I think the purpose of what the network inventory is supposed to be used
for needs to be clarified more generally, which will also answer the
question here.
I believe that fundamentally “inventory” is just that: an inventory of
what is there, or supposed to be there. Generally something that you
could attach an “asset tag” to, either physical or virtual. That is,
basically it constitutes “static” information. (Of course it can still
change over time, but changes will occur on slower time scales (e.g.
minutes to days, not subseconds) and it does not include dynamic state
of the assets themselves.) So, with inventory being understood that
way, dynamic information (such as current consumption) is simply not a
part of it. That being said, static aspects that describe properties of
the assets in the inventory might be – information such as equipment
type, which could be tied also to data sheet stuff such as power ratings.
That being said, there may also be a need for a holistic view of the
network, providing a view of current status etc. To organize this
information, utilizing a network inventory model is a natural
candidate. However, the needs for this type of use are different from
those of a traditional inventory. Crucially, this data would not be
“owned” e.g. by a controller having a network view, but by the
individual devices in question, where it would also be retrieved from or
subscribed to. This means that you would need to organize this not as
part of a network inventory, but as part of the data provided by the
device. (In the past, this would be broadly referred to as the device’s
Management Information Base, or MIB – not to be confused with MIBs in
the narrower SNMP sense).
From that perspective, dynamic metrics such as the current power
consumption do not belong in the network inventory as currently
defined. If we wanted to put it there, the next question would be why
stop there? Why not put other operational data there as well, why not
configuration data? Arguably you could organize any type of management
data into a network inventory, which would however not take into the
account the fact that in general, you will access most management data
from the device directly.
--- Alex
On 8/21/2023 7:22 AM, Tianran Zhou wrote:
Hi Daniele,
Thanks for the clarification.
Now I understand what you want to include is static value.
Best,
Tianran
------------------------------------------------------------------------
Sent from WeLink
*发件人: *Daniele Ceccarelli<daniele.i...@gmail.com>
*收件人: *Tianran Zhou<zhoutian...@huawei.com>
*抄送: *Alexander L Clemm<lud...@clemm.org>;opsawg<opsawg@ietf.org>
*主题: *Re: [OPSAWG] Some thoughts on Green Networking Metrics
*时间: *2023-08-21 20:33:57
HI Tianran, Alex,
the power consumed at a given time is highly changing over time, but
parameters like e.g. power consumption of a card at 50%, 75% and 100%
of the load don't.
I don't disagree with Alex when saying that this is not directly
inventory, but it's strictly related to it and should probably be an
augmentation of the device inventory.
BR
Daniele
On Fri, Aug 18, 2023 at 3:20 AM Tianran Zhou
<zhoutianran=40huawei....@dmarc.ietf.org> wrote:
I am not quite clear about the applicability of the inventory.
What’s the difference with hardware model or entity model.
I see energy work was related to entity mib in eman before:
https://datatracker.ietf.org/wg/eman/documents/
It seems inventory should be something static from the name. But
IMO, the energy metrics are dynamic, can will change all the time.
Best,
Tianran
*From:*OPSAWG [mailto:opsawg-boun...@ietf.org] *On Behalf Of
*Alexander L Clemm
*Sent:* Friday, August 18, 2023 7:05 AM
*To:* opsawg@ietf.org
*Subject:* Re: [OPSAWG] Some thoughts on Green Networking Metrics
Hi Daniele,
apologies for the late reply.
I think inventory is somewhat orthogonal to this, but of course
devices and equipment (including chassis, line cards, equipment
holders etc) will be considered part of inventory. Therefore via
transitive closure it is certainly conceivable to make power
consumption data accessible via inventory. This could make sense
as part of a consolidated controller view of a network. However,
on a network element itself, the network inventory aspect would
not apply but the metrics should still be available so the
device/equipment level category still applies. As to whether
device level data should be replicated as part of network
inventory data would presumably depend on the use case.
--- Alex
On 7/26/2023 6:35 PM, Daniele Ceccarelli (dceccare) wrote:
Hi Alex, all,
Just following up on the comment I did ad the mic earlier today.
The drafts speaks about metrics at: device/equipment level,
flow level, path level, network level.
The device/equipment level covers power consumption per
chassis, line card and port at different loads of traffic,
hence IMO should fall into the inventory category.
Would you agree?
Cheers,
Daniele
_______________________________________________
OPSAWG mailing list
OPSAWG@ietf.org
https://www.ietf.org/mailman/listinfo/opsawg
_______________________________________________
OPSAWG mailing list
OPSAWG@ietf.org
https://www.ietf.org/mailman/listinfo/opsawg
_______________________________________________
OPSAWG mailing list
OPSAWG@ietf.org
https://www.ietf.org/mailman/listinfo/opsawg