You are here: Home Research Pipeline Forwarding

Pipeline Forwarding

by Amministratore last modified Jan 20, 2016 10:09 PM

The Problem

IP packets are today switched asynchronously with performance characteristics that are unpredictable - thus adversely affecting especially streaming media (audio/video) applications - unless network utilization is kept (very) low. Moreover, asynchronous packet switching performs poorly at bandwidth mismatch points, i.e., when switching traffic from a high capacity channel - say a backbone optical link - to a low capacity channel - say an access wireless link. Data bursts - that typically build up across asynchronous packet switched networks - tend to overload and possibly overflow buffers feeding low capacity channels. As a consequence, the quality of the service as perceived by users decreases as a result of packet drops and large delays.

The above problems of asynchronous packet switching result in (1) unacceptable trade-off between predictable performance and network utilization and (2) poor handling of bandwidth mismatch.

Possible Solution

The NetGroup has been conducting research on an enhancement to asynchronous IP - actually applicable to any packet switched technology - that leverages on a common time reference, possibly derived from UTC (Coordinated Universal Time, a.k.a. GMT) as provided by GPS or Galileo, to realize pipeline forwarding of packets. In Pipeline Forwarding (PF) all packet switches utilize a basic time period called time frame (TF). TFs are grouped into time cycles (TCs) and TCs are further grouped into super cycles, each super cycle lasts for one UTC second. TFs are partially or totally reserved for each flow during a resource reservation phase. The TC provides the periodicity of the reserved flow. This results in a periodic schedule for IP packets to be switched and forwarded, which is repeated every TC. The basic PF operation is regulated by two simple rules: (i) all packets that must be sent in TF t by a node must be in its output ports buffers at the end of TF t-1 , and (ii) a packet p transmitted in TF  t by node n must be transmitted in TF t+d by node n+1, where d is an integer constant called forwarding delay, and TF t and TF t+d are also referred to as the forwarding TF of packet p at node n and node n+1, respectively. It follows that packets are timely moved along their path and served at well defined instants at each node. Non-pipelined (i.e., non-scheduled) packets, e.g., IP best-effort packets, can be transmitted during any unused portion of a TF, whether it is not reserved or it is reserved but currently unused. Consequently, links can be fully utilized even if flows with reserved resources generate fewer packets than expected.

The pipeline forwarding guarantees that reserved real-time traffic experiences: (i) bounded end-to-end delay, (ii) delay jitter lower than two TFs, and (iii) no congestion and resulting losses. Two implementations of the pipeline forwarding were proposed: Time-Driven Switching (TDS) and Time-Driven Priority (TDP).

An overview of pipeline forwarding, its applications, and benefits:

 

Time-Driven Switching

Time-driven switching (TDS) was proposed to realize sub-lambda or Fractional Lambda Switching in highly scalable dynamic optical networking, which requires minimum optical buffers. In TDS all packets in the same TF are switched in the same way. Consequently, header processing is not required, which results in low complexity (hence high scalability) and enables optical implementation. Scheduling through a switching fabric is based on a pre-defined schedule, which enables the implementation of a simple controller. Results showed that (especially if multiple wavelength division multiplexing channels are deployed on optical links between fractional lamba switches) high link utilization can be achieved with negligible blocking using a Banyan network without speedup. TDS is suitable for very high speed (optical) backbones, where traffic can be organized in large capacity channels handled by high performance switches.

 

A description of Fractional Lambda Switching:

  • M. Baldi, Y. Ofek, "Fractional Lambda Switching," IEEE International Conference on Communications (ICC2002), Optical Networking Symposium, New York, NY, USA, Apr. 2002, pp. 2692-2696.

A prototypal implementation of a TDS switch:

Performance evaluation of Fractional Lambda Switching:

 

 

Time-Driven Priority

Time-driven priority (TDP) is suitable when more flexibility is required at the edge of the network as offered, for example, by conventional IP destination-address-based routing. TDP is a synchronous packet scheduling technique that implements UTC-based pipeline forwarding and can be combined with conventional IP routing to achieve the abovementioned flexibility. Packets entering a switch from the same input port during the same TF can be sent out from different output ports, according to the rules that drive IP packet routing. So, routing and forwarding may be based on either the conventional IP destination-address-based routing (as mentioned above), or multi-protocol label switching (MPLS), or any other technology of choice.

A complete description of time-driven priority (TDP):

A prototypal implementation of a TDP router:

Performance evaluation of Time-Driven Priority:

 


Document Actions