[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

b/w example values

Hi!

> Now, that I agree with.  We're not going to come to agreement as our
> needs differ.

I dont see these as being different needs at all.  The discussion over
defaults/not-defaults is an engineering and design tradeoff with no
realy discernible difference to our end-users (I hope).

> I think this, along with other directions that have been taken, will
> prevent netconf from being deployed in some environments, yes.  But I
> don't work in the same environments you do.  We've proven that well
> enough.

I seriously asking you if you really think that defaults/no-defaults
will effect adoption?  I just cant see that it matters to end-users.

What environments are you working in that you think the needs are
drastically different?  I think this discussion is important.

In the case of your cidrTable example, 5 -1s (not specificed because
they are defaults) only amounts to :

  5 * (1-byte tag, 1-byte len, 4-byte value) = 30 bytes

I hardly consider an extra 30-bytes xmit'ing over a 10Mbps or the more
likely 100Mbps anything to sneeze at.

I went and asked some folks at my former employer (I sold my company
Empire Technologies to Concord in 1999 and worked for them for 3
years) about their b/w utilization for their pollers.  Here is a blurb
from their publicly available KB.  Keep in mind that this blurb is in
reference to SNMP/BER.  However, I believe its in the same ballpark as
XML encoding.  Also, in the case of netconf, a reply may be tiny with
just an "noError" value.  So, we may get back some of the bytes used
to articular defaults.

Here is the blurb from the KB regarding SNMP polling:

  If we are polling 100 elements, this is approximately 65000 bytes
  which are spread out across the roughly 30 seconds it takes to do a
  poll. This works out to 2166 bytes per second, which is 0.17% of an
  ethernet, or approximately 2 tenths of a percent utilization during
  the 30 seconds every 5 minutes you are polling. Looking at what
  percentage of the overall bandwidth over 5 minutes you are utilizing
  gives you an overall utilization of about .0017%, around 2
  hundredths of a percent utilization.

  If some of the polls are going across wide area links, the
  percentage utilization per polled element will be higher, but the
  number of devices poller per wide area link is likely to be
  lower. Taking the case of 10 devices polled over a 64K link, we have
  6500 bytes, spread out over 10 seconds (650 bytes/second). This
  works out to about 4% utilization of the full duplex 64K link during
  the 10 seconds every 5 minutes you are polling. Looking at what
  percentage of the overall bandwidth over 5 minutes you are utilizing
  gives you an overall utilization of about .027%, around 3 tenths of
  a percent.

When they say "ethernet", they really are talking about a 10 Mbps
ethernet.  

So, its pretty tiny either way.  Does this help?

Bobby

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