|
 |
 networkZONE Products for the week of December 17, 2007
Lightstorm Networks Says…
Hudson OAM Service Accelerator Boosts Capacity and Scalability of Brooklyn Carrier Grade Ethernet Switch
FPGA-based accelerator enables highly scalable Ethernet CCM processing for PBB, PBT and T-MPLS networks
Lightstorm Networks has announced that its latest carrier Ethernet solution, the Hudson OAM service accelerator, is now sampling. Following on the heels of the Brooklyn-10, Lightstorm’s carrier-grade 20Gb Layer 2 switch, Hudson significantly enhances Lightstorm’s Operations, Administration, and Maintenance (OAM) capability. Supporting both IEEE 802.1ag and ITU Y.1731 connectivity fault management (CFM) protocols, Hudson enables Brooklyn-10 based systems to process continuity check messages (CCMs) on thousands of services without the need for software intervention, which is common in many of the current solutions in the market today. This scalability is a critical element for deploying robust carrier Ethernet solutions in Provider Backbone Bridging (PBB), Provider Backbone Transport (PBT), and Transport-MPLS networks.
As networks are turning towards Ethernet-centric solutions and replacing the existing SONET/SDH and ATM/Frame Relay infrastructures, one of the primary hurdles developers are facing is providing the same or better OAM functionality. The success of carrier Ethernet depends greatly on its ability to provide service providers with the same management framework and tools that they have used with these previous technologies. Critical to that expectation is “SONET like” restoration times on services in less than 50ms. To ensure that level of conformance the network must track services health by monitoring CCMs; which could be required as often as every 3.3ms.
Next generation carrier Ethernet systems are expected to manage anywhere from 1,000 to 8,000 services which is simply not possible if monitoring of CCMs is performed in only control plane software. By offloading the filtering and processing of CCMs using the Brooklyn-10 and the Hudson service accelerator, system OEMs can return the CPU to doing exception handling and other processing functions while offering the robust protection and restoration services demanded by carrier networks.
“Up until this point, the acceptance of carrier Ethernet in communications frameworks has been limited by the lack of hardware designed to support both Ethernet OAM and maximize scalability,” said Stephen Christo, director of product marketing for Lightstorm Networks. “What we are doing is truly taking Ethernet OAM to the next level. The Hudson device effectively removes the burden of analyzing and generating continuity check messages from software, thereby scaling the number of connections while minimizing CPU utilization. As a result, the customer gains better performance and a significantly enhanced user experience.”
The Hudson device comes just three months after Lightstorm Networks officially launched and introduced the Brooklyn-10, the first Ethernet switch Application Specific Standard Product (ASSP) purpose-built for the carrier market. An industry first, the Brooklyn-10 offers the latest carrier Ethernet services at wire-rate speed - without the requirement of complex NPU or FPGA programming. The company, rapidly moving forward with its innovative suite of silicon software and hardware products, is leading the carrier Ethernet surge and meeting the ever-changing and demanding needs of today’s carrier.
EN-Genius Says…
I have had my doubts as to whether Ethernet could ever deliver the reliability and manageability it would take to displace TDM technology in carrier-grade access and transport applications, but some of the emerging standards and the emergence of silicon from vendors like Lightstorm are starting to make a believer out of me. Lightstorm’s Hudson Service Accelerator should help lower the cost of carrier-grade Ethernet by greatly expanding the number of connections that their Brooklyn L2 Ethernet switch (reviewed here October 2007) and its host processor can monitor, diagnose, and control using the First Mile Ethernet (802.3ah) standard in-band signaling standard which uses the IEEE 802.1ag and ITU Y.1731 protocols.
 Implemented on a low-cost (Cyclone or Spartan-class) FPGA, Hudson shares the PCI bus with the Brooklyn switch and its host processor. This allows it to easily check the messages to see that they are marked for the proper stream, verify packet timing, and other tasks that would normally consume a host system’s compute resources with a relatively small number of connections.
Instead of making the host look at every CFM packet, Hudson snoops the PCI bus and decides whether a particular packet should be brought to the host’s attention. In these cases, Hudson uses a DMA transfer to move the packet to the host’s memory, issues an interrupt and gives the processor a pointer to the packet. On-chip counters keep track of lost or dropped CCMs and alert the host if they exceed the threshold. While Hudson’s on-chip storage can support the parameters for a reasonably large connection table, you also have the option of using external SRAM if your requirements dictate. The accelerator also specifies the message and timing interval for heartbeat messages, and content of any error messages.
According to Lightstorm, hardware acceleration of CFM processing allows a reasonably-sized host processor (typically a PowerPC) to support close to 10x more connections (on the order of 1000 vs. 100) than if it were handling the processing itself. A hardware-based approach to monitoring connections also makes it much easier to support a SONET-like 50 ms failover time that carrier customers have come to expect from carrier-grade services.
It’s very interesting to see Lightstorm join the small band of silicon manufacturers using FPGAs to roll out a standard product (even though I suspect that many implementations of Hudson will have semi-custom feature sets). It’s too early to tell whether there is a viable business model for using programmable logic as an alternative to a CMOS foundry for producing complex merchant silicon but it is an excellent fit in this application for several reasons. First and foremost, it saves the large tooling costs associated with an ASSP whose functionality will quickly be folded into the next generation of Lightstorm’s products. Putting these functions into an FPGA also means that the accelerator can easily be modified to track emerging standards or support custom capabilities.
The Hudson service accelerator is now sampling and will be generally available in Q1 2008. Hudson will be supplied as a fully packaged bit file targeted at specific low cost FPGAs using a per system charge. Adding Hudson to a Brooklyn-10 enabled platform will increase the solution cost from $10 - $40, depending on volumes.
|
|
|
|
|
