Tutorial 2: 10th October 2019 Flashcards
QoS
Can mobile IP implement persistent connections now? Why?
No, as it isn’t deployed. The lack of a homogenous IP mobility system prevents mobile, persistent end-to-end connections
What would be needed to implement universal QoS?
A common method of pricing/tariffing, accounting, priority classifications, and how to change the priority of users/flows
What would be the effects of changing the granularity of a universal QoS?
If done on a per-packet level, it would increase the network load from a huge number more transactions
What are the effects of volatile routing in a universal QoS?
It makes tracking packets difficult, so finding the correct amount to charge users for them is also made hard
How would a universal QoS system scale? Why?
Very poorly. It would massively increase the load on a network and require more operations to do with accounting and classifying packets.
How much load would a universal QoS system place upon the Internet core? Why?
A lot, because it would have to carry out accounting and charging as well as finding the path of each packet to count the number of steps it takes and where it came from to pass a “receipt” back to the sender.
Why should you not use predefined paths in QoS systems?
It would cause a processing overhead in routers, especially for long paths, and mean that you may not take the optimal path. >= 1 nodes on the path could go down and make it unavailable. Knowledge of routes could be utilsed for DDoS attacks.
What is the difference between fast path and slow path?
Fast path = forwarding performed by hardware, in the network core. Slow path = forwarding by software, towards the network edge.
What would be the pros and cons of using “gas” as a metaphor for the computing power required to forward and deliver a packet?
It would need a single trustless and centralised currency as well as security measures for it, giving an overhead and low scalability that makes it inviable.
It would take very long to deploy even after it became feasible and a protocol was agreed.
What are the main issues with QoS mechanisms?
unscalable; any possible solution would be all-or-nothing so would be almost impossible to deploy (as would need total deployment); existing legacy applications would need to be updated or have management apps to provide an interface
What would be required to deploy QoS mechanisms across the Internet?
total participation in a scalable set of mechanisms, and updates to legacy network apps or management apps to handle the updated network
How can you handle QoS mechanisms with legacy applications?
Either update them directly or use management apps on top of them to interface to the new QoS mechanism
What is GPS with respect to scheduling algorithms?
Generalized processor sharing (GPS) is an ideal scheduling algorithm for process schedulers and network schedulers. It is related to the fair-queuing principle which groups packets into classes and shares the service capacity between them. GPS shares this capacity according to some fixed weights.
How does the heterogeneity of QoS mechanisms stop them from being compatible?
They work in different ways so they can’t be used as a part of a single mechanism. It gives an overhead of work both to implement any interfaces, if at all possible, between different QoS mechanisms, as well as in the networks: maintaining more SLAs to peer with more other networks.
How does BT use DIFFSERV?
It uses it for BT vision, which is live TV broadcasts on the same line as Internet connections; they don’t interfere despite being on the same line
