Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: [EFM]Event Time Stamp




Hi David,

You have showed clearly that the definition of the event time stamp is
consistent. Thanks!

The original motivation of my question was hoping to widen the event time
stamp to 32-bit. As others explained, this is intended only to provide
relative timing between events (well, events that are spaced no more than
65535/10/60=109 minutes apart anyway). I think a 32-bit time stamp would be
better, but I guess supplemented with local time stamping, 16-bit would be
fine too.

Thanks again.

Regards,
Yonghong Ren
Appian Communications


-----Original Message-----
From: David Law [mailto:David_Law@eur.3com.com]
Sent: Thursday, August 14, 2003 6:28 PM
To: Yonghong Ren
Cc: stds-802-3-efm@ieee.org
Subject: RE: [EFM]Event Time Stamp




Hi Yonghong,

Clause 30 is a protocol independent MIB definition which is intended to
provide
behaviors that can be referenced by protocol dependent MIBs. We provide a
GDMO
specification for IEEE 802.3 managed object classes in Annex 30A and I
understand work is ongoing within the IETF to provide the SNMP specification
as
well. You can find a presentation at the URL -[
http://www.ieee802.org/3/efm/public/sep01/law_1_0901.pdf ] which discusses
this
structure further.

In respect to your actual question, if you examine the
aOAMRemoteErrSymPeriodEvent attribute contained in the Annex 30 changes
(Page
145) you will find the syntax for the attribute defined as
IEEE802Dot3-MgmtAttributeModule.OAMErrorEvent. If you then examine the ASN.1
IEEE802Dot3-MgmtAttributeModule module additions defined in Annex 30B (Page
151)
you will find OAMErrorEvent defined as follows:

OAMErrorEvent ::= SEQUENCE {
                            OAMEventTimeStamp [1] Integer16,
                            OAMEventWindow    [2] Integer64,
                            OAMEventThreshold [3] Integer64,
                            OAMEventRunTotal  [4] Integer32
                            }

Hence as can be seen the OAMEventTimeStamp is indeed defined as a 16 bit
Integer
in Annex 30A and I hope the other Integers are defined as you would expect.

Best regards,
  David Law







Yonghong Ren <ren@appiancom.com> on 14/08/2003 22:33:05

Sent by:  Yonghong Ren <ren@appiancom.com>


To:   "'Kevin Daines'" <Kevin.Daines@worldwidepackets.com>, Yonghong Ren
      <ren@appiancom.com>, Matt Squire <MSquire@HatterasNetworks.com>,
      stds-802-3-efm@ieee.org
cc:    (David Law/GB/3Com)
Subject:  RE: [EFM]Event Time Stamp





Kevin,

As I noted in my original email, there seem to be a disagreement between
Clause 57 with 30.11.1.1.37-40.

For example, 57.5.3.1 explicitly states that the time stamp is 16-bit:

"Event Time Stamp. This two-octet field indicates the time reference when
the event was generated, in terms of 100ms intervals, encoded as a 16-bit
unsigned integer. When received from the remote DTE and if Clause 30 is
present, this maps to 30.11.1.1.37."

Then section 30.11.1.1.37 (which defines aOamRemoteErrSymPeriodEvent) has
this text:

"A SEQUENCE of four instances of the type INTEGER... The first INTEGER
represents the Event Time Stamp field..."

I would appreciate it if you would explain this further. Also, could you
refer me to the right source for this INTEGER encoding with field width you
mentioned?

Thank you very much.

Regards,
Yonghong Ren
Appian Communications.

-----Original Message-----
From: Kevin Daines [mailto:Kevin.Daines@worldwidepackets.com]
Sent: Thursday, August 14, 2003 4:48 PM
To: Yonghong Ren; Matt Squire; stds-802-3-efm@ieee.org
Subject: RE: [EFM]Event Time Stamp


The attributes found in 30.11 contain integers of various lengths, from
two-octets up to eight-octets. The referenced attributes found in
30.11.1.1.37-40 define integers corresponding to fields defined in Clause
57. The integers are sized according to the field widths.

Kevin Daines
Editor, EFM OAM


-----Original Message-----
From: Yonghong Ren [mailto:ren@appiancom.com]
Sent: Wednesday, August 06, 2003 11:53 AM
To: 'Matt Squire'; Yonghong Ren; stds-802-3-efm@ieee.org
Subject: RE: [EFM]Event Time Stamp



Thanks for the explanation!

So this time stamp gives the more or less precise timing of events that are
clumped together. If events are spaced more than 109 (65535/10/60) minutes
apart, they became meaningless. One would imagine this "micro" view needs to
be supplemented by the "macro" view of local time stamping as events are
received.

Perhaps a few sentences to explain this in the standard would be useful?

Thanks again.

-- Yonghong Ren


-----Original Message-----
From: Matt Squire [mailto:MSquire@HatterasNetworks.com]
Sent: Tuesday, August 05, 2003 10:07 PM
To: Yonghong Ren; stds-802-3-efm@ieee.org
Subject: Re: [EFM]Event Time Stamp



Here's my recollection of the justification of timestamping...

The timestamp is not intended to represent the time since system uptime.  At
least not for any piece of reliable equipment:)  It was added to provide
"relative" times between events for more accuracat reporting (e.g. if source
timestamps X & Y as happening 500 ms apart, thats probably more accurate
than the receiver getting them 650ms apart due to the delay caused by the
limited number of OAM frames).  The origin of the utility can be found in

http://www.ieee802.org/3/efm/public/mar03/oam/gerhardt_oam_1_0303.pdf.

- Matt

On Mon, 4 Aug 2003 16:38:59 -0400, Yonghong Ren <ren@appiancom.com> wrote:

> The most recent draft (2D) defines the Event Time Stamp in Event TLV as
> 16-bit, in the unit of 100ms. If we assume this "time reference" to mean
> something like a system uptime, then it's simply not wide enough.
> Interestingly, the Event Time Stamp references section 30.11.1.1.37-40,
> where the time stamps seem to be defined as 32-bit integers. So, perhaps
> it's simply a typographical error?
> I would appreciate any clarification for me. Thanks.
> -- Yonghong Ren
>