Attendance
Jean Hammond, the chair of the QoS Forum, introduced the reasoning
behind the use of scenarios. Her slides are
here
Scenarios:
- Communicate clearly
- Show detailed relevance to parties from differing domains
- Support cross domain co-working
- Allow architectures to be checked against concrete examples
Also, Business Scenarios can be carried out on the same examples to
show commercial relevance
The use of the scenarios would be, for example: in co-working with
other forums and consortia, such as the Open Group Real-time Forum and
the TeleManagement Forum
Jean then described two scenarios which are documented at http://www.opengroup.org/projects/qos_slas/
VOIP Scenario
The first related to VOIP, and described some proposed QoS-based
Traffic Handling
Amongst the SLAs that could be considered were
- Service Availability:
- Downtime not to exceed 6 min. for preceding 6-month period
- Call Setup Performance:
- Dial Response Time not to exceed 3 seconds for any call
- Call Setup Time not to exceed 3 seconds for any call
- Call Quality:
- MOS not to fall below 3.9 for preceding 6-month period
Navy Scenario
A second scenario, relating to the Navy, had been
received from Deborah Goldsmith of Mitre Corporation. The system
is designed to be on-ship, designed to carry regular applications such
as VoIP together with time critical systems. There is a series of
different interconnects between the ships and the rest of the
system. There are some emerging requirements in terms of delay
priorities; the scenario brings together traditional command-and-control
systems and the traffic requirements of a small city.
Boeing Scenario
Barl Bunje presented a summary of a Boeing Scenario; his slides are here
The scenario is based on an ERP/MRP system within Boeing which has a
large number of off-the-shelf applications that have been stitched
together, and the resulting overall system is edxtremely complicated.
The reason for looking at SLAs is not just to ensure that response
time is adequate, but to ensure that budgets are being allocated in the
best way.
There are end-to-end SLAs on transactions through the system, which
simply measure response time to the terminal, based on factory
requirements; there is an SLA for each transaction, of which there are
over 100. Availability issues are trapped separately and these
problems are dealt with differently.
There is extensive instrumentation, which grew at the same time as
the application, gathering data to measure compliance. This data
is analysed by commercial tools.
Karl Schopmeyer presented some issues from a presentation
he had prepared on Application Manageability.
There was an internal paper, developed within Boeing, which it may be
possible to make available to the group.
Action: Carl will check this.
There was discussion over whether some general principles for SLAs
could be developed from this scenario.
