Chang, Naren Mididaddi, Priyalee Kushwaha and Milena Olech go into implementation details of how to achieve these requirements. Details: https://netdevconf.info/0x19/32 cheers, jamal

1 0

Communication via Internal Shared Memory (ISM) - Time to open up
by Jamal Hadi Salim 09 Feb '25

09 Feb '25

Internal Shared Memory (ISM) is a method for synchronous, access controlled, direct memory access into dedicated memory regions between systems that share the same physical memory. With increased popularity of using shared memory semantics over the network protocols (see all the AI/ML craze going on), this talk will shed some light on these many years of experience on huge mainframes and other related efforts. In this talk Alexandra Winter will demystify for us ISM, its features and usecases, and explain several proposals that have been made to extend the usage of ISM. Alexandra will further discuss what would be the best way to provide clean interfaces in the Linux kernel, to enable future extensions. An early RFC has been sent to the netdev mailing list with a proposal for an ism layer - to finally make it open to the masses. Details: https://netdevconf.info/0x19/12 cheers, jamal

1 0

0x19: Talk, Optimizing Bandwidth Allocation with ETS and Traffic Classes
by Jamal Hadi Salim 09 Feb '25

09 Feb '25

In the current deployed approach in Linux, the operator does not have much control over how wire bandwidth is allocated between VFs. For example, when provisioned, a 50Gbps PF's equally split between two VFs, giving each 25Gbps, with traffic classes (e.g., TC0 and TC1) limited to percentages of these smaller allocations. Carolina Jubran says implementing ETS hardware support allows for control over such allocations. Using TC bandwidth allocation, the PF can instead be divided directly into 80% for TC0 and 20% for TC1 across the two VFs. In this talk, Carolina will describe the insufficiency of existing approaches to solve these problems, reference previous work, and demonstrate how this implementation improves resource allocation across VFs by applying traffic class-based bandwidth proportions, supported by real-world results and a demo. Details: https://netdevconf.info/0x19/14 cheers, jamal

1 0

0x19: Talk, mq-cake: Scaling software rate limiting across CPU cores
by Jamal Hadi Salim 09 Feb '25

09 Feb '25

Qdiscs rely on global lock to sync state across CPUs and therefore dont scale in presence of many cores (or in presence of very high bandwidth). Jonas Köppeler, Toke Høiland-Jørgensen, and Stefan Schmid implemented a multi-queue variant of sch_cake that can scale its rate limiting across hardware queues (and thus CPU cores) by sharing a bit of state on top of the mq qdisc. In this talk, they will present the implementation, performance evaluation, as well as discuss their proposal for an API that will make this work upstreamable, and applicable to other qdiscs as well. Details: https://netdevconf.info/0x19/16 cheers, jamal

1 0