Presumably you mean by imposing "tail dropping" on the queue?
As far as I know, there is not currently a parameter for this in the spec.
You could add an experimenter queue property to represent this.
However note that all queue *configuration* is still outside of the OF wire
protocol and requir
Thanks for pointing this out. openflow.org will be directed to the ONF
site soon, but the content will continue to be available as an archive.
I'm not sure contents will be updated before archiving, so we may need to
revisit this.
-Dan
On Fri, May 31, 2013 at 3:08 PM, Tharaneedharan Vilwanatha
On Tue, May 21, 2013 at 8:03 AM, Eder Leão Fernandes <
ederleaofernan...@gmail.com> wrote:
> I have the same view from Ben. But I believe that the contest is also
> asking for something like the OpenFlow Connection manager on Indigo Agent.
> (plus the unmarshaling/marshaling libraries, which are g
Traditionally, ofdatapath was the component that did the data path
forwarding. It was plumbed with data plane interfaces and received and
forwarded packets.
ofprotocol was the component that managed the connection to the openflow
controller.
As directives that updated the forwarding data path ta
ush the metadata back as a MPLS
> header before forwarding the packet.
> Is it possible to do the above sequence of operations?
>
> On Wed, Feb 27, 2013 at 12:31 AM, Dan Talayco
> wrote:
>
>> I believe there are OVS extensions that allow you to do this generically
>>
I believe there are OVS extensions that allow you to do this generically
(store a field in a register). You can do it "one-off" with 1.2 by
matching on the MPLS tag and then setting the metadata to that value. But
this is probably not what you want (as it requires a different rule for
each MPLS t
In your log, you'll see lines like
14:23:33.951 root : ERROR : RESULT: 2020 flows inserted
14:23:33.951 root : ERROR : RESULT: 1919 flows reported
This reports the number of flows for the test running that were attempted,
and those that succeeded. Note that the inserts are done
For each address (IP src, dest), a 6 bit number is provided. This is the
number of bits to ignore, starting from the low order bits. This is meant
be consistent with how wildcards are represented, but it ends up being
different from how net masks are usually represented; the netmask is the
bitwis
Have you tried
git clone git://openflow.org/openflow.git
On Mon, Nov 12, 2012 at 1:09 AM, Suneth Namal wrote:
>
>
>
> Cong Wang writes:
>
> >
> >
> > hi all,
> >
> >I'm setting up an OpenFlow Switches on NetFPGAs. When I type: "git
> clone git://openflowswitch.org/openflow.git"
> >
Just a note: There is currently work going on on a tool called Loxi. The
plan is to have a Python back-end which will effectively be a replacement
for pylibopenflow and will support multiple OF versions. The hope is that
OFTest can be ported to this without too much effort.
This is only informa
There are some known issues related to the way OVS canonicalizes flows and
the way flow_query works. These are definitely at the root of many of the
errors.
-Dan
On Tue, Oct 23, 2012 at 2:09 PM, Rob Sherwood wrote:
> Hi Maryam,
>
> Historically OVS has passed all of the OFTest tests (IIRC, I l
All versions of OpenFlow since 1.0 support vendor (later "experimenter")
extensions. You use a standard OF header (with version, xid and length)
and set the type field to 4 (OFPT_VENDOR in 1.0). The body of the message
can have any format you like after that as long as the length field in the
hea
On Tue, Oct 2, 2012 at 9:50 AM, Ben Pfaff wrote:
> On Tue, Oct 02, 2012 at 09:46:19AM -0700, Dan Talayco wrote:
> > On Tue, Oct 2, 2012 at 9:11 AM, Ben Pfaff wrote:
> >
> > > This interpretation seems needlessly wasteful. It forces the switch
> > > to keep a
On Tue, Oct 2, 2012 at 9:11 AM, Ben Pfaff wrote:
> This interpretation seems needlessly wasteful. It forces the switch
> to keep an extra copy of the original flow match around, instead of
> allowing it to convert it to an internal form and discard the
> original.
>
This is how the Indigo2 refe
I'm not sure you found this in OFTest, but there's a branch for 1.1 support
named oft-1.1. You can track it with "git checkout -t origin/oft-1.1".
There may be some useful tests in there, though the branch hasn't been
updated since Feb 2011.
-Dan
On Fri, Aug 17, 2012 at 6:43 AM, bhaskar jain w
On Tue, Aug 14, 2012 at 2:04 PM, Daniel Philip wrote:
> Hi all,
>
> By definition: When an OpenFlow switch receives a packet for the first
> time, for which it has no matching flow entries, it sends this packet to
> the controller. The controller then makes a decision on how to handle this
> pack
I haven't used Eclipse with this setup, but I expect you will still need
to install the python modules. I believe this will happen if you cd to the
tools/munger directory and do make install. Presumably Eclipse can find
the libraries once they're installed.
-Dan
On Sat, Jun 30, 2012 at 12:04
As Zoltan pointed you, you should not use --platform=remote. In fact with
your veth settings you shouldn't need to specify --platform at all.
Can you tell us a little more about your host? What version of Linux, etc?
You might have more luck running against OVS than the Stanford reference
desig
Hi Ali--
This is a little strange. Those changes went in in October and I think
they should still be there. We've verified it works on a version here.
Can you verify the SW version on your switch? The ver command should show
the 2012-03-19 date and IODS.
-Dan
On Thu, Jun 21, 2012 at 12:01
Right; the run_switch script expects the old Stanford openflow image
to be installed in the peer to oftest. I think this version will
work:
git clone git://gitosis.stanford.edu/openflow.git
You'll need to build it which hopefully will just mean
cd openflow
./boot.sh
./configure
make
-Dan
On
On Thu, Jun 7, 2012 at 12:19 AM, Dan Talayco wrote:
> First, note that the precision with which durations are reported is in
> nano-seconds in flow stats and expiration messages. This provides support
> for more accurate measurements using this info.
>
> There has not been muc
First, note that the precision with which durations are reported is in
nano-seconds in flow stats and expiration messages. This provides support
for more accurate measurements using this info.
There has not been much of a call for increasing the granularity of
time-out settings below 1 second. A
Assuming you're using veth interfaces to connect the parts together, you
should use local.py (which doesn't really do anything in the default case).
See the references to veth in oftest/tests/oft; you may need to use
--base-if-index.
-Dan
On Wed, May 23, 2012 at 1:11 AM, sonny sonny wrote:
>
Note that OpenFlow 1.2 has the notion of a group which can have a failover
algorithm. Changes to links in the group still result in a notification to
the controller, but can also move traffic to another member of the group
without controller intervention.
This in itself is not a solution, but it
xperimentally on a low bandwidth stream.
You would need to evaluate the bandwidth available on the network
connecting to the controller, and bear in mind that you'll be using 2x the
data plane bandwidth you want to encrypt. Then make sure the added latency
of doing the encryption this wa
Inline
In my case however, the packet:
>>>
>>> 1. arrives at dpid1
>>> 2. is tx to the controller
>>> 3. I insert an action to rewrite the mac address such as
>>> 00:00:00:00:00:fc
>>> 4. I verify that the flow mod was installed using dpctl on the switch
>>> 5. However, the pkt is not transmitted
From: yap...@gmail.com (mailto:yap...@gmail.com) [yap...@gmail.com
> (mailto:yap...@gmail.com)] on behalf of kk yap [yap...@stanford.edu
> (mailto:yap...@stanford.edu)]
> Sent: May 10, 2012 13:55
> To: Heming Wen
> Cc: Dan Talayco; openflow-discuss@lists.stanford.edu
> (mailto:openf
n other words, how is the OpenRoads controller setup in
> > order to control both the switches and the AP at the same time?
> >
> > Thank you,
> >
> > Heming
> >
> > From: Dan Talayco [dan.tala...@bigswitch.com
&g
I believe you're basically asking the switch to route (or at least switch)
between the management port and the dataplane. The short answer is that the
switch won't do this and it's not planned to be supported.
You probably want "physical in-band management" where the Pronto switch is
managed
com)>
> Date: Tue, May 8, 2012 at 8:54 PM
> Subject: Re: [openflow-discuss] OFTest failure
> To: Dan Talayco mailto:dtala...@stanford.edu)>
>
>
> Hi Dan,
>
>
> It is running on Ubuntu but not on fedora, I have tested this.
>
>
> On ubuntu following 3
SIC:
>
> PCI unit 0: Dev 0xb538, Rev 0x11, Chip BCM56538_B0, Driver BCM56634_B0
>
> It seems our chip is older than yours (BCM56538 < BCM56634)
> May be it is the reason ?
>
> Thanks
>
> Nam
>
> > From: "Dan Talayco" mailto:dtala...@stanford.edu)
are able to run OFTest with all modules
> in one go on OVS .
>
> -Sonny
>
> On Fri, May 4, 2012 at 10:04 PM, Dan Talayco (mailto:dtala...@stanford.edu)> wrote:
> >
> >
> > On Friday, May 4, 2012 at 4:31 AM, sonny sonny wrote:
> >
> > >
:
> C321102416A10v4B618j
> Type: Hard Disk
> Capacity: 1953.9 MB = 1.9 GB (4001760 x 512)
>
> Could you compare with your switch ? Thanks
>
> Nam
>
> On 05/03/2012 05:57 PM, Dan Talayco at gmail wrote:
> > Oops, wait. I misread th
0 0.0.0.0:47943 (http://0.0.0.0:47943)
> 0.0.0.0:* 1192/rpc.statd
> udp0 0 0.0.0.0:944 (http://0.0.0.0:944)
> 0.0.0.0:* 1192/rpc.statd
> udp0 0 :::111 :::*
(Sorry, sent too quickly)
On Thursday, May 3, 2012 at 10:06 PM, Dan Talayco wrote:
> It looks like you have the OVS veth connections and the OFTest veth
> connections on the same end.
>
>
The OFTest log shows:
>
>
>
> On Thursday, May 3, 2012 at 9:
rately,when running module, "address is in
> use" error coming.
>
> - Sonny
>
> On Thu, May 3, 2012 at 11:52 PM, Dan Talayco (mailto:dtala...@stanford.edu)> wrote:
> > But I believe I've seen some kind of discovery packets coming out of OVS,
> >
at 10:08:39AM -0700, Dan Talayco wrote:
> > OVS like many switches send out LLDPs periodically.
>
>
> No, OVS doesn't have any code for LLDP.
> ___
> openflow-discuss mailing list
> openflow-discuss@lists.stanford.edu
> (ma
Try using --relax as an option to oft.
OVS like many switches send out LLDPs periodically. --relax says "please
ignore other packets than the one I'm looking for."
-Dan
On Thursday, May 3, 2012 at 2:43 AM, sonny sonny wrote:
> >
> > Hi,
> >
> >
> >
> >
> >
> > I am running OF
Reposting to list from proper email addr.
Also, I assume you're run 'sudo make install' from oftest/tools/munger
Forwarded message:
> From: Dan Talayco at gmail
> To: Xuan-Nam Nguyen
> Cc: kalapriya k , Damien Saucez
> , openflow-discuss@lists.stanford.edu
> Date:
Hi--
I just ran OFTest against a 3290 in our lab running the test FillTableExact. I
got
RESULT: 3840 flows reported
a bit below the 4K, but about right. I'm not aware of any flow types that
might result in this kind of limit (2K that you're seeing).
I would suggest trying to run FillTab
The datapath ID is a unique id for the switch. It is normally based on the MAC
address but may be any 64-bit value.
The log level determines how much debug output is sent to the log files. It is
one of error, warn, verbose….
-Dan
On Wednesday, March 21, 2012 at 3:20 PM, Weiyang Mo wrote:
>
This will depend on what version of firmware you have installed on the switch;
I'm not sure the 3240 supports this, but it's worth a try.
-Dan
On Tuesday, March 20, 2012 at 9:53 AM, Michiel Appelman wrote:
> Hi,
>
> On 20 mrt. 2012, at 17:39, Rob Sherwood wrote:
>
> > How you seen the onlin
Are you setting ethertype to 0x800? If it's not an IP packet, the IP fields
may be ignored or forced to wildcards.
On Thursday, February 9, 2012 at 10:28 PM, Sabnis, Saurabh S wrote:
> Hi,
>
> I am using Beacon to set up flow in the switch through 'OFFlowMod' and I am
> giving just Source IP
And one more point: The test host needs to have connectivity to the
switch's interface on which it attempts to connect to a controller
(which in this case is the OFTest host).
-Dan
On Wednesday, February 8, 2012 at 1:46 PM, Rob Sherwood wrote:
> Hi,
>
> OFTest assumes that packets sent out
One more note:
In the 1.0 spec, the original queuing description provides one property which
is minimum bandwidth guarantee. It is specified from 1-1000 (tenths of a
percent). The addition of max-rate limiting for a queue has been added in many
implementations and I believe is under considera
It partly depends on how deep into the packet you want to look and what sort of
classifications you want to do.
You could first use OF as it stands to direct flows to some DPI agent.
You could also use some of the more recent extensible match mechanisms to
define matches further into the pac
See page 7 of the 1.0 spec.
1. If the action includes "set VLAN ID" then in both cases, the packet will
have a VLAN header on egress.
2. If the action includes "strip VLAN header" then for B, the packet will not
have a VLAN header on egress. For A, it is undefined and would depend on
whether th
Thanks Ben. That's a bit buried in the spec given that the value appears in
the text and the flag name, though defined in openflow.h, doesn't appear in the
spec.
-Dan
On Thursday, October 13, 2011 at 4:28 PM, Ben Pfaff wrote:
> On Thu, Oct 13, 2011 at 04:23:58PM -0700, Dan
Recently a question came up about the number of responses that may result from
a request. I don't recall seeing anything in the spec about this.
There are places where this might be optional (each port stats or flow stats
entry could be put in a different message). But there may be places where
" as the default action (with the obvious
trade-offs).
Hope this helps.
-Dan
On Tuesday, October 11, 2011 at 9:33 AM, Dan Talayco wrote:
> While I might get some pushback (hopefully I will) for saying this, I would
> argue that NORMAL is not that well defined; at least, it is hig
First, note that the --port option is to specify the TCP port on which the
controller should listen for connections from the switch. The usual OpenFlow
value is 6633.
Is Nox running at the same time as Oftest?
I suggest trying the option --test-spec=Echo to start with. This will just
attempt
Sorry, I think I just had an old version of OF checked out.
-Dan
On Monday, August 1, 2011 at 4:10 PM, Nicholas Bastin wrote:
> On Mon, Aug 1, 2011 at 12:59, Dan Talayco (mailto:dtala...@stanford.edu)> wrote:
> > while we're on the subject, I think a patch is needed
while we're on the subject, I think a patch is needed for running under Ubuntu
11.04 too?
On Friday, July 29, 2011 at 8:42 AM, Rob Sherwood wrote:
> Masahiko:
>
> Great work. Any chance you could apply your changes to the Makefile
> and commit them?
>
> [apologies if you're already familiar w
FYI: If this is based on the Stanford OpenFlow reference switch, note that
support for kernel space in 1.0 was...uh...neglected. udatapath/ was updated,
but datapath/ did not receive much attention with the 1.0 development.
Specifically, (if I recall correctly) the release criteria for the 1.0 s
There has been a lot of discussion on this over time. It depends a bit on the
nature of the non-OF switch. For example, if it uses LLDP packets or
participates in STP, you could at least recognize something about the link. You
could also have heuristics like "I see lots of MACs on this link" tho
On Thursday, April 14, 2011 at 5:04 PM, Kiran Yedavalli wrote:
> Thanks Rajiv and Dan. Looks like we need metadata in any case for this to
> work. We initially thought of this, but our platform does not support
> metadata, so was wondering if there is a way without using metadata. Looks
> like
A couple of quick notes. Hope they help.
It's possible the controller has inserted a flow entry into the flow table
before it has sent the packet back as a packet out. It's also possible the
controller has modified the packet before it sends it back as a packet out.
Finally, it's possible that
For all concerned, I've created indigo-17:
http://bugs.openflowhub.org/browse/INDIGO-17
I've received the patches from Quanta and Pica8 and will integrate them soon.
Thanks
Dan
On Monday, March 28, 2011 at 7:11 PM, Daniel E Talayco wrote:
>
> I'll be very happy to integrate it!
>
> -->)
>
>
Great question! Let us know what you decide. Those things are _loud_.
-Dan
On Monday, March 28, 2011 at 2:08 PM, Ali Sydney wrote:
> All,
> We are in the process of sourcing replacement fans for our Pronto 3290s (to
> reduce the noise level). Are there any particular specs for such fans?
> Cu
I checked my systems and both seem to respond to ping. I believe these are
using the images I downloaded from
http://www.openflow.org/wk/index.php/Pantou_:_OpenFlow_1.0_for_OpenWRT
Also just double checked that I'm able to ssh into the box.
I assume you're connected to the correct port, etc...
Announcing the availability of the indigo-1.0-web-rc3 release. This release
targets new users of the Pronto 3290. In particular, new users should be able
to upgrade to this release without connecting to the serial port. Instructions
are available here:
http://www.openflowhub.org/display/Indigo/
61 matches
Mail list logo
