| To: | Ontology Summit 2014 discussion <ontology-summit@xxxxxxxxxxxxxxxx> |
|---|---|
| From: | "John Yanosy Jr." <jyanosyjr@xxxxxxxxx> |
| Date: | Tue, 28 Jan 2014 13:47:04 -0600 |
| Message-id: | <CAMyHDHgbRvCNwqNdqW4=2jMkDC4WW2rNC=F44ENcHBb=KPapCQ@xxxxxxxxxxxxxx> |
|
Andrea and All, I found your presentation on Network Management and reuse very informative. Considering my comment and question of that day where I noted the classic hierarchy of causes and effects from network element states to emergent properties of subnetworks and the network itself. This aspect ranges from dynamic to static relationships, and from logical to quantifed probabilistic calculations and relationships,, and from externally observed and managed capabilities to network autonomous functions. .
Some typical network management and planning domains are: - Costs - Cost per network capacity - Cost per offered traffic parameter - Cost per carried traffic parameter
- Cost per user - many other types of cost relationships used to compare technology and structure alternatives - Performance and QOS (Over many different granularities of time periods depending on the average duration of the variables considered)
- Typical user perceived QOS - Network Capacity (within an offered traffic profile) - Network Offered Traffic - Network Utilization - Network Element Capacity (within an offered traffic profile)
- Network Element Offered Traffic - Network Element Utilization - Availability (typically in quantitative terms) - NE Inherent Availability - Network Service Availability
There are others but the above would suffice to investigate the relationship between ontologies, network structure, and probabilistic computation. Some thoughts about Ontology Patterns
A biological equivalent model might be useful in this regard, where some subsystems are autonomous and can be affected by other subsystems as well as external causes. In biological systems we typically only observe the effects of the subsystem on the whole system during conditions of failure or stress and that is where a doctor of medicine's knowledge can help diagnose the situation. It is this knowledge about the human body, possible pathology, observed variables, experiment or lab results to gather more information, and possible diagnoses that enables good outcomes. Similar capabilities could be built into networks and their management systems by adding knowledge about these situations in the form of patterns.
Of course there would be a functional architecture where pattern knowledge can be created, evolve, and reasoned about, both logically and from a quantitative perspective. It may be possible to create ontology patterns to represent different situational network states and possible cause- effect dependencies, which are derived from past observations. It this is possible than a hierarchy and possibly a network of ontology patterns could be used to help diagnose and manage a complex network.
Ideally the network and network manger would evolve these patterns from analysis of past events, but there is also a possibility of consortia or other organizations publishing these observed patterns that could than be used by network management systems, and possible by autonomous subsystems within the network itself.
When I look at equations for network analysis, there are variables representing captured statistics from various elements and aggregation points in the network, and implied relationships between these variables and the functions using them in their calculations, as well as higher level functional computation relationships,. It may be appropriate in these patterns to also represent some mathematical relationships between variables and functions and their relationships to real network observed statistics and NE parameters. This could have benefits for reuse to add semantics about observed variables and analysis functions or other mathematical characteristics. Whether an observed statistic is a simple aggregation over what time period, or whether it represents some computer variable over a time period based on other observed variables and parameters.
Some complexities of network planning and management. In all of the above domains there is a requirement for knowledge of the architectural relationships between network elements and the domain variables of interest.
In some cases, such as equipment costs the knowledge is somewhat static once the network is configured and the network level costs are simple summations. But for other cost parameters there is a dynamic consideration where periods of observation and parameter statistics are gathered and where relationships between NE statistics and different network level statistics are calculated probabistically. But in many of these dynamic cases there is still a logical static relationship between NE statistics and higher level network statistics. Yet! for newer dynamic routing algorithms and protocols the NE element utilization statistics varies with patterns of offered traffic at the edges of the network, since routing may adapt to NE utilization levels as these external offered traffic patterns change. So in many planning scenarios the network is stressed with varied offered traffic profiles and NE and network level statistics are calculated from simulations.
When routing patterns change dynamically there is still an underlying physical infrastructure of NE elements and there inherent capacity limits that place an upper constraint on the carried traffic benefits of routing adaptation.
There are other dynamic types of adaptation that can take place due to segmentation and rerouting of smaller information units to provide a more granular network utilization and smoothing effect.
In the planning stages there are software tools that require knowledge of the network elements characteristics, the network structure, and potential configuration setttings that affect NE functional characteristics.
With availability models there are similar aspects but now we have the possibility of mismatched planning with offered traffic causing congestion in various places in the network and thus affecting different service and NE availability statistics. Availability could be affected by planned and unplanned outages as well as mismatched traffic and capacity.
Adding semantics and knowledge representation to actually off-load some cognitive complexities could improve the network manger's ability to make decisions about configuration and administrative settings in a much simpler manner, especially if the system could represent results of what-if scenarios on the network patterns.
Congestion Discussion One of my favorite dynamic problems to understand is "congestion" and its effects on availability and user perceived QOS. There are many relationships to be represented here, but I think that logical inferences from patterns might inform possible congestion states. A performance affecting failure in one NE might affect the routing behavior to cause overloading of other NEs, and cause reduction of aggregate statistics in other NEs.
I will try to create a simple pattern example for network congestion using Protégé and share soon with the community. Sincerely John - On Mon, Jan 27, 2014 at 12:53 PM, Andrea Westerinen <arwesterinen@xxxxxxxxx> wrote:
-- Peace be with you,
John A. Yanosy Jr. Mobile: 214-336-9875 eMail: jyanosyjr@xxxxxxxxx
LinkedIn: www.linkedin.com/in/jyanosyjr
_________________________________________________________________ Msg Archives: http://ontolog.cim3.net/forum/ontology-summit/ Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontology-summit/ Unsubscribe: mailto:ontology-summit-leave@xxxxxxxxxxxxxxxx Community Files: http://ontolog.cim3.net/file/work/OntologySummit2014/ Community Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?OntologySummit2014 Community Portal: http://ontolog.cim3.net/wiki/ (01) |
| <Prev in Thread] | Current Thread | [Next in Thread> |
|---|---|---|
| ||
| Previous by Date: | Re: [ontology-summit] [Reusable Content] Characterizing or measuring reuse, Gary Berg-Cross |
|---|---|
| Next by Date: | Re: [ontology-summit] [Reusable Content] Characterizing or measuring reuse, Barkmeyer, Edward J |
| Previous by Thread: | Re: [ontology-summit] [Reusable Content] Characterizing or measuring reuse, Barkmeyer, Edward J |
| Next by Thread: | Re: [ontology-summit] [Reusable Content] Characterizing or measuring reuse, John Yanosy Jr. |
| Indexes: | [Date] [Thread] [Top] [All Lists] |