aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorVratko Polak <vrpolak@cisco.com>2020-09-24 13:11:45 +0200
committerVratko Polak <vrpolak@cisco.com>2020-09-28 15:00:43 +0000
commit34dadfe8d168b72340b497469ee6550349689f1a (patch)
tree389472804c60dc665e9a14340327504187beb51d
parente15746adc0c6b54dc74ea40fc86e794f90ecabf9 (diff)
Constants: Lower Mellanox pps limit
The previous 60 Mpps stil leads to ~30% duration stretching. + Add comment on why 36 Mpps was chosen as the new limit. Change-Id: Ic11e8ece03939bdc8680cd7bc4122373583a2f17 Signed-off-by: Vratko Polak <vrpolak@cisco.com>
-rw-r--r--resources/libraries/python/Constants.py8
1 files changed, 7 insertions, 1 deletions
diff --git a/resources/libraries/python/Constants.py b/resources/libraries/python/Constants.py
index 437c97f187..3775d98377 100644
--- a/resources/libraries/python/Constants.py
+++ b/resources/libraries/python/Constants.py
@@ -285,7 +285,13 @@ class Constants:
u"Intel-XL710": 18750000,
u"Intel-XXV710": 18750000,
u"Intel-E810CQ": 58500000,
- u"Mellanox-CX556A": 60000000, # 148809523,
+ # 2n-clx testbeds show duration stretching on high rates,
+ # depending on encapsulation TRex has to generate.
+ # 40 Mpps is still too much for dot1q (~8% stretching).
+ # 36 Mpps is around the maximal VPP throughput (l2patch 4c8t).
+ # Vxlan traffic will still show stretching at 36 Mpps (>12%),
+ # but we do not care about those tests that much.
+ u"Mellanox-CX556A": 36000000, # 148809523,
u"Amazon-Nitro-50G": 1500000,
u"virtual": 14880952,
}