From: Paul Tykodi (ptykodi@tykodi.com)
Date: Tue Feb 14 2006 - 22:50:23 EST
Dear Bill,
The host I was most recently analyzing was an IBM iSeries - AS/400 host. The
MIB itself worked flawlessly. I am not suggesting that it was somehow the
culprit for the slow printing or that it did not work correctly. The
communication started OK and then the host was concerned that a response
packet was not received in a timely fashion. It began a significant SNMP
based questioning process to determine the current hardware status of the
device and interspersed with the SNMP questions about whether the device was
in error or not came a re-transmission of a potentially lost packet just to
be safe.
Pretty soon the majority of the communication on the wire revolved around
SNMP discussions as to the device's status and data packet re-transmissions
and confirmations from the printing device that it had indeed received the
packet re-transmissions. As you mention, the whole idea of printing
information had become unfortunately a secondary concern.
In the end, all of the data was printed and no errors were reported by the
host. Unfortunately the method utilized to determine that everything was
actually fine was so intrusive on the printing process that I feel
comfortable saying I believe that a typical customer (having paid a fee for
their printing device related to its rated engine performance) would
probably not have accepted the result as commercially viable.
So my previous comment is directed more towards device managing software
product's use of MIB capabilities (especially if more interesting things to
check are added into future MIB's) and the impact that significant device
status verifications can have on the actual process (in this case printing),
which is being monitored.
Thus in the future if some type of RFC or other standards document were to
be produced, my suggestion would be to include some examples that tried to
help steer software developers implementing use of MIB data away from
creating the issue you outline in point b. below.
Thanks.
Best Regards,
/Paul
--Paul Tykodi Principal Consultant TCS - Tykodi Consulting Services LLC
Tel/Fax: 603-343-1820 Mobile: 603-866-0712 E-mail: ptykodi@tykodi.com WWW: <http://www.tykodi.com/> http://www.tykodi.com
_____
From: pmp-owner@pwg.org [mailto:pmp-owner@pwg.org] On Behalf Of wamwagner@comcast.net Sent: Tuesday, February 14, 2006 10:33 PM To: ptykodi@tykodi.com; pmp@pwg.org Cc: Paul Tykodi Subject: RE: Feedback - PMP> Minutes of the MFP Teleconference 20060214
Paul,
Thanks for sending in your observation. I have worked with printers and SNMP management for many years and have not seen anything like the sort of slowdown that you cite. Perhaps this is because I have worked with slower machines and printers/MFPs with separate NICs. At any rate, a basic SNMP tenet is that servicing of SNMP is secondary to the main purpose of the device. Indeed, reflecting this, I have seen missed or late SNMP responses during periods of high print activity.
Of course, it is desirable to have efficient MIBs, something that sometimes gets lost in this era of "human readability". Although you may have contradicting data, I would suggest that the current public MIBs are not in themselves inefficient and that the problem you observed may be due to other factors such as:
a. certain private MIBS use an indirect addressing approach, particularly for writes, which may make for some elegance but does complicate interaction
b. many management applications are terribly inefficient, repeatedly querying the same (sometimes status) variable, and often unnecessarily dumping blocks of data.
c. Drastically underpowered controllers and/or poor handling of priorities
Although I understand that it may be difficult to release such information, it would be useful to have some information on the specifics of the slow-down. the condition the management station was querying, the objects being queried, etc.
Bill Wagner, TIC
-------------- Original message -------------- From: "Paul Tykodi" <ptykodi@tykodi.com>
Dear List,
During the last year, I have been involved in some network analysis looking at how certain hosts use the current printer MIB to determine device status (including that of MFP's) and what effect a significant number of SNMP queries and responses can have on effective printing throughput (at times rather dramatic reduction in achievable throughput).
In looking at the minutes from today's meeting, I would suggest that it might be a good idea to consider whether MIB optimization should be a category for an MFP alerts project. The idea would be to at least minimally describe some best practices for MIB usage, which would result in the host obtaining the required information using the smallest SNMP query and response packet transmission overhead possible.
In case people are wondering how dramatic a reduction in PPM I have observed when SNMP traffic is significant (host trying to determine whether device is in error or not - multiple queries are sent asking more and more specific questions of the printer MIB), I have seen printers and MFP's with rated speeds in the 75 - 125 PPM range reduced to achieving actual throughput in the 10 to 20 PPM range.
HTH
Best Regards,
/Paul
--
Paul Tykodi Principal Consultant TCS - Tykodi Consulting Services LLC
Tel/Fax: 603-343-1820 Mobile: 603-866-0712 E-mail: ptykodi@tykodi.com WWW: <http://www.tykodi.com/> http://www.tykodi.com
_____
From: pmp-owner@pwg.org [mailto:pmp-owner@pwg.org] On Behalf Of Bergman, Ron Sent: Tuesday, February 14, 2006 7:02 PM To: pmp@pwg.org Subject: PMP> Minutes of the MFP Teleconference 20060214
The minutes can be found at:
ftp://ftp.pwg.org/pub/pwg/pmp/minutes/mfp/MFP_Minutes_20060214.pdf
Ron Bergman Chairman, Printer MIBs Working Group
This archive was generated by hypermail 2.1.4 : Tue Feb 14 2006 - 22:50:25 EST