| a0f984a6ee1b54c17328b05ca42675373b9dd005 |
|
27-Feb-2017 |
Alexey Kodanev <alexey.kodanev@oracle.com> |
net: dctcp01: apply tst_set_sysctl()
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com>
/external/ltp/testcases/network/dctcp/dctcp01.sh
|
| 5460eebd69695b30e36d08dac6a57be928a8f18a |
|
27-Feb-2017 |
Alexey Kodanev <alexey.kodanev@oracle.com> |
tst_netload: new options and improvements
* Add getopts parsing, make them transparent to 'netstress' tool
* cat 'netstress' output only on failure (not expected result)
* remove 'ping' type, ICMP traffic can be generated with tst_ping()
* tst_netload() sets TPASS/TFAIL flags depending on '-e' parameter --
expected result.
* allow to overwrite most significant parameters with environment
variables:
'TST_NETLOAD_CLN_NUMBER' -- number of clients,
'TST_NETLOAD_CLN_REQUESTS' -- client requests,
'TST_NETLOAD_MAX_SRV_REPLIES' -- max server replies after which
TCP connection will be closed.
Apply new usage in tests and virt_lib.sh. Change max_requests
parameter from 3 to 20 in virt_lib.sh. This will double the speed
of virt* tests but relative performance results should stay much
the same:
vxlan03 3 TINFO : run server 'netstress -R 3 -g 56792'
vxlan03 3 TINFO : check that server port in 'LISTEN' state
vxlan03 3 TINFO : run client 'netstress -l -H fd00:1:1:1::1 -a 2 -r 500000 -d res_ipv4 -g 56792'
vxlan03 3 TPASS : netstress passed, time spent '20049' ms
vxlan03 4 TINFO : time lan(20049) vxlan IPv4(33529) and IPv6(32828) ms
vxlan03 4 TINFO : IP4 vxlan over IP6 slower by 67 %
vxlan03 4 TINFO : IP6 vxlan over IP6 slower by 63 %
vs:
vxlan03 3 TINFO : run server 'netstress -R 20 -g 20201'
vxlan03 3 TINFO : check that server port in 'LISTEN' state
vxlan03 3 TINFO : run client 'netstress -l -H fd00:1:1:1::1 -a 2 -r 500000 -d res_ipv4 -g 20201'
vxlan03 3 TPASS : netstress passed, time spent '9477' ms
vxlan03 4 TINFO : time lan(9477) vxlan IPv4(16811) and IPv6(16572) ms
vxlan03 4 TINFO : IP4 vxlan over IP6 slower by 77 %
vxlan03 4 TINFO : IP6 vxlan over IP6 slower by 74 %
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com>
/external/ltp/testcases/network/dctcp/dctcp01.sh
|
| 8d61a03c86ab73f98ef4cb54d33e0b55fb8b3462 |
|
14-Nov-2016 |
Cyril Hrubis <chrubis@suse.cz> |
dctcp: Make use of tst_kvcmp
Signed-off-by: Cyril Hrubis <chrubis@suse.cz>
/external/ltp/testcases/network/dctcp/dctcp01.sh
|
| 7fc23dac69d60020cd8ce2c725ccbec89c8f2b79 |
|
20-Oct-2015 |
Alexey Kodanev <alexey.kodanev@oracle.com> |
network/dctcp: add dctcp congestion control algorithm test
The test compares cubic and dctcp congestion control algorithms
performance. One of the hosts emulates packets congestion with netem
loss 0.03%. Also ECN is enabled on the host, which is required
for dctcp algorithm to work properly.
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com>
/external/ltp/testcases/network/dctcp/dctcp01.sh
|