Hi,
On 2016年09月23日 16:31, Poonam Ghosh wrote: > Additionally if you see the Specification requirement for the EXT-335 : > Instruction: Stat-Trigger stat thresholds: Generate an event to the > controller if some of the flow statistics cross one of the stat threshold > values. > Which is what is the part which we wanted to check from the Unit test case > in ryu supporting this feature. > > We agree with you that the OFPStats compose the values w.r.t the below > parameters but it is equally important to understand that if controller is > looking for the Extensible statistics (unlike the one which is done in > OXM) so does that mean switch will internally compare these values from > the stats and trigger an EVENT to encode or prepare an OXS message format > after every X times the request is received (as part of the FLOW_MOD) from > the controller . What are your thoughts for the same? Sorry if I misunderstood what you mean. Well, whether the controller need to count/monitor OXS messages is depending on the design of the user Ryu application, I guess. But mostly at the controller side, OXS seems to be used for requesting the setting the stat thresholds with the OFPIT_STAT_TRIGGER instruction, and the controller does not need to compare these values internally. Because the statistics trigger mechanism seems to be strongly depending on the switch's implementations. Thanks, Iwase > > Argument > Value > Description > duration > Integer 32bit*2 > Time flow entry has been alive. This field is a tuple of two Integer > 32bit. The first value is duration_sec and the second is duration_nsec. > idle_time > Integer 32bit*2 > Time flow entry has been idle. > flow_count > Integer 32bit > Number of aggregated flow entries. > packet_count > Integer 64bit > Number of packets matched by a flow entry. > byte_count > Integer 64bit > Number of bytes matched by a flow entry. > > Thanks & Regards > Poonam Ghosh > Technical Architect > TCS > > > From: Poonam Ghosh/BLR/TCS > To: Ryu-devel <ryu-devel@lists.sourceforge.net>, Iwase Yusuke > <iwase.yusu...@gmail.com> > Date: 09/22/2016 11:50 PM > Subject: Re: [Ryu-devel] Query regarding the Statistics Trigger > Instruction -- OFPInstructionStatTrigger > > > First of all thanks Iwase for your reply to this thread. > > Yes, currently we do not have a functional code which supports all the > features supporting the OF version 1.5. > And that is the reason we are planning to implement the EXT features to > support the handling of Flow statistics for flexible encoding fro Egress > tables which is currently not been supported. > > I have gone through the OFV1.5 Specification for EXT-334 and EXT-335 and > started with prototyping the features requirement code and will share with > you once the patch with the EXT-335 is available. > The idea is to test the RYU controller with the EXT-335 switch > functionality so that it becomes easy to validate the requirement and to > check how the statistics is responded to the controller. > > The below query was to understand on the bitmap flags and how are the > thresholds been calculated to decide on the setting of the bitmap flags > and under what conditions the OXS stats are responded as part of the FLOW > STATS REPLY (for the corresponding FLOW_MOD). > > Primarily i will keep this thread of technical discussion Open for > discussing the RYU v1.5 changes and how can we leverage the controller > behavior to be tested with the switch functionality. > > Thanks & Regards, > Poonam Ghosh > Technical Architect,TCS > > > > > From: Iwase Yusuke <iwase.yusu...@gmail.com> > To: poonam.gh...@tcs.com > Cc: ryu-devel@lists.sourceforge.net > Date: 09/22/2016 06:07 AM > Subject: Re: [Ryu-devel] Query regarding the Statistics Trigger > Instruction -- OFPInstructionStatTrigger > > > > Hi, > > > On 2016年09月22日 17:08, Poonam Ghosh wrote: >> Hello, >> >> We have been referring the RYU source code to understand the OF Specific > instructions primarily OFV1.1, OFV1.2, OFV1.3,OFV1.4 and OFV1.5. >> Currently while simulating the tests with Instruction > OFPIT_STAT_TRIGGER(Specification EXT-335) we got couple of Open points to > understand more on the implementation part which needs more clarity to > understand. >> >> As part of the Specification details for the EXT-335 we understand that > "This Instruction is primarily required to send out Openflow Extensible > stats (as part of flexible encoding) to Controller based on certain > thresholds which are user defined". Please correct my understanding for > the same? >> >> So this is how the Instruction looks with the bitmap Flags shown below: >> >> enum ofp_stat_trigger_flags { >> OFPSTF_PERIODIC = 1 << 0, /* Trigger for all multiples of thresholds. > */ >> OFPSTF_ONLY_FIRST = 1 << 1, /* Trigger on only first reach threshold. */ >> }; >> /* Instruction structure for OFPIT_STAT_TRIGGER */ >> struct ofp_instruction_stat_trigger { >> uint16_t type; /* *OFPIT_STAT_TRIGGER* */ >> uint16_t len; >> uint32_t flags; >> struct ofp_stats thresholds; >> }; >> OFP_ASSERT(sizeof(struct ofp_instruction_stat_trigger) == 16); >> >> Based on the above understanding and the current RYU implementation it > is unclear of how the Flags and thresholds are derived and based on what? >> Because if we really see the RYU python scripts for VERSION1.5 we see > the hardcoded values for duration, idle_time , flow_count, packet_count, > and byte_count as part of the OXS fields filling. > > What do you exactly mean "hardcoded"? > "OFPStats" class can compose the arbitrary flow stats fields which are > defined in OpneFlow 1.5. > And the API is similar to OFPMatch. > > http://ryu.readthedocs.io/en/latest/ofproto_v1_5_ref.html#ryu.ofproto.ofproto_v1_5_parser.OFPStats > > >> For our understanding we are not clear on how can we decide on the above > Fields of when the Instruction trigger needs to set the*PERIODIC FLAG* or > *ONLY_FIRST* flag as part of the FLOW_STATS REPLY To controller. > > For example, if you want to send the OFPIT_STAT_TRIGGER instruction with > the PERIODIC flag and > packet_count=100 in the stats field: > instructions = [ > parser.OFPInstructionStatTrigger( > flags=ofproto.OFPSTF_PERIODIC, > thresholds=parser.OFPStats(packet_count=100), > ) > ] > > BTW, unfortunately we haven't found the switch implementation which fully > supports OpenFlow 1.5 > including EXT-335. > Do you know or have any implementation for supporting OpenFlow 1.5? > > Thanks, > Iwase > > >> >> Much appreciate your response for the same. >> >> Thanks & Regards >> Poonam Ghosh >> Technical Architect, TCS >> >> =====-----=====-----===== >> Notice: The information contained in this e-mail >> message and/or attachments to it may contain >> confidential or privileged information. If you are >> not the intended recipient, any dissemination, use, >> review, distribution, printing or copying of the >> information contained in this e-mail message >> and/or attachments to it are strictly prohibited. If >> you have received this communication in error, >> please notify us by reply e-mail or telephone and >> immediately and permanently delete the message >> and any attachments. Thank you >> >> >> >> > ------------------------------------------------------------------------------ >> >> >> >> _______________________________________________ >> Ryu-devel mailing list >> Ryu-devel@lists.sourceforge.net >> https://lists.sourceforge.net/lists/listinfo/ryu-devel >> > > > > > ------------------------------------------------------------------------------ > > > > _______________________________________________ > Ryu-devel mailing list > Ryu-devel@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/ryu-devel > ------------------------------------------------------------------------------ _______________________________________________ Ryu-devel mailing list Ryu-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/ryu-devel