(version for book website) E E 681 - Lecture 1 Kick-off Lecture: Introduction to Survivable Transport Networks Wayne D. Grover TRLabs & University of Alberta.
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
(version for book website)
E E 681 - Lecture 1E E 681 - Lecture 1
Kick-off Lecture: Introduction Kick-off Lecture: Introduction to Survivable Transport to Survivable Transport
• Graduates of E E 681 will have a basic preparation and awareness of current and emerging transport networking alternatives, mechanisms, issues, and design theory, enabling them to continue in:
– research: will be equipped to pursue a thesis project and participate in ongoing graduate research in these areas
– R&D: will be able to contribute to transport networking equipment design and product strategies
– operations: will be able to contribute to network planning and network evolution strategy
• Voice-band switching, Internet, private lines, corporate networks, ATM networks, etc. are all ‘virtual’, logical abstractions implemented within the transport network.
• The transport network sits “just above” the physical transmission systems in a layering sense.
• Individual switched connections, leased lines, pipes between IP routers, etc. do not “make their own way directly” over the fiber systems..
• Rather, traffic of all sorts is “groomed” to fill standard rate “containers” created in the transport network.
• “grooming” at or near clusters of sources (the edge or access network) tries to efficiently fill these containers so they won’t need to be opened again (i.e., processed at a call, cell or packet level), until at or near their destinations.
• Transport network thus sees a composite “demand pattern” (in STS-n units typically) that is the resultant totals of point-to-point container requirements arising from all client network / service layer requirements, e.g., trunk groups, IP pipes, leased lines, private networks, etc.
• “node”: – digital (Sonet) and / or optical (wavelength) cross-connects
• generic “link” resource: – standardized logical bandwidth units such as DS1, DS3, STS-n,
ATM VP, wavelengths, wavebands
• main survivability principles: – ring, mesh, backup-VP or p-cycle based real-time restoration re-
routing.
• “performance” measures:– restorability (of spans, nodes)– restoration time (e.g., 150 ms - 2 sec)– end-to-end path availability (e.g., 99.996 on 4,000 km HRDP)– best efforts and / or assured restoration classes – path provisioning time (seconds or days ?)
• This is why fully router-based “IP over light” (just as prior “ATM on glass”) is improbable when the short-term hype is replaced by longer term performance, complexity, cost, maintenance and operational assessments.
• The single biggest factor in IP QoS in particular is the average number of router hops in a ‘connection’...
• All other transport industries find an optimum combination of access grooming/muxing and backbone transport; pure “IP over light” implies unpacking and reloading the moving van in every city en-route.
• Or, “would you move a house brick by brick?”
• More likely structure is to stat mux and groom in one or two access stages, then launch into near mesh of non-stochastic high OCn or ?-based transport paths.
• Restorability: the fraction of working demand flows affected by a failure that are restored or for which a restoration path set solution is feasible.
• Redundancy: the ratio of spare capacity required in a network to meet restorability goals to working capacity required only to route demands without survivability concerns.
* we will return to all these concepts in greater depth. The aim today is just to create an
• Reliability: the probability that a system operates without a service-affecting failure for a given amount of time. R(t) can be thought of as the probability distribution function of time-to-first-failure from a known-good starting state.
• Availability: the probability that a continuously operating system undergoing repair after each failure is found in the “up” state at any random time in the future.
• Does a “fully restorable” network have 100% availability ?
– No. If the network restorability design is for 100% restorability to all n-failure scenarios, “(n+1) failure” scenarios may be outage-causing.
– In practice commercial / public networks used to have n = 0 (in the sense that no cable cuts would be 100% restorable). In which case addition of redundancy to get to n=1 (full restorability against any single cable cut) gives a massive boost in availability.
– But availability does not reach unity because then dual failure scenarios can then cause outage.
– --> leads to usual economic practice of : design for 100% single-failure restorability, and analyze for the dual failure (un)availability.