游客发表

RER60F11R5RCSL

发帖时间:2021-06-15 01:47:38

Following synthesis, power-aware mapping techniques may be used to optimize the netlist. These techniques include mapping highly active nodes into specific cells and mapping highly active input signals onto low capacitance input pins. When partitioning the design into multiple voltage (VDD ) domains, appropriate level shifter elements need to be inserted into the netlist to connect logic elements across multiple domains. Furthermore, signals to and from domains that may be switched on and off require special attention so as to avoid any floating net” problems.

At its founding, SPIRIT comprised the following companies:

The consortium expects to have a pilot-program-validated proposal for the standards by the end of 2003, with a general release sometime in early 2004.

RER60F11R5RCSL

With the amount of traffic growing on the net and the amount of information in a packet significantly increasing, existing lookup table architectures are clearly a bottleneck in today's system designs. The situation will only get worse as IPv6 rolls out and line rate increase.

To address this issue, designers must re-invent their lookup table architectures in order to accelerate lookups, improve scalability, and support updates/enhancements without adversely impacting overall system performance.

Several competing approaches to meeting the lookup requirements are available, ranging from simple trie-based algorithms to more complex hardware-optimized tries. While effective, these solutions do not deliver the performance required for today's networking system designs.

RER60F11R5RCSL

Fortunately, designers may find the answers they need in ternary content addressable memories (TCAMs). A new generation of TCAMs are on the way that deliver higher density, lower power, and lower cost, thus making them a viable replacement for traditional trie-based approaches. Let's see how.

Trie Climbing Considerable research has been conducted in the area of trie-based lookups. With the availability of higher density SRAM and DRAM devices, tries offer the promise of low cost, low power lookup capability. However, straightforward and well-understood algorithms such as binary and Patricia tries do fail to address current performance and density requirements and are giving way to more complex and efficient multi-way and level-compressed tries.

RER60F11R5RCSL

Table A compares alternative memory-based lookup implementations. In this table, space complexity is a function of the total memory requirement. Search complexity refers to the number of clock cycles required to perform a single search and update complexity refers to the number of clock cycles required to update a single entry in the table. System designers are increasingly able to take advantage of larger SRAM and DRAM chips but the lack of performance and non-deterministic nature of these implementations continue to limit their adoption.

Multi-Dimensional Lookups While IPv4 forwarding, in the simple cases discussed above, is suitable for packet-by-packet routers, many vendors are now developing flow-aware routers. As a result, more advanced multi-dimension lookups are needed.

ISPs are demanding the ability to provide differentiated levels of service to their customers. To provide this capability, vendors must provide the ability to not only forward the packet but to classify the packet. The required mechanisms include admission control, metering and shaping, policing, queue management and scheduling. These high level functions increase the number of lookups required per packet and force system designers to push the limits of current lookup schemes.

Take the case where there is a requirement to implement a set of rules R (R1, R2, ….) and the set of corresponding actions A (A1, A2, A3, ….), then the classification problems can be defined as:

if[(Ri=match) and (Rj=match) and …] than do Anfor any combination of i,j, and n.

The classification problem can be looked at as a combination of multiple, simple match problems. Not surprisingly, many classification schemes are based on well-known and understood search and match solutions, although some modifications or extensions are required.

热门排行

友情链接