Operating System: All
Platform: All

So, after discussion with Michal Rehák, this is what I am currently trying to achieve:

• Split giant ActionConvertor (1215 lines) and MatchConvertorImpl (2000 lines) classes to smaller pieces to be more managable and get rid of big if-else chains in these two classes
• Update all convertors in package org.opendaylight.openflowplugin.openflow.md.core.sal.convertor to use one generic method to convert between SAL and OPF model (and vice versa) if possible
• Update and fix all unit tests for convertors in above package to increase coverage and test all possibilities of different conversion types

So, in short, with these changes, convertors (converters) should be more manageable and easier to modify, when needed.

Here are links to GERRIT:

• https://git.opendaylight.org/gerrit/#/c/40863/ - ConvertorManager
• https://git.opendaylight.org/gerrit/#/c/40865 - TableFeaturesConvertor
• https://git.opendaylight.org/gerrit/#/c/40868 - TableFeaturesResponseConvertor
• https://git.opendaylight.org/gerrit/#/c/40870 - MeterConvertor
• https://git.opendaylight.org/gerrit/#/c/40875 - MeterStatsResponseConvertor, MeterConfigStatsResponseConvertor
• https://git.opendaylight.org/gerrit/#/c/40877 - PortConvertor
• https://git.opendaylight.org/gerrit/#/c/40929 - MatchConvertor
• https://git.opendaylight.org/gerrit/#/c/40937 - ActionConvertor, ActionResponseConvertor
• https://git.opendaylight.org/gerrit/#/c/40944 - GroupConvertor, GroupStatsConvertor, GroupDescStatsConvertor
• https://git.opendaylight.org/gerrit/#/c/40986 - PacketOutConvertor
• https://git.opendaylight.org/gerrit/#/c/40994 - FlowConvertor, FlowStatsResponseConvertor, FlowInstructionResponseConvertor

Attaching results of CBench tests on master and with new convertor changes, both using same configuration and on same machine.

Configuration:
controller: localhost
port: 6633
loops: 20
msPerTest: 10000
macPerSwitch: 1000
startupDelay: 1000
warmup: 1
switches: 16

To see actual graph, just open out3d.pdf, or run `gnuplot out3d.plot`.
For each branch (master, convertors) I did 3 tests to gather more data.

Attachment cbench-results.tar.gz has been added with description: CBench test results

Hi, this may have introduced a regression in NetIDE.

See https://gist.github.com/mosieurlaurin/059de066ef2849489534da921e4378bf

NetIDE cannot compile properly.

I submitted patch that will fix that compile errors to NetIDE project here: https://git.opendaylight.org/gerrit/#/c/43302