[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[ipcdn] RE: Q of docsIfCmtsChannelUtUtilization for Upstream
Hi Minnie Lu,
The other concern you have is the US Interface Utilization calculation,
I will assume that if a dynamic Burst profile adjustment is chosen by
the CMTS, the Utilization interval must be adjusted to be lower that the
dynamic Burst profile rate change ( actually half to be sort of Nyquist
compliant.)
To support your case the description should be changed for the US
interface, and I see no real benefit for overall historical data
analysis perspective.
The weighted perception of US Interface I believe is more realistic
(valuable? )and shows the bandwidth utilization sensitivity better (See
examples below)
At the end, the users consume Bandwidth with a bad or good minislots
usage/efficiencies of them, based on the Modulation profile used and CM
allocation among channels.
Also since burst types could be variable (Dynamic) the only average
constant would be user bandwidth usage at least for planning purposes
and historical recollection, weighted usage is a normalization of the
differences among channels.
Supposed Channels A, B and lets use a coins (minislots) vs coins value
(burst type)
A has 3 5 cents coins
B has 6 10 cents coins
Minislots usage = coins spend
Weighted Minislots usage -> cents usage
In Us Channels (A or B) spend coins vs spend cents ( minislots vs BW)
is the same.
With the US DS counters you can get also the U based on sharp
Total BW ( in cents is 3*5 + 6*10) = 75
Total coins = 3 + 6 = 9
If A spends 2 coins and B spends 1 coin
Spend coins = 2 + 1 = 3 U% 3/9 = 33%
Spend Cents: 2*5 + 1*10 = 20 U% = 20/75 = 26%
ChaA U = 2/3 = 66%
ChaB U = 1/6 = 16%
Supposed the other case of B more used
A Spends 1 coin and B spend 4 coins
US channel
Coins spend 1 + 4 = 5; U = 5/9 = 55%
Cents spend 1*5 + 5*10 = 55; U% = 75%
ChaA U = 1/3 = 33%
ChaB U = 5/6 = 83%
-----Original Message-----
From: Eduardo Cardona
Sent: Tuesday, November 11, 2003 11:37 AM
To: 'Minnie Lu'; 'ipcdn@ietf.org'
Cc: DOCSIS OSS Majordomo List; Jean-Francois Mule; Greg White
Subject: RE: Q of docsIfCmtsChannelUtUtilization for Upstream
Hi Minnie,
It is debatable if the current text imposes the fixed bandwidth
allocation.
We believe the text should be taken in the context of the measurement
interval. From that perspective we think It won't preclude or limit the
CMTS capabilities to change dynamically the % bandwidth allocation.
Internally at CL, Greg and I looked at the case and probably the
proposed text below may help.
-> added at the end of the below paragraph: "over the most recent
-> utilization interval."
For an upstream interface that has multiple logical
upstream channels enabled, the utilization index is a
weighted sum of utilization indices for the logical
channels. The weight for each utilization index is the
percentage of upstream minislots allocated for the
corresponding logical channel over the most recent
utilization interval.
Eduardo, Greg
-----Original Message-----
From: Minnie Lu [mailto:milu@cisco.com]
Sent: Monday, November 10, 2003 6:25 PM
To: Jean-Francois Mule; Eduardo Cardona; Greg White
Cc: milu@cisco.com; DOCSIS OSS Majordomo List
Subject: Q of docsIfCmtsChannelUtUtilization for Upstream
Hi, all,
I have a question about upstream docsIfCmtsChannelUtUtilization. In
the
description, it says:
"Upstream Channel Utilization Index:
The upstream channel utilization index is expressed as a
percentage of minislots utilized on the physical channel,
regardless of burst type. For an Initial Maintenance
region, the minislots for the complete region are
considered utilized if the CMTS received an upstream
burst within the region from any CM on the physical
channel. For contention REQ and REQ/DATA regions, the
minislots for a transmission opportunity within the
region are considered utilized if the CMTS received an
upstream burst within the opportunity from any CM on the
physical channel. For all other regions, utilized
minislots are those in which the CMTS granted
bandwidth to any unicast SID on the physical channel.
For an upstream interface that has multiple logical
upstream channels enabled, the utilization index is a
weighted sum of utilization indices for the logical
channels. The weight for each utilization index is the
percentage of upstream minislots allocated for the
corresponding logical channel.
Example:
If 75% of bandwidth is allocated to the first logical
channel and 25% to the second, and the utilization
indices for each are 60 and 40 respectively, the
utilization index for the upstream physical channel is
(60 * 0.75) + (40 * 0.25) = 55. This figure
applies to the most recent utilization interval."
Q: for multiple logical channels enabled case, why can it be still
"a percentage of minislots utilized on the physical channel,
regardless
of burst type." ? That is, (the sum of minislots utilized for all
logical
channels) /
(the sum of total minislots for all
logical
channels)
In the current description, it seems that CMTS must support fixed %
of
bandwidth allocated for each logical channel, I cannot find such
requirement in RFIv2.0 spec. Also according to the RFIv2.0 spec, some
CMTS
vendor might choose to support that the % could be changed dynamically
at
the run time for logical channels or some vendors might choose to
support
dynamical bandwidth allocation without any fixed %.
So may I suggest to remove the follow statements from the description
of
docsIfCmtsChannelUtUtilization ?
" For an upstream interface that has multiple logical
upstream channels enabled, the utilization index is a
weighted sum of utilization indices for the logical
channels. The weight for each utilization index is the
percentage of upstream minislots allocated for the
corresponding logical channel.
Example:
If 75% of bandwidth is allocated to the first logical
channel and 25% to the second, and the utilization
indices for each are 60 and 40 respectively, the
utilization index for the upstream physical channel is
(60 * 0.75) + (40 * 0.25) = 55. This figure
applies to the most recent utilization interval."
If I miss any thing, please advice me !
Thanks a lot for your help !
Minnie
_______________________________________________
IPCDN mailing list
IPCDN@ietf.org
https://www1.ietf.org/mailman/listinfo/ipcdn