istio-tcpip-bypass

# DISCONTINUATION OF PROJECT #  
This project will no longer be maintained by Intel.  
Intel has ceased development and contributions including, but not limited to, maintenance, bug fixes, new releases, or updates, to this project.  
Intel no longer accepts patches to this project.  
 If you have an ongoing need to use this project, are interested in independently developing it, or would like to maintain patches for the open source software community, please create your own fork of this project.  
  
TCP/IP Bypass with eBPF in Istio
================================

This solution aims to bypass TCP/IP stack to accelerate service mesh, it benefits two scenarios:

* p2p communication with sidecar injected
* p2p communication without sidecar injected

This solution is totally independent, which

* does not require changes to linux kernel
* does not require changes to Istio and Envoy (>= v1.10)
* does not require changes to CNI plugin

15%~20% latency decrease is expected for p2p communication on the same host.

System Requirements
~~~~~~~~~~~~~~~~~~~

* Minimal: Distribution with kernel version >= 4.18
* Optimal: Ubuntu 20.04 with Linux 5.4.0-74-generic


Build Docker Image and Load eBPF Program
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

#. Build docker image::

    $ docker build --network=host -t ${IMAGE_NAME} .

#. Load eBPF program via docker command::

    $ docker run --mount type=bind,source=/sys/fs,target=/sys/fs,bind-propagation=rshared --privileged --name tcpip-bypass  ${IMAGE_NAME}

#. Load eBPF program via setting up a deamonset::

    $ kubectl apply -f bypass-tcpip-daemonset.yaml

#. Unload eBPF program via destroying Docker container or deamonset


Debug Log
~~~~~~~~~

#. Enable debug log via modifying the debug MAP::

    $ sudo bpftool map update name debug_map key hex 0 0 0 0  value hex 1 0 0 0

#. Read log from kernel tracepipe::

    $ sudo cat /sys/kernel/debug/tracing/trace_pipe