linuxptp-users

[Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <mo...@us...>]

Hi,

I am new in Linux PTP project, I hope you can help me in using this
project.

Can anyone help me with the list of boundary clocks?I wonder whether
boundary clocks are simple switch with hardware time stamping? what special
characteristics it must have?

Than you.

BR,
*Mozhdeh *

[Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <kam...@gm...>]

Hi,

I am new in Linux PTP project. Can anyone help me with the list of boundary
clocks?

I wonder whether boundary clocks are simple switch with hardware time
stamping? what special characteristics it must have?

Than you.

BR,
*Mozhdeh *

Re: [Linuxptp-users] List of Boundary clocks

[Richard C. <ric...@gm...>]

On Thu, Mar 21, 2013 at 01:28:04PM +0100, Mozhdeh Kamel wrote:
> 
> Can anyone help me with the list of boundary clocks?I wonder whether
> boundary clocks are simple switch with hardware time stamping? what special
> characteristics it must have?

A boundary clock (BC) is more like a router than a switch. It must
intercept incoming PTP messages, and it does not forward them (except
for management messages).

Although linuxptp is designed to work as a BC, I don't know of any
that are using linuxptp. If you want to try the code out, then you
need a computer with two or more ports, all of which share a common
PTP Hardware Clock (PHC). I can think of two ways to do this.

1. Use a computer whose ports all support SW time stamping. This will
   not perform very well, but at least you can see how it works.

2. Get a computer with true HW time stamping support. Note that you
   cannot just use a box with multiple PCI cards, for example. The
   ports must all share *one* clock.

   Some ready made system on chips have this, like the Freescale P2020,
   the MPC8306, and the MPC8309. If you have one of these kits, then
   the ptp4l program should work fine.

   Another possibility is to use the DP83640 PHY, but you must make
   sure to connect the GPIOs in your design so that the PHYs can
   synchronize each other.

HTH,
Richard

Re: [Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <kam...@gm...>]

Can you please tell me the list of hardware which support both HW
timestamps and boundary clocks?

Mozhdeh


On Fri, Apr 12, 2013 at 11:46 PM, Keller, Jacob E
<jac...@in...>wrote:

>  LinuxPTP does support hardware timestamps. But some hardware doesn’t
> support boundary clocks and hardware timestamps together. Other hardware
> does.****
>
> ** **
>
> **-          **Jake****
>
> **
>

Re: [Linuxptp-users] List of Boundary clocks

[Richard C. <ric...@gm...>]

On Sun, Apr 14, 2013 at 02:35:11PM +0200, Mozhdeh Kamel wrote:
> Can you please tell me the list of hardware which support both HW
> timestamps and boundary clocks?

There only hardware that I know of is the ETSEC found on Freescale
PowerPC devices, like the P2020, the MPC85xx, and the MPC8313. These
are supported by the Linux "gianfar" driver.

I wish the list were longer.

Sorry,
Richard

Re: [Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <kam...@gm...>]

I could not find time stamping accuracy of Freescale P2020, do you know the
exact accuracy of it?

BR,
Mozhdeh


On Thu, Mar 21, 2013 at 7:58 PM, Richard Cochran
<ric...@gm...>wrote:

> On Thu, Mar 21, 2013 at 01:28:04PM +0100, Mozhdeh Kamel wrote:
> >
> > Can anyone help me with the list of boundary clocks?I wonder whether
> > boundary clocks are simple switch with hardware time stamping? what
> special
> > characteristics it must have?
>
> A boundary clock (BC) is more like a router than a switch. It must
> intercept incoming PTP messages, and it does not forward them (except
> for management messages).
>
> Although linuxptp is designed to work as a BC, I don't know of any
> that are using linuxptp. If you want to try the code out, then you
> need a computer with two or more ports, all of which share a common
> PTP Hardware Clock (PHC). I can think of two ways to do this.
>
> 1. Use a computer whose ports all support SW time stamping. This will
>    not perform very well, but at least you can see how it works.
>
> 2. Get a computer with true HW time stamping support. Note that you
>    cannot just use a box with multiple PCI cards, for example. The
>    ports must all share *one* clock.
>
>    Some ready made system on chips have this, like the Freescale P2020,
>    the MPC8306, and the MPC8309. If you have one of these kits, then
>    the ptp4l program should work fine.
>
>    Another possibility is to use the DP83640 PHY, but you must make
>    sure to connect the GPIOs in your design so that the PHYs can
>    synchronize each other.
>
> HTH,
> Richard
>



-- 
*Mozhdeh *

Re: [Linuxptp-users] List of Boundary clocks

[Keller, J. E <jac...@in...>]

> -----Original Message-----
> From: Richard Cochran [mailto:ric...@gm...]
> Sent: Thursday, March 21, 2013 11:58 AM
> To: Mozhdeh Kamel
> Cc: lin...@li...
> Subject: Re: [Linuxptp-users] List of Boundary clocks
> 
> 2. Get a computer with true HW time stamping support. Note that you
>    cannot just use a box with multiple PCI cards, for example. The
>    ports must all share *one* clock.
> 
> 
> HTH,
> Richard
> 

Also, just because the ports are on the same PCI card also does not mean they share a clock. You have to be sure the hardware supports a single clock for multiple ports.

- Jake

Re: [Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <mo...@us...>]

Thank you for the reply.

Actually I am going to design a network to run the PTP protocol(with BC).
But the main problem for me is I don't access to real hardware. I though
maybe using a PC with several network card (HW time stamping) would help.
But, share clock is another problem!!!

Now, I am thinking of possibility to use of virtual switch, router and NIC
.. Is anyone have experience of it?? Is it possible?

BR,
Mozhdeh


On Thu, Mar 21, 2013 at 9:33 PM, Keller, Jacob E
<jac...@in...>wrote:

> > -----Original Message-----
> > From: Richard Cochran [mailto:ric...@gm...]
> > Sent: Thursday, March 21, 2013 11:58 AM
> > To: Mozhdeh Kamel
> > Cc: lin...@li...
> > Subject: Re: [Linuxptp-users] List of Boundary clocks
> >
> > 2. Get a computer with true HW time stamping support. Note that you
> >    cannot just use a box with multiple PCI cards, for example. The
> >    ports must all share *one* clock.
> >
> >
> > HTH,
> > Richard
> >
>
> Also, just because the ports are on the same PCI card also does not mean
> they share a clock. You have to be sure the hardware supports a single
> clock for multiple ports.
>
> - Jake
>
>
>
> ------------------------------------------------------------------------------
> Everyone hates slow websites. So do we.
> Make your web apps faster with AppDynamics
> Download AppDynamics Lite for free today:
> http://p.sf.net/sfu/appdyn_d2d_mar
> _______________________________________________
> Linuxptp-users mailing list
> Lin...@li...
> https://lists.sourceforge.net/lists/listinfo/linuxptp-users
>

Re: [Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <kam...@gm...>]

I am totally getting confused regarding the hardware you introduced me. As
what I found, these are reference design board.. I confused how I can use
those.

I think you have read my scenario for testing IEEE1588 (bu using virtual
machines and NIC). Before that, I thought I can have a proper network
interface card that support hardware timestamps and use linuxptp  .. then I
can have boundary clock and ordinary clock and etc .. Is it wrong? But with
what you introduce.. I don't know how I can use it.


Thank you for you time. I am waiting to hear from you.

BR,


On Sun, Apr 14, 2013 at 3:01 PM, Richard Cochran
<ric...@gm...>wrote:

> On Sun, Apr 14, 2013 at 02:35:11PM +0200, Mozhdeh Kamel wrote:
> > Can you please tell me the list of hardware which support both HW
> > timestamps and boundary clocks?
>
> There only hardware that I know of is the ETSEC found on Freescale
> PowerPC devices, like the P2020, the MPC85xx, and the MPC8313. These
> are supported by the Linux "gianfar" driver.
>
> I wish the list were longer.
>
> Sorry,
> Richard
>

Re: [Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <kam...@gm...>]

Has anyone used Oregano NIC?
http://www.oreganosystems.at/?page_id=71


BR,
Mozhdeh


On Sun, Apr 14, 2013 at 3:01 PM, Richard Cochran
<ric...@gm...>wrote:

> On Sun, Apr 14, 2013 at 02:35:11PM +0200, Mozhdeh Kamel wrote:
> > Can you please tell me the list of hardware which support both HW
> > timestamps and boundary clocks?
>
> There only hardware that I know of is the ETSEC found on Freescale
> PowerPC devices, like the P2020, the MPC85xx, and the MPC8313. These
> are supported by the Linux "gianfar" driver.
>
> I wish the list were longer.
>
> Sorry,
> Richard
>



-- 
*Mozhdeh *

Re: [Linuxptp-users] List of Boundary clocks

[Richard C. <ric...@gm...>]

On Mon, Apr 15, 2013 at 09:06:31AM +0200, Mozhdeh Kamel wrote:
> Has anyone used Oregano NIC?
> http://www.oreganosystems.at/?page_id=71

I did have a chance to test an Oregan PCI card a few years back, and
it did seem to work. However, it looks like Oregano only supports
their own proprietary PTP stack.

HTH,
Richard

Re: [Linuxptp-users] List of Boundary clocks

[Richard C. <ric...@gm...>]

On Sun, Apr 14, 2013 at 04:33:36PM +0200, Mozhdeh Kamel wrote:
> I am totally getting confused regarding the hardware you introduced me. As
> what I found, these are reference design board.. I confused how I can use
> those.

Your question was...

   On Sun, Apr 14, 2013 at 02:35:11PM +0200, Mozhdeh Kamel wrote:
   > Can you please tell me the list of hardware which support both HW
   > timestamps and boundary clocks?

and the answer is that, as far as I know, only the Freescale boards
will work "out of the box" as a BC, with the Linux kernel and the
linuxptp stack. For example, the P2020RBD has three Ethernet ports.
It is a complete embedded computer, and so you could run ptp4l on the
three ports as a BC.
 
> I think you have read my scenario for testing IEEE1588 (bu using virtual
> machines and NIC). Before that, I thought I can have a proper network
> interface card that support hardware timestamps and use linuxptp  .. then I
> can have boundary clock and ordinary clock and etc .. Is it wrong? But with
> what you introduce.. I don't know how I can use it.

The important point is that a BC must have _two_or_more_ ports.
Just having one PCI card with HW time stamps is not enough. The BC
must have multiple ports all sharing the same clock.

If I understood your project idea, you are interested in the effects
of using PTP over BCs in a WAN, and you want to simulate different
kinds of network delay and jitter. In this case, I would recommend
just using a commercial BC. Getting you own home made BC working is a
whole project in itself.

HTH,
Richard

Re: [Linuxptp-users] List of Boundary clocks

[Richard C. <ric...@gm...>]

On Thu, Apr 04, 2013 at 01:04:49PM +0200, Mozhdeh Kamel wrote:
> 
> Actually I am going to design a network to run the PTP protocol(with BC).
> But the main problem for me is I don't access to real hardware. I though
> maybe using a PC with several network card (HW time stamping) would help.
> But, share clock is another problem!!!

With a small modification to ptp4l, you might be able to use multiple,
separate PCI cards together as a BC. We just need to disable* the check
in ptp4l that makes sure all the PHC clocks are the save. For example:

port1 eth0 /dev/ptp0 - ptp4l controls this clock, port connected to GM
port2 eth1 /dev/ptp1 - phc2sys instance 1 slaves this clock to /dev/ptp0
port3 eth2 /dev/ptp2 - phc2sys instance 2 slaves this clock to /dev/ptp0

In order to test how well phc2sys synchronizes the internal clocks,
you could run three separate ptp4l instances, one per port, and
connect all the ports to the same master through a switch. If you set
the ptp4l "free running" option on ports 2 and 3, then you can measure
the time error between the phc2sys servo and the PTP network time.

What PTP hardware do you have to choose from?
 
> Now, I am thinking of possibility to use of virtual switch, router and NIC
> .. Is anyone have experience of it?? Is it possible?

I am not sure what you mean by "virtual" here.

HTH,
Richard

* Or you could use a Linux kernel version before 3.5, without
  ETHTOOL_GET_TS_INFO support.

Re: [Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <kam...@gm...>]

I mean, I am thinking of having several virtual machines (such as VBox),
run the linux PTP project on each on them. and I can connect them together
through GRE or other tunnel. Use TC on linux or NeteM to add delay and
jitter .

BR,
Mozhdeh


On Sat, Apr 6, 2013 at 12:18 PM, Richard Cochran
<ric...@gm...>wrote:

> On Thu, Apr 04, 2013 at 01:04:49PM +0200, Mozhdeh Kamel wrote:
> >
> > Actually I am going to design a network to run the PTP protocol(with BC).
> > But the main problem for me is I don't access to real hardware. I though
> > maybe using a PC with several network card (HW time stamping) would help.
> > But, share clock is another problem!!!
>
> With a small modification to ptp4l, you might be able to use multiple,
> separate PCI cards together as a BC. We just need to disable* the check
> in ptp4l that makes sure all the PHC clocks are the save. For example:
>
> port1 eth0 /dev/ptp0 - ptp4l controls this clock, port connected to GM
> port2 eth1 /dev/ptp1 - phc2sys instance 1 slaves this clock to /dev/ptp0
> port3 eth2 /dev/ptp2 - phc2sys instance 2 slaves this clock to /dev/ptp0
>
> In order to test how well phc2sys synchronizes the internal clocks,
> you could run three separate ptp4l instances, one per port, and
> connect all the ports to the same master through a switch. If you set
> the ptp4l "free running" option on ports 2 and 3, then you can measure
> the time error between the phc2sys servo and the PTP network time.
>
> What PTP hardware do you have to choose from?
>
> > Now, I am thinking of possibility to use of virtual switch, router and
> NIC
> > .. Is anyone have experience of it?? Is it possible?
>
> I am not sure what you mean by "virtual" here.
>
> HTH,
> Richard
>
> * Or you could use a Linux kernel version before 3.5, without
>   ETHTOOL_GET_TS_INFO support.
>
>


-- 
*Mozhdeh *

Re: [Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <kam...@gm...>]

Thank you so much for your answers.

So, by having a real BC, and those NICS (such as Intel's 82576 chip or Intel
82574 ) with linuxptp can I emulate the network?
It is written in linuxptp readme that it implements the boundary clock, is
it means by having a real BC and connect it to the linux which has
linuxptp, it will recognize it and can be a slave for it?


Mozhdeh

On Mon, Apr 15, 2013 at 6:17 PM, Richard Cochran
<ric...@gm...>wrote:

> On Sun, Apr 14, 2013 at 04:33:36PM +0200, Mozhdeh Kamel wrote:
> > I am totally getting confused regarding the hardware you introduced me.
> As
> > what I found, these are reference design board.. I confused how I can use
> > those.
>
> Your question was...
>
>    On Sun, Apr 14, 2013 at 02:35:11PM +0200, Mozhdeh Kamel wrote:
>    > Can you please tell me the list of hardware which support both HW
>    > timestamps and boundary clocks?
>
> and the answer is that, as far as I know, only the Freescale boards
> will work "out of the box" as a BC, with the Linux kernel and the
> linuxptp stack. For example, the P2020RBD has three Ethernet ports.
> It is a complete embedded computer, and so you could run ptp4l on the
> three ports as a BC.
>
> > I think you have read my scenario for testing IEEE1588 (bu using virtual
> > machines and NIC). Before that, I thought I can have a proper network
> > interface card that support hardware timestamps and use linuxptp  ..
> then I
> > can have boundary clock and ordinary clock and etc .. Is it wrong? But
> with
> > what you introduce.. I don't know how I can use it.
>
> The important point is that a BC must have _two_or_more_ ports.
> Just having one PCI card with HW time stamps is not enough. The BC
> must have multiple ports all sharing the same clock.
>
> If I understood your project idea, you are interested in the effects
> of using PTP over BCs in a WAN, and you want to simulate different
> kinds of network delay and jitter. In this case, I would recommend
> just using a commercial BC. Getting you own home made BC working is a
> whole project in itself.
>
> HTH,
> Richard
>

Re: [Linuxptp-users] List of Boundary clocks

[Richard C. <ric...@gm...>]

On Thu, Apr 18, 2013 at 07:00:50PM +0200, Mozhdeh Kamel wrote:
> Thank you so much for your answers.
> 
> So, by having a real BC, and those NICS (such as Intel's 82576 chip or Intel
> 82574 ) with linuxptp can I emulate the network?

I am not sure how well you can "emulate the network", but a commercial
BC plus Ordinary Clocks using a PHC running linuxptp should work fine.

> It is written in linuxptp readme that it implements the boundary clock, is
> it means by having a real BC and connect it to the linux which has
> linuxptp, it will recognize it and can be a slave for it?

No, it means that linuxptp itself can be a boundary clock. However, as
I described before, you need a special hardware setup in order to
implement a BC.

HTH,
Richard

Re: [Linuxptp-users] List of Boundary clocks

[Richard C. <ric...@gm...>]

On Thu, Apr 18, 2013 at 07:00:50PM +0200, Mozhdeh Kamel wrote:
> Thank you so much for your answers.
> 
> So, by having a real BC, and those NICS (such as Intel's 82576 chip or Intel
> 82574 ) with linuxptp can I emulate the network?

BTW, if you are going with an Intel card, I can highly recommend the
i210 card. It is inexpensive and has some nice features, including a 6
pin header for the PPS signals.

Thanks,
Richard

Re: [Linuxptp-users] List of Boundary clocks

[Keller, J. E <jac...@in...>]

> -----Original Message-----
> From: Richard Cochran [mailto:ric...@gm...]
> Sent: Friday, April 19, 2013 8:43 AM
> To: Mozhdeh Kamel
> Cc: lin...@li...
> Subject: Re: [Linuxptp-users] List of Boundary clocks
> 
> On Thu, Apr 18, 2013 at 07:00:50PM +0200, Mozhdeh Kamel wrote:
> > Thank you so much for your answers.
> >
> > So, by having a real BC, and those NICS (such as Intel's 82576 chip or
> Intel
> > 82574 ) with linuxptp can I emulate the network?
> 
> BTW, if you are going with an Intel card, I can highly recommend the
> i210 card. It is inexpensive and has some nice features, including a 6
> pin header for the PPS signals.
> 
> Thanks,
> Richard
> 

This is the same device I would recommend as well.

- Jake

Re: [Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <kam...@gm...>]

Thank you so much!

I am not insisting on one brand, so if you know anything rather than Intel
that has a fair price, don't hesitate to introduce it to me.

BR,
Mozhdeh


On Fri, Apr 19, 2013 at 10:14 PM, Keller, Jacob E
<jac...@in...>wrote:

> > -----Original Message-----
> > From: Richard Cochran [mailto:ric...@gm...]
> > Sent: Friday, April 19, 2013 8:43 AM
> > To: Mozhdeh Kamel
> > Cc: lin...@li...
> > Subject: Re: [Linuxptp-users] List of Boundary clocks
> >
> > On Thu, Apr 18, 2013 at 07:00:50PM +0200, Mozhdeh Kamel wrote:
> > > Thank you so much for your answers.
> > >
> > > So, by having a real BC, and those NICS (such as Intel's 82576 chip or
> > Intel
> > > 82574 ) with linuxptp can I emulate the network?
> >
> > BTW, if you are going with an Intel card, I can highly recommend the
> > i210 card. It is inexpensive and has some nice features, including a 6
> > pin header for the PPS signals.
> >
> > Thanks,
> > Richard
> >
>
> This is the same device I would recommend as well.
>
> - Jake
>
>


-- 
*Mozhdeh *

Re: [Linuxptp-users] List of Boundary clocks

[Richard C. <ric...@gm...>]

On Mon, Apr 08, 2013 at 04:17:46PM +0200, Mozhdeh Kamel wrote:
> I mean, I am thinking of having several virtual machines (such as VBox),
> run the linux PTP project on each on them. and I can connect them together
> through GRE or other tunnel. Use TC on linux or NeteM to add delay and
> jitter .

But without hardware time stamping, you are not going to get very good
synchronization performance.

Here is another simulation idea you could explore. It would be very
interesting to be able to test the BMC in a simulated network, and to
test the clock servos as well. Take a look at Miroslav Lichvar's clock
and network simulator to get an idea of what can be done.

   http://mlichvar.fedorapeople.org/clknetsim

It would be a fair amount of programming, but you could then simulate
the hardware timestamping as exactly as you wish.

HTH,
Richard

Re: [Linuxptp-users] List of Boundary clocks

[Miroslav L. <mli...@re...>]

On Mon, Apr 08, 2013 at 08:40:44PM +0200, Richard Cochran wrote:
> On Mon, Apr 08, 2013 at 04:17:46PM +0200, Mozhdeh Kamel wrote:
> > I mean, I am thinking of having several virtual machines (such as VBox),
> > run the linux PTP project on each on them. and I can connect them together
> > through GRE or other tunnel. Use TC on linux or NeteM to add delay and
> > jitter .
> 
> But without hardware time stamping, you are not going to get very good
> synchronization performance.

Depends on what exactly she wants to test. When I don't care about
accuracy and want to test the protocol itself, I use sometimes ptp4l
in qemu with the e1000 driver. If the delay and jitter was shifted by
tc to 10s or 100s of milliseconds, maybe it could be of some value for
performance testing too.

> Here is another simulation idea you could explore. It would be very
> interesting to be able to test the BMC in a simulated network, and to
> test the clock servos as well. Take a look at Miroslav Lichvar's clock
> and network simulator to get an idea of what can be done.
> 
>    http://mlichvar.fedorapeople.org/clknetsim
> 
> It would be a fair amount of programming, but you could then simulate
> the hardware timestamping as exactly as you wish.

I'd very much like to see ptp4l supported by clknetsim. The list of
things which I know are missing for the support includes:
- multiple timers
- multiple interfaces
- multicast
- some ioctls (hw time stamping)

Thanks,

-- 
Miroslav Lichvar

Re: [Linuxptp-users] List of Boundary clocks

[Mozhdeh K. <kam...@gm...>]

The goal of my test is to study the probability of extending IEEE1588 to
larger networks. So I focus on delay between the master (GM) and slave in
different part of networks.
I chose ptplinux because it support boundary clocks(use of BC in
unavoidable in larger network) although it does not support hw timestamp,
the ptpd support hw time stamp but nor boundary clock. I decide to do
emulation and buy the necessary hardware. So,
I thought of running linuxptp on different virtual machines, connect these
VM with GRE tunnel together, connect the VM which is suppose to be boundary
clock to special  PCI that has hw timestamping. Add jitter by TC command in
linux,  and do the test. So, according to that idea, I should buy a PC with
multiple slot for PCI, buying PCI that support hw time stamp (for boundary
clocks with several port).
What do you think of this idea? Do you have any suggestion for the hardware
I should buy?

BR,
Mozhdeh
On Tue, Apr 9, 2013 at 2:34 PM, Miroslav Lichvar <mli...@re...>wrote:

> On Mon, Apr 08, 2013 at 08:40:44PM +0200, Richard Cochran wrote:
> > On Mon, Apr 08, 2013 at 04:17:46PM +0200, Mozhdeh Kamel wrote:
> > > I mean, I am thinking of having several virtual machines (such as
> VBox),
> > > run the linux PTP project on each on them. and I can connect them
> together
> > > through GRE or other tunnel. Use TC on linux or NeteM to add delay and
> > > jitter .
> >
> > But without hardware time stamping, you are not going to get very good
> > synchronization performance.
>
> Depends on what exactly she wants to test. When I don't care about
> accuracy and want to test the protocol itself, I use sometimes ptp4l
> in qemu with the e1000 driver. If the delay and jitter was shifted by
> tc to 10s or 100s of milliseconds, maybe it could be of some value for
> performance testing too.
>
> > Here is another simulation idea you could explore. It would be very
> > interesting to be able to test the BMC in a simulated network, and to
> > test the clock servos as well. Take a look at Miroslav Lichvar's clock
> > and network simulator to get an idea of what can be done.
> >
> >    http://mlichvar.fedorapeople.org/clknetsim
> >
> > It would be a fair amount of programming, but you could then simulate
> > the hardware timestamping as exactly as you wish.
>
> I'd very much like to see ptp4l supported by clknetsim. The list of
> things which I know are missing for the support includes:
> - multiple timers
> - multiple interfaces
> - multicast
> - some ioctls (hw time stamping)
>
> Thanks,
>
> --
> Miroslav Lichvar
>



-- 
*Mozhdeh *

Re: [Linuxptp-users] List of Boundary clocks

[Keller, J. E <jac...@in...>]

LinuxPTP does support hardware timestamps. But some hardware doesn't support boundary clocks and hardware timestamps together. Other hardware does.


-          Jake

From: Mozhdeh Kamel [mailto:kam...@gm...]
Sent: Friday, April 12, 2013 5:07 AM
To: Miroslav Lichvar
Cc: lin...@li...
Subject: Re: [Linuxptp-users] List of Boundary clocks

The goal of my test is to study the probability of extending IEEE1588 to larger networks. So I focus on delay between the master (GM) and slave in different part of networks.
I chose ptplinux because it support boundary clocks(use of BC in unavoidable in larger network) although it does not support hw timestamp, the ptpd support hw time stamp but nor boundary clock. I decide to do emulation and buy the necessary hardware. So,
I thought of running linuxptp on different virtual machines, connect these VM with GRE tunnel together, connect the VM which is suppose to be boundary clock to special  PCI that has hw timestamping. Add jitter by TC command in linux,  and do the test. So, according to that idea, I should buy a PC with multiple slot for PCI, buying PCI that support hw time stamp (for boundary clocks with several port).
What do you think of this idea? Do you have any suggestion for the hardware I should buy?

BR,
Mozhdeh
On Tue, Apr 9, 2013 at 2:34 PM, Miroslav Lichvar <mli...@re...<mailto:mli...@re...>> wrote:
On Mon, Apr 08, 2013 at 08:40:44PM +0200, Richard Cochran wrote:
> On Mon, Apr 08, 2013 at 04:17:46PM +0200, Mozhdeh Kamel wrote:
> > I mean, I am thinking of having several virtual machines (such as VBox),
> > run the linux PTP project on each on them. and I can connect them together
> > through GRE or other tunnel. Use TC on linux or NeteM to add delay and
> > jitter .
>
> But without hardware time stamping, you are not going to get very good
> synchronization performance.
Depends on what exactly she wants to test. When I don't care about
accuracy and want to test the protocol itself, I use sometimes ptp4l
in qemu with the e1000 driver. If the delay and jitter was shifted by
tc to 10s or 100s of milliseconds, maybe it could be of some value for
performance testing too.

> Here is another simulation idea you could explore. It would be very
> interesting to be able to test the BMC in a simulated network, and to
> test the clock servos as well. Take a look at Miroslav Lichvar's clock
> and network simulator to get an idea of what can be done.
>
>    http://mlichvar.fedorapeople.org/clknetsim
>
> It would be a fair amount of programming, but you could then simulate
> the hardware timestamping as exactly as you wish.
I'd very much like to see ptp4l supported by clknetsim. The list of
things which I know are missing for the support includes:
- multiple timers
- multiple interfaces
- multicast
- some ioctls (hw time stamping)

Thanks,

--
Miroslav Lichvar



--
Mozhdeh