{{>toc}}

h1. DPDK

Traitement haute performance en logiciel, 10 Gbit/s en petit paquets line rate

h2. Liens

* http://dpdk.org

* [[AtelierPPS2012]]

* GIT http://chiliproject.tetaneutral.net/projects/git-tetaneutral-net/repository/dpdk

<pre>

$ git clone ssh://git@git.tetaneutral.net/dpdk.git

</pre>

* Nerim Antoine https://twitter.com/nerim_net/status/517592380841930752

** Du code maison fondé sur #DPDK en prod dans le réseau : première étape vers l'élimination des systèmes propriétaires ?

* https://fd.io/wp-content/uploads/sites/34/2017/07/FDioVPPwhitepaperJuly2017.pdf

** FD.io - Vector Packet Processing One Terabit Software Router

h2. Misc

* http://lists.ceph.com/pipermail/ceph-users-ceph.com/2017-May/017814.html

** processor C-state tuning for ceph

* https://www.solid-run.com/marvell-armada-family/armada-8040-community-board/

* http://dtrace.org/blogs/rm/2016/09/15/turtles-on-the-wire-understanding-how-the-os-uses-the-modern-nic/

* https://fd.io/

* http://developerblog.redhat.com/2015/06/02/can-you-run-intels-data-plane-development-kit-dpdk-in-a-docker-container-yep/

** DPDK like

* http://www.intel.fr/content/dam/www/public/us/en/documents/datasheets/xl710-10-40-controller-datasheet.pdf

* http://www.intel.fr/content/dam/www/public/us/en/documents/datasheets/82599-10-gbe-controller-datasheet.pdf

* http://www.intel.fr/content/dam/www/public/us/en/documents/datasheets/82598-10-gbe-controller-datasheet.pdf

* MoonGen: A Scriptable High-Speed Packet Generator

** http://arxiv.org/pdf/1410.3322.pdf

** http://www.net.in.tum.de/fileadmin/bibtex/publications/papers/MoonGen_IMC2015.pdf

** https://github.com/emmericp/MoonGen

* https://www.kernel.org/doc/Documentation/networking/switchdev.txt

* https://www.usenix.org/conference/osdi14/technical-sessions/presentation/belay

** IX: A Protected Dataplane Operating System for High Throughput and Low Latency

** http://www.linuxfoundation.org/images/1/1c/Network_data_flow_through_kernel.png

** http://www.linuxfoundation.org/collaborate/workgroups/networking/kernel_flow

* https://blog.cloudflare.com/how-to-receive-a-million-packets/

** How to receive a million packets per second

* https://blog.cloudflare.com/how-to-achieve-low-latency/

* http://www.cs.cmu.edu/~dongz/papers/cuckooswitch.pdf

** We show that CUCKOOSWITCH can process 92.22 million minimum sized packets per second on a commodity server equipped with eight 10 Gbps Ethernet interfaces while maintaining a forwarding table of one billion forwarding entries.

* Improving Linux networking performance https://lwn.net/Articles/629155/

** https://www.youtube.com/watch?v=3XG9-X777Jo

** http://people.netfilter.org/hawk/presentations/LCA2015/net_stack_challenges_100G_LCA2015.pdf

* http://www.bigswitch.com/sites/default/files/presentations/onug-baremetal-2014-final.pdf

* Intel PCI id http://www.intel.com/support/network/sb/cs-012904.htm

* http://www.intel.com/content/www/us/en/ethernet-controllers/ethernet-products.html?iid=embed_portal+hdprod_ethernet#s1=Gigabit%20Ethernet&s2=all&s3=all

* Intel Pro/1000 PT vs CT http://hardforum.com/showthread.php?t=1386042

* http://images.ncix.com/forumimages/C9864B87-583B-10FB-B0DE80742781CE07.jpg

* http://www.intel.com/content/www/us/en/ethernet-controllers/ethernet-controllers.html#s1=all&s2=82572EI&s3=all

* http://www.intel.com/content/www/us/en/ethernet-controllers/ethernet-controllers.html#s1=all&s2=82574L&s3=all

* https://github.com/Pktgen/Pktgen-DPDK

* Evaluating the Suitability of Server Network Cards for Software Routers http://kfall.net/ucbpage/papers/snc.pdf

* http://www.intel.fr/content/dam/www/public/us/en/documents/guides/intel-dpdk-sample-applications-user-guide.pdf

*  rte_eth_rx_burst

** http://dpdk.org/doc/api/rte__ethdev_8h.html#a5ac93faa0f3f69545ca121dcfd9433f5

** http://dpdk.info/ml/archives/dev/2014-January/001223.html

* http://www.intel.fr/content/dam/www/public/us/en/documents/white-papers/aes-ipsec-performance-linux-paper.pdf

* http://networkbuilders.intel.com/docs/Network_Builders_RA_vBRAS_Final.pdf

* https://reannz.co.nz/news/ruru-visualise-international-traffic-flows-real-time/

* https://github.com/SnabbCo/snabbswitch

** https://github.com/snabbco/snabb/

** https://lwn.net/Articles/713918/

** https://github.com/SnabbCo/snabbswitch/blob/master/src/apps/intel/intel10g.lua

** http://blog.ipspace.net/2014/06/snabb-switch-and-nfv-on-openstack-in.html

** https://github.com/SnabbCo/snabbswitch/wiki/Snabb-Lab

** https://twitter.com/lukego/status/327714050219847680

* http://www.lagopus.org/

h2. Note sileht dpdk:

Extract from: http://www.intel.com/content/dam/www/public/us/en/documents/guides/intel-dpdk-getting-started-guide.pdf

h3. configuration hugepages: 

* 2M (1024*2k): hugepages=1024 (at runtime: echo 1024 > /sys/kernel/mm/hugepages/hugepages-2048kB/nr_hugepages)

* 4G (4x1G): default_hugepagesz=1G hugepagesz=1G hugepages=4 (only works at boot time via grub)

<pre>

mkdir /mnt/huge

mount -t hugetlbfs nodev /mnt/huge

</pre>

h3. Compile and load modules: 

_Note: source tools/setup.sh is a helper tools for this but works only with IGB driver not e1000e_

<pre>

# apt-get install make gcc linux-headers-amd64

# make config T=x86_64-default-linuxapp-gcc

# make T=x86_64-default-linuxapp-gcc

..

Build complete

# modprobe uio  (I think this is not useful)

# insmod build/kmod/rte_kni.ko (I think this is not useful)

# insmod build/kmod/igb_uio.ko (I think this is not useful)

# ./tools/pci_unbind.py --status

Network devices using IGB_UIO driver

====================================

<none>

Network devices using kernel driver

===================================

0000:00:19.0 '82579LM Gigabit Network Connection' if=eth1 drv=e1000e unused=<none>

0000:01:00.0 '82599EB 10-Gigabit SFI/SFP+ Network Connection' if=eth6 drv=ixgbe unused=igb_uio *Active*

0000:01:00.1 '82599EB 10-Gigabit SFI/SFP+ Network Connection' if=eth7 drv=ixgbe unused=igb_uio

0000:04:00.0 '82571EB Gigabit Ethernet Controller' if=eth2 drv=e1000e unused=igb_uio *Active*

0000:04:00.1 '82571EB Gigabit Ethernet Controller' if=eth3 drv=e1000e unused=igb_uio

0000:05:00.0 '82571EB Gigabit Ethernet Controller' if=eth4 drv=e1000e unused=igb_uio

0000:05:00.1 '82571EB Gigabit Ethernet Controller' if=eth5 drv=e1000e unused=igb_uio

0000:06:00.0 'RTL8111/8168B PCI Express Gigabit Ethernet controller' if=eth0 drv=r8169 unused=<none> *Active*

Other network devices

=====================

<none>

</pre>

test sur eth7:

<pre>

# ip link set eth7 down

# ./tools/pci_unbind.py --bind=igb_uio 0000:01:00.1

</pre>

h3. Prepare examples programs:

<pre>

# export RTE_SDK=/root/sileht/dpdk-1.6.0r1

# export RTE_TARGET=build

# cd /root/sileht/

# cp -r $RTE_SDK/examples/helloworld my_rte_app

# cd my_rte_app

# make

</pre>

h3. Tests

<pre>

</pre>

h2. Limitation SFP carte intel:

Une fois le module igb_uio changé sur les:

<pre>

01:00.0 Ethernet controller: Intel Corporation 82599ES 10-Gigabit SFI/SFP+ Network Connection (rev 01)

</pre>

Un flags de l'eeprom change et le module ixgbe ne charge plus les ports de la carte qui n'ont pas des SFP+ intel avec le message d'erreur suivant:

<pre>

[  563.069683] ixgbe 0000:01:00.0: failed to load because an unsupported SFP+ or QSFP module type was detected.

[  563.069786] ixgbe 0000:01:00.0: Reload the driver after installing a supported module.

</pre>

Pour contourner le problème:

<pre>

modprobe ixgbe allow_unsupported_sfp=1 # allow_any_sfp=1 suivant le kernel

</pre>

ou 

<pre>

echo "options ixgbe allow_unsupported_sfp=1" > /etc/modprobe.d/ixgbe_any_sfp.conf

update-initramfs -u -k all

</pre>

La carte fonctionnera à nouveau avec le message suivant:

<pre>

[  908.361767] ixgbe 0000:01:00.0 (unregistered net_device): WARNING: Intel (R) Network Connections are quality tested using Intel (R) Ethernet Optics.  Using untested modules is not supported and may cause unstable operation or damage to the module or the adapter.  Intel Corporation is not responsible for any harm caused by using untested modules.

</pre>

h2. Packet Journey

https://www.gandi.net/news/en/2015-12-16/6308-packet-journey_a_free_software_router_for_linux_based_on_dpdk/

https://github.com/gandi/packet-journey

h2. Atelier 2016

h3. Atelier Hardware

En salle milieu a Myrys.

http://pad.tetaneutral.net/p/dpdk

<pre>

core i7 2600 @ 3.40GHz

 fpu vme de pse tsc msr pae mce cx8 

 apic sep mtrr pge mca cmov pat pse36 clflush dts acpi mmx fxsr sse sse2 

 ss ht tm pbe syscall nx rdtscp lm constant_tsc arch_perfmon pebs bts rep_good nopl 

 xtopology nonstop_tsc aperfmperf eagerfpu pni pclmulqdq dtes64 monitor ds_cpl vmx smx est tm2 ssse3 

 cx16 xtpr pdcm pcid sse4_1 sse4_2 

 x2apic popcnt tsc_deadline_timer aes xsave avx lahf_lm ida arat epb xsaveopt pln pts dtherm tpr_shadow vnmi flexpriority ept vpid

16G RAM

SSD 250G

</pre>

http://www.intel.fr/content/dam/www/public/us/en/documents/guides/gbe-server-selection-guide.pdf

<pre>

h7   82599ES 8086:10fb 8086:7a11 

g1   82599ES (VM maposmatic)

g2   X520    8086:154d 8086:7b11

g3   82599ES

g4   X520 

g5   X520

g6   X520

n7   X520 (VM openstack)

g8   X520

g9   82598EB 8086:10f1 8086:a21f

g10  X520

g11  82598EB

g12  82598EB

stri X520

t1   82598EB 

t2   82598EB 

t3   82598EB 

stock 82598EB

stock 82598EB

stock X520

</pre>

old obsolete:

<pre>

host eth DHCP          10G-a                                           10G-b

t1   54:04:a6:ef:a8:c7 00:1b:21:d9:28:55 fe80::21b:21ff:fed9:2855(t2b) 00:1b:21:d9:28:54 fe80::21b:21ff:fed9:2854(t3b)

t2   e0:69:95:2e:e6:a2 00:1b:21:d9:1a:e1 fe80::21b:21ff:fed9:1ae1(t3a) 00:1b:21:d9:1a:e0 fe80::21b:21ff:fed9:1ae0(t1a)

t3   bc:ae:c5:e1:b2:6d 00:1b:21:d9:df:3d fe80::21b:21ff:fed9:df3d(t2a) 00:1b:21:d9:df:3c fe80::21b:21ff:fed9:df3c(t1b)

1G e1000e

t1 00:15:17:a6:cb:19

t3 00:15:17:16:bd:de

</pre>

h3. Atelier Liens

https://github.com/Gandi/packet-journey

http://dpdk.org/doc/guides/linux_gsg/build_dpdk.html

http://dpdk.org/doc/quick-start

http://dpdk.org/browse/apps/pktgen-dpdk/tree/README.md

http://dpdk.org/doc/guides/testpmd_app_ug/index.html

http://dpdk.readthedocs.org/en/v2.2.0/

https://github.com/emmericp/MoonGen

http://www.slideshare.net/movilforum/7-hands-on

http://p4.org/wp-content/uploads/2015/11/Princeton-Demo-Poster.pdf

http://dpdk.org/ml/archives/users/

http://dpdk.org/ml/archives/dev/

http://openvswitch.org/support/ovscon2015/

https://events.ccc.de/congress/2015/Fahrplan/events/7523.html lecture: DDoS mitigation EPIC FAIL collection

https://github.com/aragua/libpacket

https://github.com/groud/dpdkcap

h3. Atelier Issues

* Issues opened

** https://github.com/Gandi/packet-journey/issues/20 

** https://github.com/pktgen/Pktgen-DPDK/issues/52

** https://github.com/emmericp/MoonGen/issues/106

** http://dpdk.org/ml/archives/dev/2015-December/030798.html

** http://dpdk.org/ml/archives/dev/2016-April/037331.html

** https://github.com/Gandi/packet-journey/pull/36

** https://github.com/Gandi/packet-journey/pull/38

h3. Atelier Build

http://pad.tetaneutral.net/p/dpdk

<pre>

# From debian jessie basic install

# Packages

apt-get -y build-dep linux-image-amd64

apt-get -y install linux-headers-$(uname -r)

apt-get -y install git linux-tools emacs-nox libpcap-dev tcpdump iperf iperf3 screen rsync socat bzip2 xz-utils cmake python-sphinx tcpdump

# huge pages, dans /etc/rc.local par exemple

modprobe uio_pci_generic

mkdir -p /mnt/huge

mount -t hugetlbfs nodev /mnt/huge

# or GRUB_CMDLINE_LINUX="default_hugepagesz=1G hugepagesz=1G hugepages=4" :

echo 1024 > /sys/devices/system/node/node0/hugepages/hugepages-2048kB/nr_hugepages

# Sources

mkdir ftp

cd ftp

wget http://dpdk.org/browse/dpdk/snapshot/dpdk-2.2.0.tar.gz

wget http://dpdk.org/browse/apps/pktgen-dpdk/snapshot/pktgen-2.9.12.tar.gz

cd ..

git clone https://github.com/Gandi/packet-journey.git

git clone http://dpdk.org/git/apps/pktgen-dpdk

git clone http://dpdk.org/git/dpdk-stable

git clone https://github.com/emmericp/MoonGen.git

# Build DPDK

tar xf ftp/dpdk-2.2.0.tar.gz

cd dpdk-2.2.0

make config T=x86_64-native-linuxapp-gcc

make install T=x86_64-native-linuxapp-gcc  # about 2mn40s on core i7 2.6 GHz, 4m35s on Celeron 3205U 1.5 GHz

export RTE_SDK=/root/dpdk-2.2.0

export RTE_TARGET=x86_64-native-linuxapp-gcc

/root/dpdk-2.2.0/tools/dpdk_nic_bind.py -s

/root/dpdk-2.2.0/tools/dpdk_nic_bind.py --bind=uio_pci_generic 0000:01:00.0

/root/dpdk-2.2.0/tools/dpdk_nic_bind.py --bind=uio_pci_generic 0000:01:00.1

/root/dpdk-2.2.0/x86_64-native-linuxapp-gcc/app/testpmd -c 7 -n 2 -- -i

# http://www.intel.fr/content/dam/www/public/us/en/documents/guides/dpdk-testpmd-application-user-guide.pdf

# Build pktgen (2.9.13+ doesn't compile)

cd ..

tar xf ftp/pktgen-2.9.12.tar.gz

cd pktgen-2.9.12

make # about 24s

./app/app/x86_64-native-linuxapp-gcc/pktgen -c 7 -n 2   --proc-type auto -- -P  -N -m '1.0,2.1'

# save toto, edit toto and change both dst.mac start N xxxx and add start 0 and start 1 at the end of the file

./app/app/x86_64-native-linuxapp-gcc/pktgen -c 7 -n 2   --proc-type auto -- -P  -N -m '1.0,2.1' -f toto

# Build MoonGen

cd ..

cd MoonGen

git submodule update --init

./build.sh # about 42 seconds

./build/MoonGen ./examples/benchmark/udp-throughput.lua 1:1

# TODO

cd ../packet-journey

make

cd ../pktgen-dpdk

make

cd docs

make html

</pre>

Patch pour packet journey sur dpdk 16.11: https://github.com/Gandi/packet-journey/pull/36