RAS (Routing As a Service) [31] takes a different approach from the mentioned proposals so far
and suggests resolving the user-provider tussle by delegating the control of end-to-end route com- putation and forwarding infrastructure for customised routes to third parties. The third parties obtain a global view of the Internet topology through business relationships with ISPs along the paths. They buy virtual links from several ASes to connect the virtual routers (VRs) across multiple domains. Although the ASes have control over the traffic flow between VRs, the third party can set the forwarding state of them. In this way, ISPs have the control to engineer their network and cus- tomers do not need to negotiate with providers directly and can deal with Routing Service Providers (RSPs) instead. The benefit for ISPs will be the control over the number and size of the Virtual Links (VLs) they sell to RSPs. On the other hand, users can enjoy the performance they require without going through the hassle of dealing with individual ASes along the end-to-end path. A few competing RSPs can then provide a range of services based on choice for the user. These services include avoiding undesirable ASes, blocking unwanted traffic, and guaranteed QoS.
Figure 2.5. The Main Components of the RAS Architecture [31]
The Forwarding Infrastructure (FI), One or More Routing Service Providers (RSPs), and RAS Clients In a similar effort, Path Brokering [32] suggests the use of third party entities that act as retail- ers of end-to-end paths. Path brokers have a global view of Internet topology and the ability to compute user-desired paths and enforce them via MPLS. They also get a hold of payment mecha- nisms between the end-users and ISPs so that they do not have to negotiate directly. Path brokers in this sense act as middlemen who get transit offering data (including performance and cost) in a standard format like XML from the providers and receive path queries with constraint attributes from the end-users. Upon receiving path queries form the end-users, they compute suitable paths based on the defined path constraints and reply with some path offers that are signed with the bro- ker’s private key. Then, after the end-user has confirmed the choice of path to the broker, the broker
will forward a transit request to the operator with the start and end time for the transit service re- quested by the end-user. Once the provider approves the request, transit labels will be sent to the broker and ultimately to the end-user to label its packets across the network.
Figure 2.6. Path Query Steps in Path Brokering [32]
A number of studies highlight the importance of a payment mechanism for successful adoption of source controlled routing approaches by the ISPs.
The authors of [33] question the conventional wisdom which believes that global topological maps are needed for ETE path computation in source routing and suggest that end systems can ob- tain such information for relevant parts of the network through ETE measurements done during data transfer or using probe packets. Also, providers can disseminate statistical routing information computed over a longer period of time to path computing entities (e.g. 3rd party path providers).
The authors suggest the use of 3rd party path providers to give Interdomain ETE paths or path segments to the end-systems after considering the end-user multiple constraints (e.g. QoS level, disjoint path, etc.) in heuristic path computation algorithms. They also mention some techniques like path caching, proactive path computation, and cooperative path computation by multiple path providers as methods to improve path computational cost and latency. However, the above two scalability issues as well as the cost of information dissemination for a given topology and traffic matrix is not known yet.
Laskowski et al. [20] investigate the validity of the end-user empowerment idea by exploring various combinations of routing (either ISP or source controlled) and contracting (access or route or hop pricing) approaches. Their analysis shows that a mere user-controlled routing does not im- prove the overall network performance. Instead, exploiting this technology in conjunction with a system of hop prices (as an overlay architecture) will restore the competition in ISPs market and keeps the Internet architecture evolvable.
A study [34] analyses the effect of user-directed routing on three important objectives of ISPs, which are network control, privacy, and profits. Based on these, the authors challenge the dominant wisdom that user-directed routing weakens providers’ control over their network. They argue that a flexible payment system allows the ISPs to engineer traffic in their networks. The ISPs can set
prices for individual routes in the network and engineer their traffic just as they do using link costs. However, the privacy of ISPs and their internal workings is not guaranteed in user-directed routing as a result of more transparency. This has a number of disadvantages e.g. facilitating the observa- tion of competitors into one’s internal network as well as increasing vulnerability to their possibly destructive traffic manipulation tactics. They suggest that regulation can be a way to mitigate this shortcoming in the future.
Yang [35] stresses that the blockage of certain applications (e.g. BitTorrent) and traffic differen- tiation are signs of bandwidth scarcity. The paper proposes that ISPs auction their bandwidth and let end-users place bids in packet headers to value their communications. At congestion times, ISPs use this bidding information to serve the most valuable packets and price the senders according to the auction model. To help the users in the bidding process, each application may have a default bid value, which can be changed by the user. In addition, a software agent assists the user in translation of prices to bid values.
Araujo et al. [36] takes a similar approach to exploit congestion pricing principles and enable the providers manage their traffic in an end-user controlled multipath routing environment by charging the user traffic based on the congestion it causes. The authors suggest the use of path- fragment dissemination algorithms like Pathlet Routing to develop a widespread congestion-aware multipath routing architecture that provides the total congestion cost of end-to-end routes for the source nodes.
Finally, [37] presents several security defence mechanisms for user controlled routing. The main technique uses lightweight cryptographic constraints on forwarding entries, which prevents a range of attacks including eavesdropping, loops, and traffic amplification. The other technique uses a 3- way handshake against any forwarding entry to prevent an attacker from inserting entries that are pointing to other end-users. The third technique prohibits the replication of traffic by the attacker. The link capacity is used as a limiting factor that controls the amount of traffic that can be generat- ed by a source. It also ensures that the rate of packet loss for each link in the network topology does not exceed to a defined value.