## Accelerate Istio-CNI with ebpf Xu Yizhou & Guo Ruijing ## Agenda • Istio-CNI • tcp/ip stack overhead between sidecar and service • Background knowledge of ebpf • Acceleration for Inbound/Outbound/Envoy [Image](/uploads/documents/5/a/b/b/5abb1b8f1b8f9d74adba9f84c56cea7a/p3_1.jpg) ## Tcp/ip stack overhead between sidecar and service Overhead sidecar traffic from 3 scopes • Inbound • Outbound • Envoy to Envoy(same key may conflict due to same src/dst ip address  ## Use pod ip as hash key Use pod_ip to generate a unique key is a way to

## Using C libraries in your Modern C++ Embedded Project EtherCAT Stack, C++17, and Modern C++ Idioms  ciere consulting 4e/p13_2.jpg) ## Outline Beckhoff SSC extern "C" Beckhoff SSC - Slave Stack Code “The EtherCAT Slave Stack Code (SSC) is a code written in ANSI C. Its modular and simple structure enables fast THINGS WE DON'T PUT ON SLIDES } } extern "C" { // This is the main loop in the SSC stack. void MainLoop(void); } int main() { init(); while (1U) { // The SSC main loop

## +24 ## Hidden Overhead of a Function API ## OLEKSANDR BACHERIKOV ## What we do at Snap with C++  Neural style transfer add rsp, 16 mov eax, ebx pop rbx ret mov rbx, rax jmp .15 #### x64 msvc v19.40 VS17.10 ## Stack unwinding on exception push rbx sub rsp, 32 lea rcx, QWORD PTR ptr[$rsp] call value_ptr() mov the function call is avoided Cons (unless we fully trust the user and don’t have exceptions): • Stack unwind on exception is still necessary • Extra bool flag is required to know if the object ☐ was

## Designing an ultra low-overhead multithreading runtime for Nim Mamy Ratsimbazafy mamy@numforge.co ## Hello! ## I am Mamy Ratsimbazafy During the day blockchain/Ethereum 2 developer (in Nim) During space Hardware and software multithreading: definitions and use-cases Parallel APIs Sources of overhead and runtime design Minimum viable runtime plan in a weekend ## Understanding the design space futures? (in Nim “Flowvar” to distinguish from IO-tasks futures) Synchronization - Scheduling overhead Thread-safe memory management ## Dataflow parallelism Alternative names - Pipeline parallelism

Web on Reactive Stack Version 5.3.36-SNAPSHOT ## Table of Contents 1. Spring WebFlux ..... 2 1.1. Overview ..... 2 1.1.1. Define "Reactive" ..... 2 1.1.2. Reactive API ..... 3 1.1.3. .. 173 Reactive Libraries ..... 176 This part of the documentation covers support for reactive-stack web applications built on a Reactive Streams API to run on non-blocking servers, such as Netty, Undertow reactive WebClient, support for testing, and reactive libraries. For Servlet-stack web applications, see Web on Servlet Stack. ### Chapter 1. Spring WebFlux The original web framework included in the Spring

Web on Servlet Stack Version 5.3.36-SNAPSHOT ## Table of Contents 1. Spring Web MVC ..... 2 1.1. DispatcherServlet ..... 2 1.1.1. Context Hierarchy ..... 4 1.1.2. Special Bean Types ..... 7 1 251 5.5. Further Resources ..... 252 This part of the documentation covers support for Servlet-stack web applications built on the Servlet API and deployed to Servlet containers. Individual chapters Spring MVC, View Technologies, CORS Support, and WebSocket Support. For reactive-stack web applications, see Web on Reactive Stack. ### Chapter 1. Spring Web MVC Spring Web MVC is the original web framework

## How and When You Should Measure CPU Overhead of eBPF Programs eBPF Summit ## Why should I profile eBPF programs? ## CI variance tracking ●●●● name TCPLatency/eBPF/kprobe/sys_bind TCPLatency/eB TCPLatency/eBPF/kprobe/sys_socket TCPLatency/eBPF/kprobe/tcp_cleanup_rbuf TCPLatency/eBPF/kprobe/tcp_close TCPLatency/eBPF/kprobe/tcp_sendmsg time/op 650ns ±15% 1.00μs ±23% 256ns ±15% 1.84μs ±13% 605ns ±13% Benchmarking + CI/CD – Sampling profiler in production ## How does it work? ## - Adds ~20ns of overhead per run ☐ ☐ ☐ // pseudo-code if (bpf_stats_enabled) { u64 start = sched_clock(); run_ebpf_program();

## +23 ## I s std::mdspan a Zero-overhead Abstraction? ## OLEKSANDR BACHERIKOV ## I s std::mdspan a Zero-overhead Abstraction? Oleksandr Bacherikov Snap Inc ## What is std::mdspan? It's a view Wrong! std::layout_stride supports only all strides specified at runtime. If we target zero overhead, we have to specify one of the strides as 1 at compile time. What does the Standard offer us instead + 4*r8], xmm0 inc r8 cmp rcx, r8 jne .LBB0_4 jmp .LBB0_5 .LBB0_6: ret rsp is the register for the stack pointer mov rax, qword ptr [rsp + 8] test rax, rax je .LBB0_6 lea rsi, [rsp + 8] mov rcx, qword

14a60/p7_2.jpg) ## • KubeBus Protocol Stack • L2/L3/L4 over TCP - KubeBus Client/Server Proxy to proxy Http packets • Multiplex https connections over TCP connection, i.e. "L7 tunnel" ##