1 .. This work is licensed under a Creative Commons Attribution 4.0 International License.
2 .. http://creativecommons.org/licenses/by/4.0
3 .. (c) Cisco Systems, Inc
11 - NFVBENCH-152 Add service_mode method for debugging purpose
12 - NFVBENCH-150 Add support for VXLAN latency
13 - NFVBENCH-146 Add cache_size option
14 - NFVBENCH-151 Allocate hugepages on two NUMAs in nfvbenchvm
15 - NFVBENCH-149 Negative latency exception during NDR/PDR search
16 - NFVBENCH-148 Increase the waiting time based on # of instances
21 - NFVBENCH-147 Incorrect URL used for admin check in credentials
22 - Release the validation check for VxLAN networks
23 - NFVBENCH-145 Config file not found. No explicit error
24 - NFVBENCH-144 Trex cannot take account NFVBench config (platform thread id 0)
26 - NFVBENCH-140 Retrieve High Dynamic Range latency histograms with TRex v2.59
27 - NFVBENCH-143 Trex cannot start due to invalid config (platform None)
28 - NFVBENCH-141 Fix Openstack user admin role check
29 - NFVBENCH-139 Fix master_thread_id and latency_thread_id property checking
30 - NFVBENCH-95 Add HdrHistogram encodes returned by TRex to JSON results
31 - NFVBENCH-138 Use yaml.safe_load() instead of unsafe yaml load
32 - NFVBENCH-137 NFVbench generates wrong L4 checksums for VxLAN traffic
37 - Add L3 traffic management with Neutron routers
43 Major release highlights:
46 - test VM can now have idle interfaces
47 - test VM can be launched with multiqueue enabled
48 - upgrade to TRex v2.56
53 NFVbench will now follow its own project release numbering (x.y.z) which is independent of the OPNFV release numbering (opnfv-x.y.z)
55 Major release highlights:
57 - Dedicated edge networks for each chain
58 - Enhanced chain analysis
59 - Code refactoring and enhanced unit testing
60 - Miscellaneous enhancement
62 Dedicated edge networks for each chain
63 --------------------------------------
64 NFVbench 1.x only supported shared edge networks for all chains.
65 For example, 20xPVP would create only 2 edge networks (left and right) shared by all chains.
66 With NFVbench 2.0, chain networks are dedicated (unshared) by default with an option in
67 the nfvbench configuration to share them. A 20xPVP run will create 2x20 networks instead.
69 Enhanced chain analysis
70 -----------------------
71 The new chain analysis improves at multiple levels:
73 - there is now one table for each direction (forward and reverse) that both read from left to right
74 - per-chain packet counters and latency
75 - all-chain aggregate packet counters and latency
76 - supports both shared and dedicated chain networks
78 Code refactoring and enhanced unit testing
79 ------------------------------------------
80 The overall code structure is now better partitioned in the following functions:
82 - staging and resource discovery
86 The staging algorithm was rewritten to be:
88 - a lot more robust to errors and to handle better resource reuse use cases.
89 For example when a network with a matching name is discovered the new code will verify that the
90 network is associated to the right VM instance
91 - a lot more strict when it comes to the inventory of MAC addresses. For example the association
92 from each VM MAC to a chain index for each Trex port is handled in a much more strict manner.
94 Although not all code is unit tested, the most critical parts are unit tested with the use of
95 the mock library. The resulting unit test code can run in isolation without needing a real system under test.
101 Over 30 Jira tickets have been addressed in this release (Jira NFVBENCH-55 to NFVBENCH-78)
103 The Fraser release adds the following new features:
105 - support for benchmarking non-OpenStack environments (with external setup and no OpenStack openrc file)
106 - PVVP packet path with SRIOV at the edge and vswitch between VMs
107 - support logging events and results through fluentd
109 Enhancements and main bug fixes:
111 - end to end connectivity for larger chain count is now much more accurate for large chain count - avoiding excessive drops
112 - use newer version of TRex (2.32)
113 - use newer version of testpmd DPDK
114 - NDR/PDR uses actual TX rate to calculate drops - resulting in more accurate results
115 - add pylint to unit testing
116 - add self sufficient and standalone unit testing (without actual testbed)
119 OPNFV Euphrates Release
120 =======================
122 This is the introductory release for NFVbench. In this release, NFVbench provides the following features/capabilities:
124 - standalone installation with a single Docker container integrating the open source TRex traffic generator
125 - can measure data plane performance for any NFVi full stack
126 - can setup automatically service chains with the following packet paths:
127 - PVP (physical-VM-physical)
128 - PVVP (physical-VM-VM-physical) intra-node and inter-node
129 - can setup multiple service chains
132 - supports any external service chain (pre-set externally) that can do basic IPv4 routing
134 - drop rate and latency for any given fixed rate
135 - NDR (No Drop Rate) and PDR (Partial Drop Rate) with configurable drop rates
136 - traffic specification
137 - any fixed frame size or IMIX
138 - uni or bidirectional traffic
139 - any number of flows
140 - vlan tagging can be enabled or disabled
144 - fully configurable runs with yaml-JSON configuration
145 - detailed results in JSON format
146 - summary tabular results
147 - can send logs and results to one or more fluentd aggregators (per configuration)