HTML Version available at
Minutes of the HTTP MIB BOF
Monday, March 4, 1996
35th IETF, Los Angeles
Carl Kalbfleisch presented the status of the work on the MIB to
date. A Summary of Carl's talk is included below*. The most
current MIB may be found at
http://http-mib.onramp.net/draft_2.my
There was a great deal of discussion and some confusion over the
purpose and requirements for this MIB. Regarding the
customer/user requirements for the MIB ...
We were fortunate to have representatives from the Centre for Earth
Observation (CEO) in Europe at the BOF. (More information is included
below**). These individuals, Rui Meneses and Dirk Vangulik, are
participating in CEO related projects for enabling the use and access
to EO information and data (satellite pictures) across multiple
locations in Europe.
Rui Meneses deals in particular with projects related with
management of distributed information services, both within the
CEO project and the DESIRE project
(http://www.surfnet.nl/surfnet/projects/desire), where a survey questionnaire
will be published on line at http://www.netmode.ntua/gr/musiq,
aiming at assessing the requirements of end-users, IS providers, and
developers on distributed IS management. The results of this
questionnaire could be used to draft the requirements document
and interested parties are encouraged to review the draft
questionnaires at the following location
http://www.netmode.ntua.gr/musiq/final.html
An unfortunate discovery at the BOF was that most attendees had
their own set of requirements. Most of the BOF time was
devoted to defining and explaining why it is necessary to manage
web servers, and what is required.
The conclusion at the end of the BOF was to put together a charter,
goals, and set of requirements for a WEB Server Management
MIB Working Group, and to investigate other working groups to
determine areas of overlap (specifically MADMAN and
applicationMIB).
NEW TITLE: Web Server Management MIB
(includes proxy and client)
PURPOSE: Provide a consistent method for managing heterogeneous web
servers in medium to large enterprises.
USERS AND CUSTOMERS:
* Internet Service Providers managing their networks and their
customers' web pages on web servers in those networks
* End users managing large networks of web servers
* Producers of specialized web servers
* Network managers already using SNMP to manage their networks
and who now want to extend this capability to manage web servers
REQUIREMENTS:
Web server management may be divided into three areas:
(1) Tracking specific activity in the web server
-Error generating and reporting
-Capacity planning
-Quality of service (in real time)
-"Log Digester"
(2) Retrieval services -- an abstraction decoupling the
information space from the underlying transport mechanism
(3) Document information store -- managing documents and their
dependencies among applications
Most of the discussion in the meeting revolved around the first
item:
Tracking Specific Activity in the Web Server
1. Is my server saturated? How many user/connections were
refused? What were the 500-level errors?
2. How long is it taking the server to respond? How many
queues? holding time? connection handling characteristics?
3. Real time versus after the fact errors. Real time implies
the quality of service, i.e. being able to respond to a
consistent set of problems rapidly.
4. After-the-fact or "logged" errors
Is the current log data satisfactory for this?
The log file mechanism found in most web server software
packages contains too much data in a format that is
difficult to use for SNMP management. A summary of this
data is required.
5. Counting page hits (use a table, not a log file).
6. Manage configuration?
7. Manage performance?
Threads management was discussed. How is this resource
utilized (almost full, 50%, little)? Check the current
active incoming and outgoing threads (connections) and compare
them with max-threads-allowed and min-threads allowed.
How to affect this resource utilization (by changing the
allowed boundaries - setting max-threads-allowed and/or
min-threads-allowed to different values).
*Summary of Carl Kalbfleisch's Presentation at the BOF in Los
Angeles, March 4, 1996.
--------------------------------------
HTTP-MIB
Background
--------------------------------------
* September 1995
o I Started at OnRamp Technologies.
o Looked for information on SNMP management of Web servers.
o Found Management of a WWW server using SNMP, by Carlos Picoto and
Pedro Veiga.
o No standard efforts underway.
* October 1995
o half a dozen folks expressed interest from various email list and
USENET posts
o http-mib@onramp.net mailing list started
* November 1995
o Created first draft of a MIB
o There are at least two implementations based on this draft
* December 1995
o Discuss draft
o not on agenda for Dallas IETF
* January 1996
o Created second draft of a MIB
o Extends MADMAN efforts
o Documents several threads from mailing list
* February 1995
o Preliminary work on document tables
o Preliminary work on access and other errors
o discussion on timeouts
* March 1996
o BOF at LA IETF Meeting
----------------------------------------------------------------------------
Last Update: March 1, 1996
Carl W. Kalbfleisch
cwk@onramp.net
===================================================================
**The CEO project has implemented a set of three MIBs to allow
management of WEB based Information Service:
(1) Information Store MIB
(2) Retrieval Services MIB
(3) http MIB -- this latter in strong collaboration with the
work developed at the http-MIB mailing list (through Harrie
Hazewinkel, Mark Gamble, and Erik van Hengstum)
Information related to these MIBs can be found at (location will
change):
http://case.cs.utwente.nl:1234/ISMIB.asn1
http://case.cs.utwente.nl:1234/RSMIB.asn1
http://case.cs.utwente.nl:1234/HTTPMIB.asn1