RE: [EFM] OAM - Faye's seven points
In looking at infrastructure deployment architectures, I agree with your
statement "In another words, there are two management segments: One from
EMS to head-end and one from head-end to CPE(s)."
For the purposes of OAM in EFM, it is the link between the "head-end to
CPE(s). This the reason that I referred to the EFM deployments
infrastructure a "closed system" in my presentation
http://www.ieee802.org/3/efm/public/sep01/bynum_3_0901.pdf on the EFM web page.
At 10:20 AM 9/24/01 -0700, Faye Ly wrote:
>I am pretty sure your points are valid. Based on my past
>experience (that is not the only scenario, I do recognize
>that!), here are some points.
>The assumptions I stated 'to avoid sending a technician' are:
>1. CPE is cheap, therefore it will be quite expensive to
>have an alternative dial up management interface dedicated
>for each CPE. The only way any management commands
>from carrier can get to CPE is through the headend.
>2. When a CPE is in trouble, not necessarily just the EFM
>link that is bad. This can be software hanging or subscriber
>line went bad. (This means reset can reset both the CPE
>OR the subscriber line).
>3. If the first reset CPE command timed out, falls back to
>'link keep-alive' stage and determine if the link is good/bad?
>Note that if the 'reset' command is issued by EMS to the
>headend or CPE directly, EMS usually does have retries.
>This becomes a design issue with headend to handle
>the proxy commands correctly that not to congest it's
>management links to the CPEs.
>In another words, there are two management segments:
>One from EMS to head-end
>and one from head-end to CPE(s)
>The later is closer to device management (like ILMI in
>ATM) than legacy network management.
>These are my assumptions, please correct me if I am
>From: Geoff Thompson [mailto:gthompso@xxxxxxxxxxxxxxxxxx]
>Sent: Thursday, September 20, 2001 9:47 AM
>To: Faye Ly
>Subject: RE: [EFM] OAM - Faye's seven points
>At 04:54 PM 9/18/01 -0700, Faye Ly wrote:
>>Some OAM traffic is more critical than others. For example -
>>OAM command like 'reset' (in our case, reset CPE) should not be retried.
>Actually, I don't agree. Resets should be confirmed by the entity being
>reset. If a confirmation is not received in a "reasonable" amount of time
>then the protocol should try some set number of times before giving up and
>assuming that communication has been lost with the entity.
>>Certainly don't want to reset the CPE a couple of times just because
>>network is slow.
>Agreed, but that doesn't negate what I said above rather it says that your
>reset retry protocol should have a reasonable amount of time between retrys.
>>Giving up means sending a technician to the field to actually toggle the
>>power button on the CPE.
>If this is the case then the equipment vendor will have done a bad job of
>systems design. Hopefully there will be enough information elsewhere in
>the system to help figure out if the cable has gone open or hopelessly
>noisy or the far end power has gone down or anyone of a number of other
>real live faults as opposed to a far-end microprocessor program counter
>going off into the weeds. That is the result of poor design.
>>This is very expensive.
>>The whole reason of requesting for a dedicated OAM channel/IPG/whatever
>>is to gurantee that no actual human needs to be sent to the
>>field. Maybe this is not do-able but we ought to try our best.
>I do not believe that there is any correlation between the need to
>send/not send a technician to the field and the presence of a separate OAM
>>On a side note -
>>Can you please clarify the statement "P2P PHYs do not drop packets"?
>P2P PHYS don't drop packets any more (or any less) than any other piece of
>There are only 2 places for bits to go in a setup that consists of:
>P2P PHY |_____MEDIUM_______| P2P PHY
>1) Go where they are supposed to.
>2) Not go there, in which case you can't communicate with anything in the
>When the link is reestablished (or gone around) then there are already
>plenty of counters to look at in the existing MAC management to count the
>>This is good. I don't need to keep all those dropped packets/bytes
>>error counters then. Thanks.