Broadcom Corp. has just introduced what it claims is the world’s highest performance multicore network processor architecture: the XLP980, capapble of 1 trillion operations per second.
Manufactured in a 28nm CMOS process, the XLP980 is part of the company’s new XLP900 Series optimized for deployment of network functions such as hardware acceleration, virtualization and deep packet inspection.
The 28nm chip uses 20 quad-threaded, out-of-order MIPS cores to drive throughput to 160 Gbits/s, scalable to 1.28Tbits/s using multi-chip coherency.
According to Ron Jankov, Broadcom Senior Vice President and General Manager, Processor and Wireless Infrastructure, driving the development of this kind of performance is the fact that leading service providers worldwide are rolling out high-speed 4G LTE/LTE-Advanced services to keep pace with the exponential growth in connected devices, video traffic and subscriber growth.
“To address this, equipment manufacturers are developing a new class of systems that offer the speed, quality of service (QoS), security and power efficiency required by next-generation mobile infrastructure networks,” he said. “This is driving the need for a new generation of more intelligent communications processors to deliver unparalleled performance and scalability.”
With the new network processor Broadcom is competing for this market with companies such as Cavium Networks with its MIPS based designs and Intel Corp. with its Xeon processors.
The flagship processor features end-to-end virtualization, advanced security capabilities such as deep-packet inspection (DPI), and innovative network and application intelligence technology with wire-speed networking and multi-layer QoS capabilities.
In addition to a 160Gbps application performance scalable to 1.28 terabits per second, the new processor is designed to support full end-to-end hardware virtualization support for the CPU cores, I/Os, hardware accelerators and on-chip interconnects and is capable of managing hundreds of hundreds of virtual machines (VMs), with protected memory, resources and I/Os, in either KVM (Kernel-based Virtual Machine) or QEMU (open-sourced) configurations.
Because network service providers are always susceptible to security attacks, said Jankov, the new series incorporates what the company believes is the industry’s highest 100Gbps encryption and authentication performance, scalable to 800Gbps.
“Securing these networks has become a top priority for network owners and managers,” he said, ”leading to the need to inspect, encrypt, authenticate and secure Internet traffic against increasingly complex malware and intrusion threats, all at wire speeds.”
In addition to 160Gbps of network packet processing at 128-byte packets (20 cores with 80 issues and 80 threads), other key features include: 40Gbps deep-packet inspection performance with grammar processing, scalable to 320Gbp; 20Gbps compression and decompression, scalable to 160Gbps; and RAID5/6 engines with de-duplication support.
The XLP900 Series is now sampling.