Age | Commit message (Collapse) | Author | Files | Lines |
|
A support of traffic capture was added into T-Rex.
It allows to write a functional tests, which will
verify proper vSwitch functionality by inspection
of packets received by T-Rex.
A testcase example was added into integration
testcases.
JIRA: VSPERF-556
Change-Id: I5ad28479ca2ec29760b68f24510af1a6d74866ae
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
Reviewed-by: Richard Elias <richardx.elias@intel.com>
|
|
Traffic can be captured also at DUT side. Two options are supported by
vsperf:
1) Traffic is captured inside VM after it has been processed
by vSwitch. This can be used for verification of vSwitch frame
modification functionality, including HW offloading at ingress side.
2) Another NIC under the test (NIC2) is added into standard VSPERF DUT
setup. Traffic is then forwarded from TGen through NIC and vSwitch
to NIC2 and then over patch cable back to NIC2, the vSwitch, NIC and
to the traffic generator. This setup supports also verification of HW
offloading at egress side of NIC2 and thus it can be used for validation
of smart NICS.
Both options above are traffic generator agnostic if compared to direct
support of traffic capture by traffic generator. This patch introduces
example testcases for both options.
Detailed documentation will be pushed as a standalone patch.
JIRA: VSPERF-556
Change-Id: I23e12e45768ae4dbe9442f74d8391c3d5b2c7895
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Richard Elias <richardx.elias@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
- Adds T-Rex verification step as implemented by other trafficgens
- Adds check in rfc2544 loop for if no packets are received to fail test
immediately
- Refactors the trial code to support the verification option
- Adds trial_run function
- Removed unused line speed configuration setting
JIRA: VSPERF-553
Change-Id: Ie324fe8fb6bf79fe0dc337b91af2bf83e901a8ab
Signed-off-by: Christian Trautman <ctrautma@redhat.com>
|
|
Adds packet learning option which is available on other
Traffic generators as a feature. This adds this feature to the
Trex implementation inside of VSPerf.
JIRA: VSPERF-547
Change-Id: Iaf4d0721b22eb780c25e29295c112d4fcb47b22c
Signed-off-by: Christian Trautman <ctrautma@redhat.com>
|
|
A set of regression tests for verification of OVS/DPDK
features was introduced. Following features are tested:
* DPDK NIC Support
* DPDK Hotplug Support
* RX Checksum Support
* Flow Control Support
* Multiqueue Support
* Vhost User
* Virtual Devices Support
* NUMA Support
* Jumbo Frame Support
* Rate Limiting
* Quality of Service
JIRA: VSPERF-539
Change-Id: I6c83e9c38e29f197a5236ac68c88d7045fef775b
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Signed-off-by: Martin Goldammer <martinx.goldammer@intel.com>
Signed-off-by: Richard Elias <richardx.elias@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
If VPP can't be executed properly, then VPP tests will be skipped
for VERIFY and MERGE jobs. It sometimes happens, that VPP can't
be executed at builders used for VERIFY & MERGE jobs execution.
These failures are not caused by patches themselves, but they
are blocking standard review & merge process. Thus VPP tests
will be suppressed in case that VPP can't be started at all.
This change is not related to DAILY job, which is executed
at POD dedicated to VSPERF project. It means that VPP tests
are always executed by DAILY job.
Change-Id: I4752601561b94a82482c403305bc3f24779f538b
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
|
|
|
|
Newer versions of VPP require an explicit definition of CLI socket
file for proper functionality. Otherwise vppctl is not able to connect
to running vpp process and thus vsperf is not able to configure
and control vpp.
Change-Id: Id0df75fd0894e22b1b4d7fa05e16f49efe20d3ab
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
VSPERF verifies if PATHS to the tools (vSwitch, VNF, etc.)
are valid. This check was modified to take into the account
tools required by selected testcase. It will help to avoid
runtime errors triggered by missing tools, which
are not essential for testcase execution.
Change-Id: Ifcfea983cad1bdbc8c1f08c7a0281001005410df
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
|
|
Currently all tools are loaded via loader utility, except load_gen.
Load_gens were loaded directly through component_factory.
This patch adds support to load load_gens through loader utility.
The changes are as follows:
1. Configuration changes:The common.conf include configuration of
directory. testcases.conf includes changes to load configuration,
where tool is no more part of the load-configuration. The
custom.conf has configuration of LOADGEN to be used - this configuration
replaces the earlier 'tool' configuration parameter.
2. loader_utility_changes: In loader.py, loadgen_loader is
defined, which is used in new get_loadgen function.
3. component_factory changes: in create_loadgen, similar
to other tools, the function just retuns the object of loadgen_class.
4. Renaming of Dummy load_gen: Loader fails to load properly a dummy
loadgen due to name-clash with dummy in pkt_gen. To avoid this name clash
dummy is renamed to dummyloadgen.
5. testcase changes: create_loadgen is now called with output of
loader's get_loadgen_class.
6. Fixed Pylint Errors and extra-space at the end.
7. Included CLI options support for --loadgen and --list-loadgens.
Thanks to Martin K.
8. Added the missing loadgen parameter in testcases.conf.
9. Fixed the missing comma error.
10. Added CI change in build-vsperf.sh
11. Fixed configuration reading in stress/stress.py
JIRA: VSPERF-533
Change-Id: I3fbb259618825a12fef55320a748a4f02509190b
Signed-off-by: Sridhar K. N. Rao <sridhar.rao@spirent.com>
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
|
|
This patch introduces several improvements and small bugfixes
of teststeps. These changes were identified during implementation
of OVS/DPDK regression tests.
Patch content:
* teststeps: step aliases were implemented
* teststeps: improved filtering by regex for any step, which returns string
or list of stings; filter will process all lines
* teststeps: support for log object
* teststeps: support for trafficgen get_results call
* teststeps: configurable suppression of step validation
* trafficgen: remove old results before traffic is executed
* trafficgen: support for flow control on/off (IxNet)
* trafficgen: support for configurable learning frames (IxNet)
* trafficgen: support for runtime changes of TRAFFICGEN_PKT_SIZES, _DURATION
and _LOSSRATE
* vnf: flush pexpect output of previous commands
* vnf: use execute_and_wait() to ensure correct cmds order
* vnf: dpdk vHost User interface name set according to its type,
e.g. dpdkvhostuserclient
* vswitch: support for OVS restart
* decap: simplify configuration of tunneling decapsulation tests
* settings: values of all configuration options are restored after TC execution
* modified formatting of test description used by --list
* testcase name and description is logged before its execution
* small bugfixes
JIRA: VSPERF-539
Change-Id: I550ba0d897ece89abd3f33d6d66f545c4d863e7b
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
|
|
Adds incremental source multi-stream functionality to
T-Rex traffic generator code in VSPerf.
JIRA: VSPERF-532
Change-Id: Ib5ba326699d89350ac1715c9a4276e5fa46a133e
Signed-off-by: Christian Trautman <ctrautma@redhat.com>
|
|
Number of VPP workers (aka PMD threads) was not specified
explicitly in default configuration file. Configuration
has been fixed to use the same cores as OVS with DPDK.
Change-Id: Icd3e5d6f429435fbc1ff580891f4b182d1c48729
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
Adds option to enable promiscuous mode for Trex traffic
generator to allow for SR-IOV tests to work correctly.
Trex by default only accepts packets based on destination
MAC address matching what is configured on server in
/etc/trex-cfg.yaml. This removes that limitation since
SR-IOV tests require mac addresses to be modified in the
generated packets to correctly route packets to the relevant
VFs.
JIRA: VSPERF-531
Change-Id: I19b64f364c8a1b656994c82c96f47a68b92486d8
Signed-off-by: Christian Trautman <ctrautma@redhat.com>
|
|
Changes introduced by patch:
* Ensure that correct results with given lossreate are returned
* Add threshold value to specify the accurancy of RFC2544 Throughput
measurement. Default value is 0.05.
JIRA: VSPERF-530
Change-Id: I8148deba9039c3a0feb6394d6671aa10c5afaf0a
Signed-off-by: Martin Goldammer <martin.goldammer6@gmail.com>
Reviewed-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
Topology are two physical servers, on first is trex and second is VSPERF.
Trex is running in stateless mode this means that on server where is located
trex repo is running trex binary file and VSPERF working with server via
python API.
JIRA: VSPERF-528
Change-Id: Id8819495325ebc13fdce365f4af0e040ce68cd0e
Signed-off-by: Martin Goldammer <martin.goldammer6@gmail.com>
Reviewed-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
Data reported to result DB were updated to comply
with recent API version and requirements.
JIRA: VSPERF-488
Change-Id: I6ab02ea54d99099e88b0d5ab6e9745ee8297f544
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Signed-off-by: Mars Toktonaliev <mars.toktonaliev@nokia.com>
Signed-off-by: Morgan Richomme <morgan.richomme@orange.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
|
|
A set of tests was introduced with focus on L3, L4 and VxLAN performance
of OVS and VPP. New testcases were created for phy2phy network scenario.
In case of PVP and PVVP, only OVS testcases are available.
Notes:
* two sets of OVS P2P testcases were created, one creates unique flow
for each IP address involved in the test (performance sensitive);
Second set inserts just one flow with large network mask (tests with
_mask suffix).
* three different types of VPP P2P L3 testcases were created to
demonstrate performance impact of multi ARP entries or IP routes.
* VPP multi ARP record based testcases use a set of "workarounds"
to load a large number of ARP entries. It is not possible to use
"set ip arp count" syntax, as it doesn't work well for large count
values (e.g. 60K)
* OVS VxLAN testcases utilize existing OP2P deployment scenario and thus
it can be used also with GRE and GENEVE tunnel types. Tunnel type
to be used is defined by test configuration option "Tunnel Type".
JIRA: VSPERF-518
Change-Id: I65adad976f12d8625d918a1996eb42693c511ee1
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
Reviewed-by: Cian Ferriter <cian.ferriter@intel.com>
|
|
Support of vhost-user client mode has been added for both
OVS and VPP. VSPERF configures vhost-user server mode
in vswitches by default for backward compatibility.
Support of vhost-user server mode is deprecated
in OVS and it can be removed in future releases. Once it will
happen, we have to change vsperf default behavior to
usage of vhost-user client ports at vswitch side.
This patch also fixes processing of boolean configuration
parameters passed via --test-params option. In the past
it was not possible to set boolean value to False through
--test-params option.
JIRA: VSPERF-517
Change-Id: I65e341f820edd6c720043a4acc74c5140b05db18
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
Versions of VSPERF tools were updated to use OVS
with DPDK 17.02 support by default. Currently used
VPP version uses DPDK 17.02 already.
VSPERF was updated to support new path to the dpdk
devbind tool.
JIRA: VSPERF-516
Change-Id: I004d0531553c974d604635d50f601cd4326f2f7d
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
A set of improvements was introduced to enhance step driven
testcases capabilities.
Details:
* delay among test steps is configurable by TEST_STEP_DELAY
parameter
* step driven tool function exec was renamed to exec_shell
* new step driven tool function exec_python was introduced to execute
a python code
* new step driven object sleep was introduced to pause test execution
for defined number of seconds.
* fixed bug in settings.validate_getValue() to correctly validate
access of parameters modified by TEST_PARAMS
* new #PARAM() macro was introduced to allow references among
configuration parameters
* multistream support has been added into ixnetrfc2544v2.tcl,
which is used for tunneling protocols test (op2p deployment)
* fixed bug in op2p deployment to list interfaces and flows
from both bridges involved in the test
* test report updated to state exact rfcxxxx type of traffic
type, e.g. rfc2544_continuous
* test report of step driven testcases was updated to contain
measured values from traffic generator in CSV report
* method for ovs flow comparison was modified to normalize
IPv4 CIDR network addr (e.g. 10.0.0.5/8 => 10.0.0.0/8)
JIRA: VSPERF-512
Change-Id: Ib4f38dcdfbf3820dd766b25520da0ad0c81f3293
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
Reviewed-by: Ciara Loftus <ciara.loftus@intel.com>
|
|
JIRA: VSPERF-510
This patch is used to implement the mechanism to pass extra qemu
cpu options required to achieve better results/performance. Also
used to implement a mechanism to pin the vCPU threads to another
pCPU core which helps in reducing the noise from housekeeping
thread and packet loss.
Change-Id: Ic40fd47d7c4f5556f7e240c6ca671a0535d06ece
Signed-off-by: Gundarapu Kalyan Reddy <reddyx.gundarapu@intel.com>
|
|
VSPERF makefiles were updated to automatically download and compile
VPP. VPP will not be installed into the OS, but it will be executed
from VSPERF's src subdirectory. Thus underlying OS is not affected
by VPP build required by VSPERF. It also allows quick switch among
different versions of various tools used by VSPERF.
As part of this patch, VERIFY and MERGE CI jobs were updated to
build VPP and test it with basic set of integration tests.
JIRA: VSPERF-493
Change-Id: I958b9031c4fefc87c4c63a471d2ba1a0db1eaaa4
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
|
|
|
|
For some testcases it is useful to have only one connection between
traffic generator and DUT. Thus IxNet class was extended to support
such connection. Feature is enabled if both IXIA ports are set to
the same value.
New integration testcase ixnet_pvp_tput_1nic was created to demonstrate
this feature.
Unused code was removed from 3rd_party/ixia/ixnetrfc2544.tcl to simplify
introduction of new features and to speed up IXIA configuration.
JIRA: VSPERF-503
Change-Id: I92ab24fa58eeb1ccb7f00ac7f3530ffca7b99241
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
Adds final verification option to Xena traffic generator
inside VSPerf. Works by running the Xena2544.exe again
after the initial run is complete with modified values
inside the json file. If the final verify fails it will
revert the settings and continue the in a pattern based
on if the smart search is enabled.
JIRA: VSPERF-506
Change-Id: Ic272628c489b88989f67549bd020af7354dc9b91
Signed-off-by: Christian Trautman <ctrautma@redhat.com>
|
|
Add jumbo frame support for all packet forwarding applications
inside guest for pxp testing. Enable jumbo frame support for
OvsVanilla, OvsDpdkVhostUser, and Packet forwarding.
Add minor fix for missing method issue when running packet
forwarding tests. The get_version method was missing in
the packet_forwaring core module.
JIRA: VSPERF-501
Change-Id: Ia99975f47c64259ed2566bde3c85b2779c309e80
Signed-off-by: Christian Trautman <ctrautma@redhat.com>
|
|
During testing it is sometimes useful to generate pure L2 traffic
without L3 or L4 related headers. IxNetwork TCL script was updated
to support this option.
Generic TRAFFIC array was enhanced by on/off switch for L3 and L4
headers. Thus it can be implemented by other traffic generators
in the future, if this feature will prove itself useful.
JIRA: VSPERF-500
Change-Id: I723c703e5d6ef609fd5b7db366871278a4730203
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
Restriction of 64K streams was limited only to L4 stream
type. In case of L3 and L2 stream type, it is possible to
define unlimited number of streams to be simulated
by traffic generator.
JIRA: VSPERF-498
Change-Id: I4ce3253299473127d135c01543999a8593b9556f
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Bill Michalowski <bmichalo@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
Support of VPP was implemented into VSPERF. Initial implementation
uses step driven testcases to configure P2P, PVP and PVVP network
scenarios. These testcases were prepared for three RFC2544 traffic
types, i.e. throughput, continuous stream and back to back.
VPP configuration is driven by new configuration option VSWITCH_VPP_ARGS.
It is possible to use three types of l2 port connection supported
by VPP, i.e. l2 xconnect (default), l2patch and l2 bridge features.
Configuration is driven by parameter VSWITCH_VPP_L2_CONNECT_MODE.
JIRA: VSPERF-495
Change-Id: Idebef9b10fb0d70796adb3405fec77302de00a7e
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Trevor Cooper <trevor.cooper@intel.com>
|
|
This patch adds RFC-2889 address-learning-rate test and
address-caching-capacity tests. The patch includes following changes
1. 01_testcases.conf: testcase definitions are added for new tests
2. results_constants.py: 3 new results constants are added, removed
unsupported constants
3. testcenter-rfc2889-rest.py: Added support to these two new
tests. Removed un-necessary synchronization of db-file. 5 new
configuration parameters are added - which are explained in (4).
4. trafficgen.rst: Configuration description for rfc2889 tests are
added. Stale information is removed.
5. testcenter.py: Support for two new tests are added. Separate
functions for results analysis and parameter settings are added.
6. traffic_controller_rfc2889.py: Calls to functions in testcenter.py,
based on the traffic_type, are added.
Removed trailing whitespaces from trafficgen.rst
JIRA: VSPERF-395
Change-Id: Ie6ca87519b7736e7410d7f0e60904579934af24a
Signed-off-by: Sridhar K. N. Rao <sridhar.rao@spirent.com>
|
|
|
|
Modifies RTE_TARGET in conf/10_custom.conf which overwrites
default value in 02_vswitch.conf file.
JIRA: VSPERF-359
Change-Id: I83561d94543a6543764cd76eb4376ccdc091a4cd
Signed-off-by: Christian Trautman <ctrautma@redhat.com>
|
|
Documentation was updated as follows:
* description of vsperf downloading was added
* VNF section was updated
* additions to installation and virtualnev related parts
* file permissions of all documentaton files
was unified on 644
* README.md was updated with links to doc files
and online docs
* 3rd_party/readme was renamed and updated to follow
markdown syntax
JIRA: VSPERF-441
Change-Id: I648c6a8f0bb283bbc0875fd6fecd2539ca0c9056
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Bill Michalowski <bmichalo@redhat.com>
Reviewed-by: Antonio Fischetti <antonio.fischetti@intel.com>
Reviewed-by: Sridhar Rao <sridhar.rao@spirent.com>
Reviewed-by: Martin Goldammer <martinx.goldammer@intel.com>
|
|
Adds a configurable learning preemptive option when running
xena continuous traffic tests. This resolves a missing packet
or two when running OVS tests at flow rates that should not
drop packets.
JIRA: VSPERF-473
Change-Id: I056578ed9c9181287d97aea55a002f09aac14e81
Signed-off-by: Christian Trautman <ctrautma@redhat.com>
|
|
Traffic generated by traffic generator is based on default
values and their modifications specific to particular testing
scenario. Traffic default values were defined inside VSPERF
code and it was not possible to change them. This patch introduces
new TRAFFIC dictionary inside 03_traffic.conf. Thus user can
modify any of TRAFFIC values either in configuration file
or by CLI or by 'Parameters' section of testcase definition.
Following CLI options were obsoleted by this patch:
'bidirectional', 'traffic_type', 'iload', 'multistream',
'stream_type' and 'pre-installed_flows'
Following CLI option was renamed to be consistent with other options:
'tunnel_type' => 'TUNNEL_TYPE'
Following sections of testcase definition were obsoleted:
"Traffic Type", "biDirectional", "MultiStream", "Stream Type",
"Pre-installed Flows", "Flow Type" and "iLoad"
New TRAFFIC dictionary should be used instead of old CLI options
and old testcase definition sections. Testcase definitons,
yardstick sample testcases and documentation were updated to reflect
configuration changes.
JIRA: VSPERF-433
Change-Id: I03a388c766491d5688e715f6d7b51e8e0377ec27
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Bill Michalowski <bmichalo@redhat.com>
Reviewed-by: Antonio Fischetti <antonio.fischetti@intel.com>
Reviewed-by: <sridhar.rao@spirent.com>
|
|
Requires DPDK 16.11 or greater to support vdev flags.
Initial support for TestPMD to support guests.
Allows vsperf to execute TestPMD as a switch for pvp
test scenarios.
Can be increased in functionality later to support
multiple guest configs.
JIRA: VSPERF-406
Change-Id: I67a5a355c990ca6cfcbb5845a2beaf1c1f21f5f0
Signed-off-by: Christian Trautman <ctrautma@redhat.com>
|
|
Adds installer scripts for RHEL 7.3 OS.
Modified dpdk makefile to allow upstream dpdk to build.
Prevents upstream OVS from building due to kernel incompatibilities.
Provides instructions on setting up RHEL 7.3 for usage with binary
OVS packages when testing with OVS Vanilla.
Tested under RHEL 7.3
JIRA: VSPerf-322
Change-Id: If294d465f28f60a1c514a7b661ce4696ac3b5b73
Signed-off-by: Christian Trautman <ctrautma@redhat.com>
|
|
|
|
Following changes are added following first review
1. Descriptions in integration/01_testcases.conf
2. Comment in conf/01_testcases.conf
3. Checking all cases of traffic types in traffic_controller_rfc2544.py
4. Removing whitespaces
5. Missing changes (throughput->rfc2544_throughput)
6. Syntax error in core/traffic_controller_rfc2544.py
Currently the traffic-type values that are provided as part of the
test case definition needs some fix to support additional traffic
- apart from rfc2544. The Current and the proposed changed is summarized
in the below table:
|--------------------------------------------------------------------|
| Parameter | Existing traffic_type value | Changed to |
----------------------------------------------------------------------
| Throughput | rfc2544 OR throughput | rfc2544_throughput |
----------------------------------------------------------------------
| Back2Back | back2back | rfc2544_back2back |
----------------------------------------------------------------------
| Continuous | continuous | rfc2544_continuous |
----------------------------------------------------------------------
JIRA: VSPERF-420
Change-Id: I22bd897f6389d2b568a6a3735c67f0003e813445
Signed-off-by: Sridhar K. N. Rao <sridhar.rao@spirent.com>
|
|
Vswitchperf specific class in yardstick was updated to reflect
new functionality of --test-params. These changes have impact
on vsperf documentation and yardstick samples.
As part of this fix following changes were made:
* trafficgen mode can be executed with default configuration
* yardstick specific TC samples were updated
* dummy traffic generator was enhanced to support back2back
* dummy traffic generator was enhanced to accept traffic
results from configuration
JIRA: VSPERF-422
JIRA: VSPERF-205
Change-Id: Ibeaa6ef46bde453e3ca5dee6e4568fc8c3853521
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Bill Michalowski <bmichalo@redhat.com>
Reviewed-by: Antonio Fischetti <antonio.fischetti@intel.com>
Reviewed-by: <sridhar.rao@spirent.com>
|
|
VSPERF by default requires allocation of hugepages with total
size of 10GB. However for MERGE and VERIFY jobs less memory is
required and default size of 10GB can cause issues. Thus amount
of memory was limited and other minor enhancements were introduced:
* sysstat system package is installed at nodes running VERIFY
and MERGE jobs
* WHITELIST_NICS was set to empty list by default to avoid
automatic bind of NICs to DPDK. Old default value could
cut off network connection to DUT.
* imports were fixed at both integration.py and testcase.py
JIRA: VSPERF-391
Change-Id: Ic07d68f4591ae697c1b8edeae6c0cffe47c065ce
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Bill Michalowski <bmichalo@redhat.com>
Reviewed-by: Antonio Fischetti <antonio.fischetti@intel.com>
|
|
|
|
In the past, only a few configuration parameters could be
modified via --test-params CLI argument and it was not easy
to find out their complete list.
This patch adds support for generic modification of any
configuration parameter via CLI argument --test-params
or by "Parameters" section of testcase definition.
Thus it is possible to customize a vsperf configuration
environment per testcase or for each vsperf execution.
Old CLI parameters duration, pkt_sizes, rfc2544_tests
and rfc2889_trials were renamed to TRAFFICGEN_DURATION,
TRAFFICGEN_PKT_SIZES, TRAFFICGEN_RFC2544_TESTS and
TRAFFICGEN_RFC2889_TRIALS to be consistent with
other configuration parameters.
JIRA: VSPERF-375
Change-Id: I50a1f4ff7250d754aa8af0295a9c7c1be8151175
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Bill Michalowski <bmichalo@redhat.com>
Reviewed-by: Antonio Fischetti <antonio.fischetti@intel.com>
Reviewed-by: <sridhar.rao@spirent.com>
|
|
IXIA traffic generator can be controlled by two different
VSPERF's wrappers IxNet (preferred) and Ixia (legacy and
outdated). It has been found, that Ixia wrapper can't be used
because of missing "packetsize" traffic item. Also only
IXIA OS 6.60 was supported and never IXIA OS versions
were not working. Both limitations were fixed.
Also reported results were fixed to show correct values
for MBPS items.
JIRA: VSPERF-417
Change-Id: I191aa9fee2683877ce373b6aa7cd1c72e9847568
Signed-off-by: Martin Klozik <martinx.klozik@intel.com>
Reviewed-by: Al Morton <acmorton@att.com>
Reviewed-by: Christian Trautman <ctrautma@redhat.com>
Reviewed-by: Bill Michalowski <bmichalo@redhat.com>
Reviewed-by: Christo Kleu <christo.kleu@netronome.com>
|