I TopologyII MobilityIII SecurityIV Quality of Service (QoS)V WirelessVI Band-AidsVII EconomicsVIII Other Research AreasReferences1EE228a - Lecture 3 - Spring 2006Moving ForwardJean Walrand, scribed by Phoebus ChenAbstractThis lecture surveys several research areas in communication networking research. First, we covered an example of proposalsto change the routing topology of the internet away from shortest path routing to reduce congestion bottlenecks. Then, wecovered issues when networking under different types of environments and with different types of constraints/requirements:mobile networking, network security, quality of service (QoS), and wireless networking. This was followed by a discussions ofsome issues with transitioning/patching old network deployments and incorporating economics for pricing networks. Lastly, webriefly touched on some related research areas that might intersect networking research.I. TOPOLOGYThe 100 × 100 project [1] is a proposal to link 100 million homes with 100 Megabit/second internet connections. In thisproposal, instead of using shortest path routing for the internet backbone, the network would use a two-step routing strategy tospread out network traffic so that the network load is more uniform. This two step strategy involves first routing packets fromthe source to a random switch, and then routing packets from the random switch to the final destination (See Figure 1). Thisidea was borrowed from a similar proposal for routing data between microprocessors. In addition to this change in routing, the100x100 project proposes to apply local back pressure/feedback to reduce congestion as opposed to only applying end-to-endback pressure to reduce congestion in TCP. The 100 × 100 project is meant to replace OSPF and BGP.Of course, there are costs and technical limitations to be considered for these proposed changes to the routing protocols. Forinstance, a maximum optical fiber length of around 500 km limits the selection of random switches for routing by the source.Also, the random routing in the first step of the routing scheme needs the switches in the internet backbone to be connected, whichmakes adding new switches to scale the network more difficult. Many other questions remain, including how many switchingcenters are needed and the costs for deploying this new routing topology.Fig. 1. The routing principle for the 100x100 project involves first routing to a random destination and then to the final destination to distribute networkload.II. MOBILITYIntroducing mobility to network communications brings up unique challenges. The current implementation of DHCP (DynamicHost Configuration Protocol) is too slow to allow users to migrate from one local network to another seamlessly. There is asignificant delay before the user can renew his IP address and establish a connection with a new network. Mobile IP allows usersto migrate from network to network by indirection — data is routed to a computer with a home address, which then reroutes thedata to the mobile computer, which has a temporary care-of address [2]. This extra layer of indirection can result in an inefficientuse of bandwidth, especially when a mobile computer is on the subnetwork containing the sender and the computer with the homeaddress is far away.Of course, mobility also brings up the issue of needing to reauthenticate the user as he moves through different networks.For instance, can the same session for an application be maintained as a user moves from one access point/cell to a neighboringaccess point/cell, without the need to reauthenticate? Also, it raises the question of whether an identity should be associated withone user that may be using different devices (call forwarding) or with the device, which may have many different users.2On a similar note, applications should be able to migrate seamlessly across different devices. These applications are referredto as device-based services. The communication layers should be able to “reformat” the data for the different devices. An exampleof currently deployed device-based services is the ability to read email both on a personal computer and a Blackberry.Mobility also opens the doors to location-based services, which are services made available to the user depending on thecontext of their location. In a cell phone, for instance, one can get an estimate of the user’s position based on triangulation usingthe received signal strength of various cell phone base stations at the receiver and the position of the base stations. Some cellphones also come equipped with GPS to assist triangulation. The accuracy of these location services can range from kilometerswhen triangulating with very few base stations to tens of meters when using GPS-assisted triangulation. As an example of wherelocation-based services are useful, one can imagine that if the user is on a bus, he can use his cell phone to pull up a map ofwhere the bus is on its route, an estimate of when it will arrive at the bus station, and a map of where to walk to transfer to thenext bus. Another example is a service that provides automated traffic reports tailored to your position on the highway.The convenience of location-based services also brings up the issue of privacy. Systems should be designed such that theuser’s position throughout the day is not disclosed to the public.III. SECURITYSecurity is a huge issue in the internet today, with the number of different attacks mounting each day. Unfortunately, theinternet was not initially designed with security as a high priority, particularly because it was hard to foresee security implicationsin such a large and complex system. Quoting the Computer Science and Telecommunications Board and the National Academies,“In short, it appears that our nation’s dependence on NISs (Networked Information Services) has grown faster than our ability toaddress vulnerabilities.”New software tools are being developed to help detect and handle network attacks. For instance, there is ongoing researchto develop software programs that apply statistical analysis to traffic patterns to detect when a network is being attacked so thatnetwork administrators can take preventive action. One good indicator that a machine is searching for other computers to attackis when we see a lot of sync packets probing a network with very few syn-ack packets responding. Other research efforts aredirected at developing software tools to traceback the origin of
View Full Document
Unlocking...