1 OVS DPDK ADVANCED INSTALL GUIDE
2 =================================
6 1. [Overview](#overview)
7 2. [Building Shared Library](#build)
8 3. [System configuration](#sysconf)
9 4. [Performance Tuning](#perftune)
10 5. [OVS Testcases](#ovstc)
11 6. [Vhost Walkthrough](#vhost)
13 8. [Rate Limiting](#rl)
16 ## <a name="overview"></a> 1. Overview
18 The Advanced Install Guide explains how to improve OVS performance using
19 DPDK datapath. This guide also provides information on tuning, system configuration,
20 troubleshooting, static code analysis and testcases.
22 ## <a name="build"></a> 2. Building Shared Library
24 DPDK can be built as static or shared library and shall be linked by applications
25 using DPDK datapath. The section lists steps to build shared library and dynamically
26 link DPDK against OVS.
28 Note: Minor performance loss is seen with OVS when using shared DPDK library as
29 compared to static library.
31 Check section [INSTALL DPDK], [INSTALL OVS] of INSTALL.DPDK on download instructions
34 * Configure the DPDK library
36 Set `CONFIG_RTE_BUILD_SHARED_LIB=y` in `config/common_base`
37 to generate shared DPDK library
40 * Build and install DPDK
42 For Default install (without IVSHMEM), set `export DPDK_TARGET=x86_64-native-linuxapp-gcc`
43 For IVSHMEM case, set `export DPDK_TARGET=x86_64-ivshmem-linuxapp-gcc`
46 export DPDK_DIR=/usr/src/dpdk-16.04
47 export DPDK_BUILD=$DPDK_DIR/$DPDK_TARGET
48 make install T=$DPDK_TARGET DESTDIR=install
51 * Build, Install and Setup OVS.
53 Export the DPDK shared library location and setup OVS as listed in
54 section 3.3 of INSTALL.DPDK.
56 `export LD_LIBRARY_PATH=$DPDK_DIR/x86_64-native-linuxapp-gcc/lib`
58 ## <a name="sysconf"></a> 3. System Configuration
60 To achieve optimal OVS performance, the system can be configured and that includes
61 BIOS tweaks, Grub cmdline additions, better understanding of NUMA nodes and
62 apt selection of PCIe slots for NIC placement.
64 ### 3.1 Recommended BIOS settings
67 | Settings | values | comments
68 |---------------------------|-----------|-----------
69 | C3 power state | Disabled | -
70 | C6 power state | Disabled | -
71 | MLC Streamer | Enabled | -
72 | MLC Spacial prefetcher | Enabled | -
73 | DCU Data prefetcher | Enabled | -
75 | CPU power and performance | Performance -
76 | Memory RAS and perf | | -
77 config-> NUMA optimized | Enabled | -
80 ### 3.2 PCIe Slot Selection
82 The fastpath performance also depends on factors like the NIC placement,
83 Channel speeds between PCIe slot and CPU, proximity of PCIe slot to the CPU
84 cores running DPDK application. Listed below are the steps to identify
87 - Retrieve host details using cmd `dmidecode -t baseboard | grep "Product Name"`
88 - Download the technical specification for Product listed eg: S2600WT2.
89 - Check the Product Architecture Overview on the Riser slot placement,
90 CPU sharing info and also PCIe channel speeds.
92 example: On S2600WT, CPU1 and CPU2 share Riser Slot 1 with Channel speed between
93 CPU1 and Riser Slot1 at 32GB/s, CPU2 and Riser Slot1 at 16GB/s. Running DPDK app
94 on CPU1 cores and NIC inserted in to Riser card Slots will optimize OVS performance
97 - Check the Riser Card #1 - Root Port mapping information, on the available slots
98 and individual bus speeds. In S2600WT slot 1, slot 2 has high bus speeds and are
99 potential slots for NIC placement.
101 ### 3.3 Advanced Hugepage setup
103 Allocate and mount 1G Huge pages:
105 - For persistent allocation of huge pages, add the following options to the kernel bootline
107 Add `default_hugepagesz=1GB hugepagesz=1G hugepages=N`
109 For platforms supporting multiple huge page sizes, Add options
111 `default_hugepagesz=<size> hugepagesz=<size> hugepages=N`
112 where 'N' = Number of huge pages requested, 'size' = huge page size,
113 optional suffix [kKmMgG]
115 - For run-time allocation of huge pages
117 `echo N > /sys/devices/system/node/nodeX/hugepages/hugepages-1048576kB/nr_hugepages`
118 where 'N' = Number of huge pages requested, 'X' = NUMA Node
120 Note: For run-time allocation of 1G huge pages, Contiguous Memory Allocator(CONFIG_CMA)
121 has to be supported by kernel, check your Linux distro.
125 `mount -t hugetlbfs -o pagesize=1G none /dev/hugepages`
127 Note: Mount hugepages if not already mounted by default.
129 ### 3.4 Enable Hyperthreading
131 Requires BIOS changes
133 With HT/SMT enabled, A Physical core appears as two logical cores.
134 SMT can be utilized to spawn worker threads on logical cores of the same
135 physical core there by saving additional cores.
137 With DPDK, When pinning pmd threads to logical cores, care must be taken
138 to set the correct bits in the pmd-cpu-mask to ensure that the pmd threads are
139 pinned to SMT siblings.
141 Example System configuration:
142 Dual socket Machine, 2x 10 core processors, HT enabled, 40 logical cores
144 To use two logical cores which share the same physical core for pmd threads,
145 the following command can be used to identify a pair of logical cores.
147 `cat /sys/devices/system/cpu/cpuN/topology/thread_siblings_list`, where N is the
150 In this example, it would show that cores 1 and 21 share the same physical core.
151 The pmd-cpu-mask to enable two pmd threads running on these two logical cores
152 (one physical core) is.
154 `ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=100002`
156 ### 3.5 Isolate cores
158 'isolcpus' option can be used to isolate cores from the linux scheduler.
159 The isolated cores can then be used to dedicatedly run HPC applications/threads.
160 This helps in better application performance due to zero context switching and
161 minimal cache thrashing. To run platform logic on core 0 and isolate cores
162 between 1 and 19 from scheduler, Add `isolcpus=1-19` to GRUB cmdline.
164 Note: It has been verified that core isolation has minimal advantage due to
165 mature Linux scheduler in some circumstances.
167 ### 3.6 NUMA/Cluster on Die
169 Ideally inter NUMA datapaths should be avoided where possible as packets
170 will go across QPI and there may be a slight performance penalty when
171 compared with intra NUMA datapaths. On Intel Xeon Processor E5 v3,
172 Cluster On Die is introduced on models that have 10 cores or more.
173 This makes it possible to logically split a socket into two NUMA regions
174 and again it is preferred where possible to keep critical datapaths
175 within the one cluster.
177 It is good practice to ensure that threads that are in the datapath are
178 pinned to cores in the same NUMA area. e.g. pmd threads and QEMU vCPUs
179 responsible for forwarding. If DPDK is built with
180 CONFIG_RTE_LIBRTE_VHOST_NUMA=y, vHost User ports automatically
181 detect the NUMA socket of the QEMU vCPUs and will be serviced by a PMD
182 from the same node provided a core on this node is enabled in the
185 ### 3.7 Compiler Optimizations
187 The default compiler optimization level is '-O2'. Changing this to
188 more aggressive compiler optimization such as '-O3 -march=native'
189 with gcc(verified on 5.3.1) can produce performance gains though not
190 siginificant. '-march=native' will produce optimized code on local machine
191 and should be used when SW compilation is done on Testbed.
193 ## <a name="perftune"></a> 4. Performance Tuning
197 For superior performance, DPDK pmd threads and Qemu vCPU threads
198 needs to be affinitized accordingly.
200 * PMD thread Affinity
202 A poll mode driver (pmd) thread handles the I/O of all DPDK
203 interfaces assigned to it. A pmd thread shall poll the ports
204 for incoming packets, switch the packets and send to tx port.
205 pmd thread is CPU bound, and needs to be affinitized to isolated
206 cores for optimum performance.
208 By setting a bit in the mask, a pmd thread is created and pinned
209 to the corresponding CPU core. e.g. to run a pmd thread on core 2
211 `ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=4`
213 Note: pmd thread on a NUMA node is only created if there is
214 at least one DPDK interface from that NUMA node added to OVS.
216 * Qemu vCPU thread Affinity
218 A VM performing simple packet forwarding or running complex packet
219 pipelines has to ensure that the vCPU threads performing the work has
220 as much CPU occupancy as possible.
222 Example: On a multicore VM, multiple QEMU vCPU threads shall be spawned.
223 when the DPDK 'testpmd' application that does packet forwarding
224 is invoked, 'taskset' cmd should be used to affinitize the vCPU threads
225 to the dedicated isolated cores on the host system.
227 ### 4.2 Multiple poll mode driver threads
229 With pmd multi-threading support, OVS creates one pmd thread
230 for each NUMA node by default. However, it can be seen that in cases
231 where there are multiple ports/rxq's producing traffic, performance
232 can be improved by creating multiple pmd threads running on separate
233 cores. These pmd threads can then share the workload by each being
234 responsible for different ports/rxq's. Assignment of ports/rxq's to
235 pmd threads is done automatically.
237 A set bit in the mask means a pmd thread is created and pinned
238 to the corresponding CPU core. e.g. to run pmd threads on core 1 and 2
240 `ovs-vsctl set Open_vSwitch . other_config:pmd-cpu-mask=6`
242 For example, when using dpdk and dpdkvhostuser ports in a bi-directional
243 VM loopback as shown below, spreading the workload over 2 or 4 pmd
244 threads shows significant improvements as there will be more total CPU
247 NIC port0 <-> OVS <-> VM <-> OVS <-> NIC port 1
249 ### 4.3 DPDK physical port Rx Queues
251 `ovs-vsctl set Interface <DPDK interface> options:n_rxq=<integer>`
253 The command above sets the number of rx queues for DPDK physical interface.
254 The rx queues are assigned to pmd threads on the same NUMA node in a
257 ### 4.4 Exact Match Cache
259 Each pmd thread contains one EMC. After initial flow setup in the
260 datapath, the EMC contains a single table and provides the lowest level
261 (fastest) switching for DPDK ports. If there is a miss in the EMC then
262 the next level where switching will occur is the datapath classifier.
263 Missing in the EMC and looking up in the datapath classifier incurs a
264 significant performance penalty. If lookup misses occur in the EMC
265 because it is too small to handle the number of flows, its size can
266 be increased. The EMC size can be modified by editing the define
267 EM_FLOW_HASH_SHIFT in lib/dpif-netdev.c.
269 As mentioned above an EMC is per pmd thread. So an alternative way of
270 increasing the aggregate amount of possible flow entries in EMC and
271 avoiding datapath classifier lookups is to have multiple pmd threads
272 running. This can be done as described in section 4.2.
274 ### 4.5 Rx Mergeable buffers
276 Rx Mergeable buffers is a virtio feature that allows chaining of multiple
277 virtio descriptors to handle large packet sizes. As such, large packets
278 are handled by reserving and chaining multiple free descriptors
279 together. Mergeable buffer support is negotiated between the virtio
280 driver and virtio device and is supported by the DPDK vhost library.
281 This behavior is typically supported and enabled by default, however
282 in the case where the user knows that rx mergeable buffers are not needed
283 i.e. jumbo frames are not needed, it can be forced off by adding
284 mrg_rxbuf=off to the QEMU command line options. By not reserving multiple
285 chains of descriptors it will make more individual virtio descriptors
286 available for rx to the guest using dpdkvhost ports and this can improve
289 ## <a name="ovstc"></a> 5. OVS Testcases
290 ### 5.1 PHY-VM-PHY [VHOST LOOPBACK]
292 The section 5.2 in INSTALL.DPDK guide lists steps for PVP loopback testcase
293 and packet forwarding using DPDK testpmd application in the Guest VM.
294 For users wanting to do packet forwarding using kernel stack below are the steps.
297 ifconfig eth1 1.1.1.2/24
298 ifconfig eth2 1.1.2.2/24
299 systemctl stop firewalld.service
300 systemctl stop iptables.service
301 sysctl -w net.ipv4.ip_forward=1
302 sysctl -w net.ipv4.conf.all.rp_filter=0
303 sysctl -w net.ipv4.conf.eth1.rp_filter=0
304 sysctl -w net.ipv4.conf.eth2.rp_filter=0
305 route add -net 1.1.2.0/24 eth2
306 route add -net 1.1.1.0/24 eth1
307 arp -s 1.1.2.99 DE:AD:BE:EF:CA:FE
308 arp -s 1.1.1.99 DE:AD:BE:EF:CA:EE
311 ### 5.2 PHY-VM-PHY [IVSHMEM]
313 The steps (1-5) in 3.3 section of INSTALL.DPDK guide will create & initialize DB,
314 start vswitchd and add dpdk devices to bridge br0.
316 1. Add DPDK ring port to the bridge
319 ovs-vsctl add-port br0 dpdkr0 -- set Interface dpdkr0 type=dpdkr
322 2. Build modified Qemu (Qemu-2.2.1 + ivshmem-qemu-2.2.1.patch)
326 wget http://wiki.qemu.org/download/qemu-2.2.1.tar.bz2
327 tar -jxvf qemu-2.2.1.tar.bz2
328 cd /usr/src/qemu-2.2.1
329 wget https://raw.githubusercontent.com/netgroup-polito/un-orchestrator/master/orchestrator/compute_controller/plugins/kvm-libvirt/patches/ivshmem-qemu-2.2.1.patch
330 patch -p1 < ivshmem-qemu-2.2.1.patch
331 ./configure --target-list=x86_64-softmmu --enable-debug --extra-cflags='-g'
335 3. Generate Qemu commandline
338 mkdir -p /usr/src/cmdline_generator
339 cd /usr/src/cmdline_generator
340 wget https://raw.githubusercontent.com/netgroup-polito/un-orchestrator/master/orchestrator/compute_controller/plugins/kvm-libvirt/cmdline_generator/cmdline_generator.c
341 wget https://raw.githubusercontent.com/netgroup-polito/un-orchestrator/master/orchestrator/compute_controller/plugins/kvm-libvirt/cmdline_generator/Makefile
342 export RTE_SDK=/usr/src/dpdk-16.04
343 export RTE_TARGET=x86_64-ivshmem-linuxapp-gcc
345 ./build/cmdline_generator -m -p dpdkr0 XXX
346 cmdline=`cat OVSMEMPOOL`
352 export VM_NAME=ivshmem-vm
353 export QCOW2_IMAGE=/root/CentOS7_x86_64.qcow2
354 export QEMU_BIN=/usr/src/qemu-2.2.1/x86_64-softmmu/qemu-system-x86_64
356 taskset 0x20 $QEMU_BIN -cpu host -smp 2,cores=2 -hda $QCOW2_IMAGE -m 4096 --enable-kvm -name $VM_NAME -nographic -vnc :2 -pidfile /tmp/vm1.pid $cmdline
359 5. Running sample "dpdk ring" app in VM
362 echo 1024 > /proc/sys/vm/nr_hugepages
363 mount -t hugetlbfs nodev /dev/hugepages (if not already mounted)
365 # Build the DPDK ring application in the VM
366 export RTE_SDK=/root/dpdk-16.04
367 export RTE_TARGET=x86_64-ivshmem-linuxapp-gcc
370 # Run dpdkring application
371 ./build/dpdkr -c 1 -n 4 -- -n 0
372 where "-n 0" refers to ring '0' i.e dpdkr0
375 ## <a name="vhost"></a> 6. Vhost Walkthrough
377 DPDK 16.04 supports two types of vhost:
379 1. vhost-user - enabled default
381 2. vhost-cuse - Legacy, disabled by default
389 - Adding vhost-user ports to Switch
391 Unlike DPDK ring ports, DPDK vhost-user ports can have arbitrary names,
392 except that forward and backward slashes are prohibited in the names.
394 For vhost-user, the name of the port type is `dpdkvhostuser`
397 ovs-vsctl add-port br0 vhost-user-1 -- set Interface vhost-user-1
401 This action creates a socket located at
402 `/usr/local/var/run/openvswitch/vhost-user-1`, which you must provide
403 to your VM on the QEMU command line. More instructions on this can be
404 found in the next section "Adding vhost-user ports to VM"
406 Note: If you wish for the vhost-user sockets to be created in a
407 sub-directory of `/usr/local/var/run/openvswitch`, you may specify
408 this directory in the ovsdb like so:
410 `./utilities/ovs-vsctl --no-wait \
411 set Open_vSwitch . other_config:vhost-sock-dir=subdir`
413 - Adding vhost-user ports to VM
417 Pass the following parameters to QEMU to attach a vhost-user device:
420 -chardev socket,id=char1,path=/usr/local/var/run/openvswitch/vhost-user-1
421 -netdev type=vhost-user,id=mynet1,chardev=char1,vhostforce
422 -device virtio-net-pci,mac=00:00:00:00:00:01,netdev=mynet1
425 where vhost-user-1 is the name of the vhost-user port added
427 Repeat the above parameters for multiple devices, changing the
428 chardev path and id as necessary. Note that a separate and different
429 chardev path needs to be specified for each vhost-user device. For
430 example you have a second vhost-user port named 'vhost-user-2', you
431 append your QEMU command line with an additional set of parameters:
434 -chardev socket,id=char2,path=/usr/local/var/run/openvswitch/vhost-user-2
435 -netdev type=vhost-user,id=mynet2,chardev=char2,vhostforce
436 -device virtio-net-pci,mac=00:00:00:00:00:02,netdev=mynet2
439 2. Configure huge pages.
441 QEMU must allocate the VM's memory on hugetlbfs. vhost-user ports access
442 a virtio-net device's virtual rings and packet buffers mapping the VM's
443 physical memory on hugetlbfs. To enable vhost-user ports to map the VM's
444 memory into their process address space, pass the following parameters
448 -object memory-backend-file,id=mem,size=4096M,mem-path=/dev/hugepages,
449 share=on -numa node,memdev=mem -mem-prealloc
452 3. Enable multiqueue support(OPTIONAL)
454 QEMU needs to be configured to use multiqueue.
455 The $q below is the number of queues.
456 The $v is the number of vectors, which is '$q x 2 + 2'.
459 -chardev socket,id=char2,path=/usr/local/var/run/openvswitch/vhost-user-2
460 -netdev type=vhost-user,id=mynet2,chardev=char2,vhostforce,queues=$q
461 -device virtio-net-pci,mac=00:00:00:00:00:02,netdev=mynet2,mq=on,vectors=$v
464 The vhost-user interface will be automatically reconfigured with required
465 number of rx and tx queues after connection of virtio device.
466 Manual configuration of `n_rxq` is not supported because OVS will work
467 properly only if `n_rxq` will match number of queues configured in QEMU.
469 A least 2 PMDs should be configured for the vswitch when using multiqueue.
470 Using a single PMD will cause traffic to be enqueued to the same vhost
471 queue rather than being distributed among different vhost queues for a
472 vhost-user interface.
474 If traffic destined for a VM configured with multiqueue arrives to the
475 vswitch via a physical DPDK port, then the number of rxqs should also be
476 set to at least 2 for that physical DPDK port. This is required to increase
477 the probability that a different PMD will handle the multiqueue
478 transmission to the guest using a different vhost queue.
480 If one wishes to use multiple queues for an interface in the guest, the
481 driver in the guest operating system must be configured to do so. It is
482 recommended that the number of queues configured be equal to '$q'.
484 For example, this can be done for the Linux kernel virtio-net driver with:
487 ethtool -L <DEV> combined <$q>
489 where `-L`: Changes the numbers of channels of the specified network device
490 and `combined`: Changes the number of multi-purpose channels.
492 - VM Configuration with libvirt
494 * change the user/group, access control policty and restart libvirtd.
496 - In `/etc/libvirt/qemu.conf` add/edit the following lines
503 - Disable SELinux or set to permissive mode
507 - Restart the libvirtd process, For example, on Fedora
509 `systemctl restart libvirtd.service`
513 - Copy the xml configuration from [Guest VM using libvirt] in to workspace.
517 `virsh create demovm.xml`
519 - Connect to the guest console
521 `virsh console demovm`
525 The demovm xml configuration is aimed at achieving out of box performance
528 - The vcpus are pinned to the cores of the CPU socket 0 using vcpupin.
530 - Configure NUMA cell and memory shared using memAccess='shared'.
532 - Disable mrg_rxbuf='off'.
534 Note: For information on libvirt and further tuning refer [libvirt].
542 - Enable vhost-cuse support
544 1. Enable vhost cuse support in DPDK
546 Set `CONFIG_RTE_LIBRTE_VHOST_USER=n` in config/common_linuxapp and follow the
547 steps in 2.2 section of INSTALL.DPDK guide to build DPDK with cuse support.
548 OVS will detect that DPDK has vhost-cuse libraries compiled and in turn will enable
549 support for it in the switch and disable vhost-user support.
551 2. Insert the Cuse module
555 3. Build and insert the `eventfd_link` module
558 cd $DPDK_DIR/lib/librte_vhost/eventfd_link/
560 insmod $DPDK_DIR/lib/librte_vhost/eventfd_link.ko
563 - Adding vhost-cuse ports to Switch
565 Unlike DPDK ring ports, DPDK vhost-cuse ports can have arbitrary names.
566 For vhost-cuse, the name of the port type is `dpdkvhostcuse`
569 ovs-vsctl add-port br0 vhost-cuse-1 -- set Interface vhost-cuse-1
573 When attaching vhost-cuse ports to QEMU, the name provided during the
574 add-port operation must match the ifname parameter on the QEMU cmd line.
576 - Adding vhost-cuse ports to VM
578 vhost-cuse ports use a Linux* character device to communicate with QEMU.
579 By default it is set to `/dev/vhost-net`. It is possible to reuse this
580 standard device for DPDK vhost, which makes setup a little simpler but it
581 is better practice to specify an alternative character device in order to
582 avoid any conflicts if kernel vhost is to be used in parallel.
584 1. This step is only needed if using an alternative character device.
587 ./utilities/ovs-vsctl --no-wait set Open_vSwitch . \
588 other_config:cuse-dev-name=my-vhost-net
591 In the example above, the character device to be used will be
594 2. In case of reusing kernel vhost character device, there would be conflict
595 user should remove it.
597 `rm -rf /dev/vhost-net`
599 3. Configure virtio-net adapters
601 The following parameters must be passed to the QEMU binary, repeat
602 the below parameters for multiple devices.
605 -netdev tap,id=<id>,script=no,downscript=no,ifname=<name>,vhost=on
606 -device virtio-net-pci,netdev=net1,mac=<mac>
609 The DPDK vhost library will negotiate its own features, so they
610 need not be passed in as command line params. Note that as offloads
611 are disabled this is the equivalent of setting
613 `csum=off,gso=off,guest_tso4=off,guest_tso6=off,guest_ecn=off`
615 When using an alternative character device, it must be explicitly
616 passed to QEMU using the `vhostfd` argument
619 -netdev tap,id=<id>,script=no,downscript=no,ifname=<name>,vhost=on,
620 vhostfd=<open_fd> -device virtio-net-pci,netdev=net1,mac=<mac>
623 The open file descriptor must be passed to QEMU running as a child
624 process. This could be done with a simple python script.
628 fd = os.open("/dev/usvhost", os.O_RDWR)
629 subprocess.call("qemu-system-x86_64 .... -netdev tap,id=vhostnet0,\
630 vhost=on,vhostfd=" + fd +"...", shell=True)
633 4. Configure huge pages
635 QEMU must allocate the VM's memory on hugetlbfs. Vhost ports access a
636 virtio-net device's virtual rings and packet buffers mapping the VM's
637 physical memory on hugetlbfs. To enable vhost-ports to map the VM's
638 memory into their process address space, pass the following parameters
641 `-object memory-backend-file,id=mem,size=4096M,mem-path=/dev/hugepages,
642 share=on -numa node,memdev=mem -mem-prealloc`
644 - VM Configuration with QEMU wrapper
646 The QEMU wrapper script automatically detects and calls QEMU with the
647 necessary parameters. It performs the following actions:
649 * Automatically detects the location of the hugetlbfs and inserts this
650 into the command line parameters.
651 * Automatically open file descriptors for each virtio-net device and
652 inserts this into the command line parameters.
653 * Calls QEMU passing both the command line parameters passed to the
654 script itself and those it has auto-detected.
656 Before use, you **must** edit the configuration parameters section of the
657 script to point to the correct emulator location and set additional
658 settings. Of these settings, `emul_path` and `us_vhost_path` **must** be
659 set. All other settings are optional.
661 To use directly from the command line simply pass the wrapper some of the
662 QEMU parameters: it will configure the rest. For example:
665 qemu-wrap.py -cpu host -boot c -hda <disk image> -m 4096 -smp 4
666 --enable-kvm -nographic -vnc none -net none -netdev tap,id=net1,
667 script=no,downscript=no,ifname=if1,vhost=on -device virtio-net-pci,
668 netdev=net1,mac=00:00:00:00:00:01
671 - VM Configuration with libvirt
673 If you are using libvirt, you must enable libvirt to access the character
674 device by adding it to controllers cgroup for libvirtd using the following
677 1. In `/etc/libvirt/qemu.conf` add/edit the following lines:
680 clear_emulator_capabilities = 0
683 cgroup_device_acl = [
684 "/dev/null", "/dev/full", "/dev/zero",
685 "/dev/random", "/dev/urandom",
686 "/dev/ptmx", "/dev/kvm", "/dev/kqemu",
687 "/dev/rtc", "/dev/hpet", "/dev/net/tun",
688 "/dev/<my-vhost-device>",
692 <my-vhost-device> refers to "vhost-net" if using the `/dev/vhost-net`
693 device. If you have specificed a different name in the database
694 using the "other_config:cuse-dev-name" parameter, please specify that
697 2. Disable SELinux or set to permissive mode
699 3. Restart the libvirtd process
700 For example, on Fedora:
702 `systemctl restart libvirtd.service`
704 After successfully editing the configuration, you may launch your
705 vhost-enabled VM. The XML describing the VM can be configured like so
706 within the <qemu:commandline> section:
708 1. Set up shared hugepages:
711 <qemu:arg value='-object'/>
712 <qemu:arg value='memory-backend-file,id=mem,size=4096M,mem-path=/dev/hugepages,share=on'/>
713 <qemu:arg value='-numa'/>
714 <qemu:arg value='node,memdev=mem'/>
715 <qemu:arg value='-mem-prealloc'/>
718 2. Set up your tap devices:
721 <qemu:arg value='-netdev'/>
722 <qemu:arg value='type=tap,id=net1,script=no,downscript=no,ifname=vhost0,vhost=on'/>
723 <qemu:arg value='-device'/>
724 <qemu:arg value='virtio-net-pci,netdev=net1,mac=00:00:00:00:00:01'/>
727 Repeat for as many devices as are desired, modifying the id, ifname
728 and mac as necessary.
730 Again, if you are using an alternative character device (other than
731 `/dev/vhost-net`), please specify the file descriptor like so:
733 `<qemu:arg value='type=tap,id=net3,script=no,downscript=no,ifname=vhost0,vhost=on,vhostfd=<open_fd>'/>`
735 Where <open_fd> refers to the open file descriptor of the character device.
736 Instructions of how to retrieve the file descriptor can be found in the
737 "DPDK vhost VM configuration" section.
738 Alternatively, the process is automated with the qemu-wrap.py script,
739 detailed in the next section.
741 Now you may launch your VM using virt-manager, or like so:
743 `virsh create my_vhost_vm.xml`
745 - VM Configuration with libvirt & QEMU wrapper
747 To use the qemu-wrapper script in conjuntion with libvirt, follow the
748 steps in the previous section before proceeding with the following steps:
750 1. Place `qemu-wrap.py` in libvirtd binary search PATH ($PATH)
751 Ideally in the same directory that the QEMU binary is located.
753 2. Ensure that the script has the same owner/group and file permissions
756 3. Update the VM xml file using "virsh edit VM.xml"
758 Set the VM to use the launch script.
759 Set the emulator path contained in the `<emulator><emulator/>` tags.
760 For example, replace `<emulator>/usr/bin/qemu-kvm<emulator/>` with
761 `<emulator>/usr/bin/qemu-wrap.py<emulator/>`
763 4. Edit the Configuration Parameters section of the script to point to
764 the correct emulator location and set any additional options. If you are
765 using a alternative character device name, please set "us_vhost_path" to the
766 location of that device. The script will automatically detect and insert
767 the correct "vhostfd" value in the QEMU command line arguments.
769 5. Use virt-manager to launch the VM
771 ### 6.3 DPDK backend inside VM
773 Please note that additional configuration is required if you want to run
774 ovs-vswitchd with DPDK backend inside a QEMU virtual machine. Ovs-vswitchd
775 creates separate DPDK TX queues for each CPU core available. This operation
776 fails inside QEMU virtual machine because, by default, VirtIO NIC provided
777 to the guest is configured to support only single TX queue and single RX
778 queue. To change this behavior, you need to turn on 'mq' (multiqueue)
779 property of all virtio-net-pci devices emulated by QEMU and used by DPDK.
780 You may do it manually (by changing QEMU command line) or, if you use
781 Libvirt, by adding the following string:
783 `<driver name='vhost' queues='N'/>`
785 to <interface> sections of all network devices used by DPDK. Parameter 'N'
786 determines how many queues can be used by the guest.This may not work with
787 old versions of QEMU found in some distros and need Qemu version >= 2.2.
789 ## <a name="qos"></a> 7. QOS
791 Here is an example on QOS usage.
792 Assuming you have a vhost-user port transmitting traffic consisting of
793 packets of size 64 bytes, the following command would limit the egress
794 transmission rate of the port to ~1,000,000 packets per second
796 `ovs-vsctl set port vhost-user0 qos=@newqos -- --id=@newqos create qos
797 type=egress-policer other-config:cir=46000000 other-config:cbs=2048`
799 To examine the QoS configuration of the port:
801 `ovs-appctl -t ovs-vswitchd qos/show vhost-user0`
803 To clear the QoS configuration from the port and ovsdb use the following:
805 `ovs-vsctl destroy QoS vhost-user0 -- clear Port vhost-user0 qos`
807 For more details regarding egress-policer parameters please refer to the
810 ## <a name="rl"></a> 8. Rate Limiting
812 Here is an example on Ingress Policing usage.
813 Assuming you have a vhost-user port receiving traffic consisting of
814 packets of size 64 bytes, the following command would limit the reception
815 rate of the port to ~1,000,000 packets per second:
817 `ovs-vsctl set interface vhost-user0 ingress_policing_rate=368000
818 ingress_policing_burst=1000`
820 To examine the ingress policer configuration of the port:
822 `ovs-vsctl list interface vhost-user0`
824 To clear the ingress policer configuration from the port use the following:
826 `ovs-vsctl set interface vhost-user0 ingress_policing_rate=0`
828 For more details regarding ingress-policer see the vswitch.xml.
830 ## <a name="vsperf"></a> 9. Vsperf
832 Vsperf project goal is to develop vSwitch test framework that can be used to
833 validate the suitability of different vSwitch implementations in a Telco deployment
834 environment. More information can be found in below link.
836 https://wiki.opnfv.org/display/vsperf/VSperf+Home
842 Please report problems to bugs@openvswitch.org.
845 [INSTALL.userspace.md]:INSTALL.userspace.md
846 [INSTALL.md]:INSTALL.md
847 [DPDK Linux GSG]: http://www.dpdk.org/doc/guides/linux_gsg/build_dpdk.html#binding-and-unbinding-network-ports-to-from-the-igb-uioor-vfio-modules
848 [DPDK Docs]: http://dpdk.org/doc
849 [libvirt]: http://libvirt.org/formatdomain.html
850 [Guest VM using libvirt]: INSTALL.DPDK.md#ovstc
851 [INSTALL DPDK]: INSTALL.DPDK.md#build
852 [INSTALL OVS]: INSTALL.DPDK.md#build