Serial RapidIO and 10 Gigabit Ethernet

The explosive growth of multimedia content in mobile applications has spawned yet another evolution of wireless networks. In order to keep up with the increasing demand for voice and data traffic, high-performance systems must be developed. To stay ahead in such a highly competitive market, system designers must choose a scalable interconnect standard to provide better performance, cost, power consumption, and reliability. Serial RapidIO (S-RIO) is the only protocol that can meet the specific needs of wireless baseband processing applications, but 10 Gigabit Ethernet (10GE) is sometimes seen as an alternative solution.

To understand the difference between 10GE and S-RIO, you must first understand the history of the development of these two standards. The Ethernet standard is designed for large-scale local area networks or wide area networks. The premise is that there are powerful processors on the network that can strengthen the software protocol stack. Transactions on these kinds of networks usually actually contain a large amount of data communication traffic, and have a high data carrying capacity. However, wireless baseband processing has different requirements. In wireless systems, high-performance processing components both carry control plane and data plane traffic, and need to communicate quickly and reliably with each other. Serial RapidIO is specifically designed to support this type of demand.

Performance is important

3G + and 4G wireless protocols and advanced antenna system (AAS) architectures require powerful and predictable system performance. To meet this demand, special attention has been given to latency and protocol efficiency in the RapidIO standard. In order to minimize latency, RapidIO is designed with a short packet header and a simple routing architecture based on target ID (DesTID). This simple transport layer function achieves a through switch delay within 100 ns. RapidIO provides the function of "stomp" the data packet in transit, so that the switch device can forward it without receiving the entire data packet ("stomp" is a control symbol used in the RapidIO protocol to encounter In the event of an error, the partially transmitted data packet is cancelled). This function improves the overall performance of the system without affecting reliability. If an error is detected in the forwarded data packet, the RapidIO port can perform "stomp" processing on the transmitted data packet without software intervention.

In contrast, the header of a 10GE packet is much longer, which results in increased latency in communicating with other processing components. For LANs and WANs, the longer packet length can compensate for the longer header length and related overhead. However, compared with the traffic in the LAN / WAN model, the traffic in wireless applications is usually shorter in length and larger in number. For this kind of traffic with less load, the longer packet header in the 10GE standard will adversely affect the efficiency of each transaction and reduce the overall performance of the system. The delay of the layer 2 switch in the 10GE standard is also much larger than the delay in S-RIO, and in the best case it is far more than 200ns. Ethernet cannot perform "stomp" processing on data packets in transit. Instead, it will “try its best” to deliver the data packet and rely on layer 3 or 4 software to verify the integrity of the data packet content.

Reliability without compromising performance

Reliability is the flash point of RapidIO. The acknowledgment design built into the hardware prevents data packet loss and guarantees the transmission of data packets. The mechanism to ensure reliable data packet transmission is implemented at the physical layer in pure hardware. Implementing the transmission of data packets in hardware has a significant positive impact on the performance of the system. Because reliable transmission can be achieved without software intervention, transactions will not introduce delays due to software routines. For the 10GE standard, the physical layer only does its best to manage the transmission of data packets (that is, there may be packet loss). To ensure the successful transmission of data packets, it needs to be handled by the TCP layer. In the TCP layer, a reliable transmission mechanism may be implemented in software.

Accessing the software layer takes valuable time, consumes processing resources, and is detrimental to system performance. This software intervention in the 10GE design may introduce a system delay of more than 10μs. In contrast, the system delay of S-RIO is only about 1μs. The retransmission of data packets in the 10GE system is more complicated and may consume dozens of subtleties. For the S-RIO design, data packet retransmission is transparent and completely handled by the hardware, which can be completed in less than 1μs. At the same time, relying on software to ensure the transmission of data packets will cause uncertain system delays. Depending on the software routine that was running when the packet arrived, the time required to complete the packet transfer operation is somewhat unpredictable. 10GE relies on software to ensure the transmission of data packets, so it is not a good choice for systems that want to obtain a short and deterministic delay.

Consider system costs, not just equipment costs

In order to run the software used to implement the protocol stack, 10GE requires a processor. This management of the software stack increases processing overhead and reduces system efficiency. Ethernet may consume 15% to 30% of processing performance for software stack management. For a $ 100 processor, this is equivalent to an invisible system interconnect cost of $ 15-30 per processor. In the RapidIO system, the protocol will minimize the dependence on software, thereby reducing the load on the processor, and at the same time the saved expenses can be used on a faster processor, thereby achieving a higher performance system at a lower cost.

System power consumption

Since the processor load required to process the RapidIO protocol is reduced, the system power consumption will also be reduced. Because the use of multi-gigahertz processors for protocol management is reduced, the power consumption of the system can be kept to a minimum. Therefore, it can also reduce thermal management costs and reduce system complexity. In 10GE systems, the use of more processing resources will result in higher system power consumption.

Scalability provides a competitive advantage

Competition in the wireless infrastructure market is fierce. Because vendors are competing with each other to provide more users on each line card, it is crucial to build a scalable system. System designers must design an architecture that allows them to adapt to changes in performance with simple modifications. The RapidIO system uses the same register set and can easily expand from the lowest rate to the highest rate. Available port rates are 1, 2, 2.5, 4, 5, 8, 10, 16, and 20Gb / s. On the contrary, in order to upgrade Ethernet from 1 Gb rate to 10 Gb bandwidth, a new set of registers is required, and a large number of system software changes are required. Another important component missing from the 10GE ecosystem is the small port count switch. Most devices have to deal with a large number of backplanes or aggregation devices with many 1GE ports but only two 10GE ports.

The advantages of S-RIO include: low latency and certainty, low system processor load, high reliability, and low coupling between the processor and protocol management software. These advantages make S-RIO the default protocol choice for wireless applications. Although 10GE is a viable option for wireless baseband applications, it is usually abandoned by designers because of its own disadvantages. By using RapidIO, designers can build scalable systems to maximize performance while minimizing power consumption and costs, and can shorten completion time in a highly competitive market.

Multi-Layer Circuit boards. In order to increase the area that can be wired, the multilayer board uses more single or double-sided wiring boards. Using a double-sided inner layer and two single-sided outer layers, the printed circuit board which is alternately arranged by the positioning system and the insulating bonding material and the conductive pattern is interconnected according to design requirements becomes four layers, similiar stack up for 6 layer, 8 layers etc. The layers are more than two called multiple layers printed circuit board. The number of layers of the board does not mean that there are several layers of independent wiring layers. In special cases, empty layers are added to control the thickness of the board. Usually the number of layers is even and the outermost two layers are included.

Multi Layer PCB

Multilayer Pcb,Multilayer Printed Pcb,Multilayer Printed Circuit Board ,Multi-Layer Circuit Board

Chuangying Electronics Co.,Ltd ,