settingsLogin | Registersettings

[Openstack] Rackspace abandons Open vSwitch ?

0 votes

Hi All,

As per this news article listed below, Rackspace is abandoning Open
vSwitch. Is this where everyone else going in general ? An excerpt from
the article

"With the new release Rackspace also decided to get away from Open vSwitch,
the open source software defined network platform used as a plug-in for
Neutron (the network element of OpenStack), conceding it was not quite
ready for production and high-volume workloads."

http://www.datacenterknowledge.com/archives/2014/09/25/rackspace-re-architects-its-openstack-private-clouds/

Regards,
Raghu.


Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
asked Sep 27, 2014 in openstack by Raghu_Vadapalli (400 points)   1 2
retagged Feb 25, 2015 by admin

20 Responses

0 votes

On Sat, Sep 27, 2014 at 3:50 AM, Raghu Vadapalli rvatspacket@gmail.com wrote:
As per this news article listed below, Rackspace is abandoning Open vSwitch.
Is this where everyone else going in general ?

That conclusion is inaccurate. The entirety of the public cloud runs
openvswitch for both public/servicenet connectivity as well as
isolated tenant network features.The article is referring to the
private cloud distribution no longer choosing to use the Neutron
OpenVswitch plugin
(https://github.com/openstack/neutron/tree/master/neutron/plugins/openvswitch)
as it is being deprecated. The ML2 plugin replaces this and can use a
variety of mechanisms including openvswitch.

The article's conclusion that openvswitch is not ready for production
and high-volume workloads is ludicrous. Versions 2.0+ perform very
well with multithreading in the vswitchd process and megaflows in the
datapath. However it is important to point out that datapath
performance is very much related to the flows programmed. A poorly
written flow set will result in bad performance. Tuning the flows and
optimizing the ability for megaflow'ing is the key to high throughput.

Happy Hacking!

7-11

responded Sep 27, 2014 by =?UTF-8?Q?Jason_K=C3 (260 points)   1
0 votes

On 27.09.2014 06:37, Jason Kölker wrote:
On Sat, Sep 27, 2014 at 3:50 AM, Raghu Vadapalli rvatspacket@gmail.com wrote:

As per this news article listed below, Rackspace is abandoning Open vSwitch.
Is this where everyone else going in general ?

That conclusion is inaccurate. The entirety of the public cloud runs
openvswitch for both public/servicenet connectivity as well as
isolated tenant network features.The article is referring to the
private cloud distribution no longer choosing to use the Neutron
OpenVswitch plugin
(https://github.com/openstack/neutron/tree/master/neutron/plugins/openvswitch)
as it is being deprecated. The ML2 plugin replaces this and can use a
variety of mechanisms including openvswitch.

The article's conclusion that openvswitch is not ready for production
and high-volume workloads is ludicrous. Versions 2.0+ perform very
well with multithreading in the vswitchd process and megaflows in the
datapath. However it is important to point out that datapath
performance is very much related to the flows programmed. A poorly
written flow set will result in bad performance. Tuning the flows and
optimizing the ability for megaflow'ing is the key to high throughput.

That's the theory though and the article seems to talk about practical
problem Rackspace ran into with OVS so it would have been nice to learn
what specifically was the problem.

What are the alternatives though? As far as I know the regular linux
bridge lacks most of the features of OVS and these are the only to
options I've played with so far. Is the a third alternative out there
that they've switched to?

Regards,
Dennis

responded Sep 27, 2014 by Dennis_Jacobfeuerbor (140 points)   1
0 votes

-----Original Message-----
From: Dennis Jacobfeuerborn [mailto:dennisml@conversis.de]
Sent: 27 September 2014 13:26
To: openstack@lists.openstack.org
Subject: Re: [Openstack] Rackspace abandons Open vSwitch ?

On 27.09.2014 06:37, Jason Kölker wrote:

On Sat, Sep 27, 2014 at 3:50 AM, Raghu Vadapalli rvatspacket@gmail.com
wrote:

As per this news article listed below, Rackspace is abandoning Open vSwitch.
Is this where everyone else going in general ?

That conclusion is inaccurate. The entirety of the public cloud runs
openvswitch for both public/servicenet connectivity as well as
isolated tenant network features.The article is referring to the
private cloud distribution no longer choosing to use the Neutron
OpenVswitch plugin
(https://github.com/openstack/neutron/tree/master/neutron/plugins/open
vswitch) as it is being deprecated. The ML2 plugin replaces this and
can use a variety of mechanisms including openvswitch.

The article's conclusion that openvswitch is not ready for production
and high-volume workloads is ludicrous. Versions 2.0+ perform very
well with multithreading in the vswitchd process and megaflows in the
datapath. However it is important to point out that datapath
performance is very much related to the flows programmed. A poorly
written flow set will result in bad performance. Tuning the flows and
optimizing the ability for megaflow'ing is the key to high throughput.

That's the theory though and the article seems to talk about practical problem
Rackspace ran into with OVS so it would have been nice to learn what
specifically was the problem.

What are the alternatives though? As far as I know the regular linux bridge lacks
most of the features of OVS and these are the only to options I've played with so
far. Is the a third alternative out there that they've switched to?

I also am not clear on what the best option is for a scalable open source Neutron plug in. The surveys are regularly reporting open vswitch is the most commonly used but it is not clear if that is the best one for production at scale. Are there any references of real life usage of OVS at the 1000s of hypervisors level ?

Tim

responded Sep 27, 2014 by Tim_Bell (16,440 points)   1 5 10
0 votes

I'm not the best to answer the specific technical problems we had, I'm sure others from Rackspace private cloud will chime in on that. At a high level though we did have customers see issues with high numbers of flows, including the 64k limit, where it would simply fall over. We were also having trouble in general with kernel/ ovs mismatches, regressions between OVS versions, and kernel panics.

We moved to Linuxbridge/ML2 and its corresponding neutron plugin and it appears to be much more stable so far. From a feature perspective we haven't seen a huge drop in terms of what we actually use.

Jason

On Sep 27, 2014, at 11:41, Tim Bell Tim.Bell@cern.ch wrote:

-----Original Message-----
From: Dennis Jacobfeuerborn [mailto:dennisml@conversis.de]
Sent: 27 September 2014 13:26
To: openstack@lists.openstack.org
Subject: Re: [Openstack] Rackspace abandons Open vSwitch ?

On 27.09.2014 06:37, Jason Kölker wrote:
On Sat, Sep 27, 2014 at 3:50 AM, Raghu Vadapalli rvatspacket@gmail.com
wrote:

As per this news article listed below, Rackspace is abandoning Open vSwitch.
Is this where everyone else going in general ?

That conclusion is inaccurate. The entirety of the public cloud runs
openvswitch for both public/servicenet connectivity as well as
isolated tenant network features.The article is referring to the
private cloud distribution no longer choosing to use the Neutron
OpenVswitch plugin
(https://github.com/openstack/neutron/tree/master/neutron/plugins/open
vswitch) as it is being deprecated. The ML2 plugin replaces this and
can use a variety of mechanisms including openvswitch.

The article's conclusion that openvswitch is not ready for production
and high-volume workloads is ludicrous. Versions 2.0+ perform very
well with multithreading in the vswitchd process and megaflows in the
datapath. However it is important to point out that datapath
performance is very much related to the flows programmed. A poorly
written flow set will result in bad performance. Tuning the flows and
optimizing the ability for megaflow'ing is the key to high throughput.

That's the theory though and the article seems to talk about practical problem
Rackspace ran into with OVS so it would have been nice to learn what
specifically was the problem.

What are the alternatives though? As far as I know the regular linux bridge lacks
most of the features of OVS and these are the only to options I've played with so
far. Is the a third alternative out there that they've switched to?

I also am not clear on what the best option is for a scalable open source Neutron plug in. The surveys are regularly reporting open vswitch is the most commonly used but it is not clear if that is the best one for production at scale. Are there any references of real life usage of OVS at the 1000s of hypervisors level ?

Tim

Regards,
Dennis


Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack


Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack


Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
responded Sep 27, 2014 by Jason_Cannavale (140 points)   1
0 votes

On Sat, Sep 27, 2014 at 4:29 PM, Tim Bell Tim.Bell@cern.ch wrote:
I also am not clear on what the best option is for a scalable open source Neutron plug in. The surveys are regularly reporting open vswitch is the most commonly used but it is not clear if that is the best one for production at scale. Are there any references of real life usage of OVS at the 1000s of hypervisors level ?

In public cloud we run our own opensource Neutron plugin
(https://github.com/rackerlabs/quark) which is conceptually similar to
ML2. The only real reason we don't run ML2 is that we proposed the
Quark model of minimizing the L2 operations into tiny backend drivers
and it was not well received. We went on developing Quark in the open,
and a month or two later ML2 came along, but we weren't in a position
to change direction. If we were starting over today I think ML2 and
Ryu's ofagent would be the direction I would like to go for 1000's of
hypervisors.

I think much of the confusion as to "OVS" performance is due to the
Neutron plugin sharing the same name as the data plane system. With
OVS (not the Neutron plugin) we have 10,000's of hypervisors running
it as their dataplane. Since the majority of our vifs are just bridged
to a publicnet or servicenet fabric, we use the Quark Neutron plugin
to model the VM's connections, and then udev rules on each hypervisor
to actually program static flows per vif into the OVS dataplane.

We haven't opensourced the udev scripts, as they are public cloud
specific and probably not generally applicable (business logic
unfortunately infected them ;( ); however, I've got it in the back of
my mind to reimplement them using Ryu as the controller framework and
I've slowly been adding in remote ovsdb replication and two draft
rfc's (https://tools.ietf.org/html/draft-ietf-l2vpn-evpn and
https://tools.ietf.org/html/draft-sd-l2vpn-evpn-overlay) in my github
fork of Ryu to that end with the eventual goal of getting them ready
to be submitted to upstream Ryu. That will allow us to decouple the
flushing of flows to the dataplane from the flow generation and
opensource it for anyone to use.

TL;DR I don't think there is an opensource scalable (depending on your
definition of scale ;) ) drop and go Neutron ovs dataplane yet. The
private cloud choice of ML2 and LinuxBridge mech is probably the best
option as it currently stands.

Happy Hacking!

7-11


Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
responded Sep 27, 2014 by =?UTF-8?Q?Jason_K=C3 (260 points)   1
0 votes

On 27.09.2014 06:37, Jason Kölker wrote:

On Sat, Sep 27, 2014 at 3:50 AM, Raghu Vadapalli rvatspac...@gmail.com
wrote:

As per this news article listed below, Rackspace is abandoning Open vSwitch.
Is this where everyone else going in general ?

That conclusion is inaccurate. The entirety of the public cloud runs
openvswitch for both public/servicenet connectivity as well as
isolated tenant network features.The article is referring to the
private cloud distribution no longer choosing to use the Neutron
OpenVswitch plugin
(https://github.com/openstack/neutron/tree/master/neutron/plugins/openvswitch)
as it is being deprecated. The ML2 plugin replaces this and can use a
variety of mechanisms including openvswitch.

The article's conclusion that openvswitch is not ready for production
and high-volume workloads is ludicrous. Versions 2.0+ perform very
well with multithreading in the vswitchd process and megaflows in the
datapath. However it is important to point out that datapath
performance is very much related to the flows programmed. A poorly
written flow set will result in bad performance. Tuning the flows and
optimizing the ability for megaflow'ing is the key to high throughput.

That's the theory though and the article seems to talk about practical
problem Rackspace ran into with OVS so it would have been nice to learn
what specifically was the problem.

What are the alternatives though? As far as I know the regular linux
bridge lacks most of the features of OVS and these are the only to
options I've played with so far. Is the a third alternative out there
that they've switched to?

One alternative is OpenContrail vRouter as ML3 plugin. It meets the scale and feature requirements.

Maria

responded Sep 29, 2014 by NAPIERALA,_MARIA_H (260 points)   2 3
0 votes


Raghu V. Vadapalli
Principal Software Architect
Smart Packet Solutions

On Monday, Sep 29, 2014 at 1:33 PM, NAPIERALAMARIA H mn1921@att.com, wrote:

On 27.09.2014 06:37, Jason Kölker wrote:

On Sat, Sep 27, 2014 at 3:50 AM, Raghu Vadapalli rvatspac...@gmail.com

wrote:

As per this news article listed below, Rackspace is abandoning Open vSwitch.

Is this where everyone else going in general ?

That conclusion is inaccurate. The entirety of the public cloud runs

openvswitch for both public/servicenet connectivity as well as

isolated tenant network features.The article is referring to the

private cloud distribution no longer choosing to use the Neutron

OpenVswitch plugin

(https://github.com/openstack/neutron/tree/master/neutron/plugins/openvswitch)

as it is being deprecated. The ML2 plugin replaces this and can use a

variety of mechanisms including openvswitch.

The article's conclusion that openvswitch is not ready for production

and high-volume workloads is ludicrous. Versions 2.0+ perform very

well with multithreading in the vswitchd process and megaflows in the

datapath. However it is important to point out that datapath

performance is very much related to the flows programmed. A poorly

written flow set will result in bad performance. Tuning the flows and

optimizing the ability for megaflow'ing is the key to high throughput.

That's the theory though and the article seems to talk about practical

problem Rackspace ran into with OVS so it would have been nice to learn

what specifically was the problem.

What are the alternatives though? As far as I know the regular linux

bridge lacks most of the features of OVS and these are the only to

options I've played with so far. Is the a third alternative out there

that they've switched to?

One alternative is OpenContrail vRouter as ML3 plugin. It meets the scale and feature requirements.

Maria

​did you mean to say contrail with ML2 not ML3 ? I am not quite sure contrail integration but is ml3 available in icehouse yet ?_______________________________________________
Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack

responded Sep 29, 2014 by Raghu_Vadapalli (400 points)   1 2
0 votes

did you mean to say contrail with ML2 not ML3 ?

Should be ML3 since it provides layer 3 networking.

I am not quite sure contrail integration but is ml3 available in icehouse yet ?

It is not.

From: Raghu Vadapalli [mailto:rvatspacket@gmail.com]
Sent: Monday, September 29, 2014 1:38 PM
To: NAPIERALA, MARIA H
Cc: dennisml@conversis.de; openstack@lists.openstack.org
Subject: Re: [Openstack] Rackspace abandons Open vSwitch ?


Raghu V. Vadapalli
Principal Software Architect
Smart Packet Solutions

On Monday, Sep 29, 2014 at 1:33 PM, NAPIERALAMARIA H mn1921@att.com, wrote:

On 27.09.2014 06:37, Jason Kölker wrote:

On Sat, Sep 27, 2014 at 3:50 AM, Raghu Vadapalli rvatspac...@gmail.com
wrote:

As per this news article listed below, Rackspace is abandoning Open vSwitch.
Is this where everyone else going in general ?

That conclusion is inaccurate. The entirety of the public cloud runs
openvswitch for both public/servicenet connectivity as well as
isolated tenant network features.The article is referring to the
private cloud distribution no longer choosing to use the Neutron
OpenVswitch plugin
(https://github.com/openstack/neutron/tree/master/neutron/plugins/openvswitch)
as it is being deprecated. The ML2 plugin replaces this and can use a
variety of mechanisms including openvswitch.

The article's conclusion that openvswitch is not ready for production
and high-volume workloads is ludicrous. Versions 2.0+ perform very
well with multithreading in the vswitchd process and megaflows in the
datapath. However it is important to point out that datapath
performance is very much related to the flows programmed. A poorly
written flow set will result in bad performance. Tuning the flows and
optimizing the ability for megaflow'ing is the key to high throughput.

That's the theory though and the article seems to talk about practical
problem Rackspace ran into with OVS so it would have been nice to learn
what specifically was the problem.

What are the alternatives though? As far as I know the regular linux
bridge lacks most of the features of OVS and these are the only to
options I've played with so far. Is the a third alternative out there
that they've switched to?

One alternative is OpenContrail vRouter as ML3 plugin. It meets the scale and feature requirements.

Maria

​did you mean to say contrail with ML2 not ML3 ? I am not quite sure contrail integration but is ml3 available in icehouse yet ?


Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
responded Sep 29, 2014 by NAPIERALA,_MARIA_H (260 points)   2 3
0 votes

Are there any references for people running OpenContrail at scale ?

From: NAPIERALA, MARIA H [mailto:mn1921@att.com]
Sent: 29 September 2014 19:26
To: dennisml@conversis.de
Cc: openstack@lists.openstack.org
Subject: Re: [Openstack] Rackspace abandons Open vSwitch ?

......

What are the alternatives though? As far as I know the regular linux
bridge lacks most of the features of OVS and these are the only to
options I've played with so far. Is the a third alternative out there
that they've switched to?

One alternative is OpenContrail vRouter as ML3 plugin. It meets the scale and feature requirements.

Maria


Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
responded Sep 29, 2014 by Tim_Bell (16,440 points)   1 5 10
0 votes

On 09/29/2014 01:52 PM, Tim Bell wrote:

Are there any references for people running OpenContrail at scale ?

Though reference are good to have, in general, L3 networks are known to
scale better than L2 networks.
Having said that, the complexity of two large frameworks OpenStack +
OpenContrail working together nicely in
deployment is not known to me. Any ideas ?

From:NAPIERALA, MARIA H [mailto:mn1921@att.com]
Sent: 29 September 2014 19:26
To: dennisml@conversis.de
Cc: openstack@lists.openstack.org
Subject: Re: [Openstack] Rackspace abandons Open vSwitch ?

……

What are the alternatives though? As far as I know the regular linux

bridge lacks most of the features of OVS and these are the only to

options I've played with so far. Is the a third alternative out there

that they've switched to?

One alternative is OpenContrail vRouter as ML3 plugin. It meets the
scale and feature requirements.

Maria


Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack


Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
Post to : openstack@lists.openstack.org
Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
responded Sep 29, 2014 by Raghu_Vadapalli (400 points)   1 2
...