RE: How does XML help Network Operators

["Allen, Keith" <kallen@tri.sbc.com>]

Oops, sorry.  David Perkins caught me.

Forgot a division in there.  How many seconds in a minute??  (Hey, it was a
long day.)

I still think my point stands.  I am looking at things from the NMS point of
view, not the device point of view, because we (service providers) need to
find an NMS as well as devices that can utilize this interface.  If we have
one NMS managing the whole network then it will have to be on the "north"
end of a lot of backups and restores.  That's a lot of work for one system.
And even if we have more than one, it will likely be a relatively small
number, not thousands.

Still, here are some real numbers.  (I probably should have led with these,
I just thought a VoIP scenario would pique your imagination.)  On peak days,
we have processed 10,000 DSL orders, most of them additions.  That's across
the whole company, and we use a single NMS for DSL "provisioning."  I think
we are still batching them up and processing them during a window at night,
which results in pretty high volumes in a short period of time.  We have
plans, though, to go to a more real-time approach.  Even if you spread those
10,000 orders out over a 10 hour day, you still get less than 4 seconds to
process an order  (This time, my 10 year-old son checked my work: 10,000
orders / 10 hours = 1,000 orders/hr.;  3,600 seconds/hr / 1,000 orders/hr =
3.6 seconds/order.  If this is wrong blame him!).  I don't think that's
enough time to restore the entire state of a 2,000 line DSLAM.  Also,
sometimes processing an order requires touching more than one device.  In
our network, for example, the DSLAMs are being networked hierarchically.
Processing an order might mean configuring 2 DSLAMs as well as some ATM
switches.  I know, the NMS can be multi-threaded, we can have more than one
instance, etc.  I still think it would be pushing it.

David asked about an example such as the number of voice orders in the Bay
area.  In all of Pacific Bell, we AVERAGE 107,000 orders per day.  We divide
California into two regions, north and south.  So, figure about 50,000
orders per day for Northern California, of which of course the Bay area
would be a part.  I couldn't find peak day information, but there can be a
good amount of variance, so a peak day could be well over 50,000 orders.

This was the original list of operator requirements I saw in one of Bert's
messages:

- dump/save configuration
- load/restore configuration
- activate a new configuration at time X or at reboot
- rollback a configuration
- network level configuration, as in the following points
  - mutli-device transactions (e.g. activate a specific
    configuration at mutliple devices in sync
  - rollback at multiple devices in sync
- who can do what on which devices, as in the
  following points
  - what are the identities of who can do what
  - how can they be grouped
  - how do we do access control fo such groups
- transport security
  - how is the transport properly secured

It just didn't give me the feeling that this group was working towards the
kind of high-volume, interactive protocol I think we need.

Keith Allen
.


 -----Original Message-----
From: 	David T. Perkins [mailto:dperkins@dsperkins.com] 
Sent:	Tuesday, October 01, 2002 3:10 PM
To:	Allen, Keith; xmlconf
Subject:	RE: How does XML help Network Operators

HI,

In the discussions so far, there is not a limitation to bulk configuration
retrieval from a managed system and bulk configuration restore to a
managed system.

In the numbers that you provided, which are...
>A large network operator with 60 million telephone
>subscribers figures that some day a good chunk of those subscribers might
be
>served with VoIP.  Just to pick an easy number, lets say at some point they
>are converting 1 million subscribers per year (which, as I'm sure you can
>calculate, means it would take 60 years to evolve from POTS to VoIP).  Even
>at that leisurely pace, they are converting on the order of 10 subscribers
>per second each 8 hour working day.
I have some problem with these. But first lets verify....
(1000000 conversions per year)/(6 work days per week * 52 weeks)=3205 per
day
and 
(3205 per day)/(8 hours per day)= 400 per hour
and
(400 per hour)/(60 min per hour) = 6.7 per minute

So, I don't see 10 per second. (Also, these are changes over the WHOLE
network, and not per device.) What would the numbers turn out when
the number of devices was factored in? With automation, why an 8 hour
day or only 6 days per week?

These numbers are guesses. It would be much better to get REAL numbers.
For example, in a given meto area (such as the SF bay area), how many
phone service adds/deletes/changes occur per day. What is the total
number of lines in service.

Bert and all of the rest of us want REAL data and not projections.

Regards,
/david t. perkins

--
to unsubscribe send a message to xmlconf-request@ops.ietf.org with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/xmlconf/>