full story
Alliance Semiconductor Announces First Packet-Forwarding ICs That Enable Up to OC192 Speeds in Next-Generation Internet Core Routers
SAN JOSE, Calif.--(BUSINESS WIRE)--June 28, 1999--
Announcement Supports New Corporate Strategy to Leverage Existing
Expertise to Develop ASSPs for High-Demand Market Segments
The first silicon solution that lets Internet Router OEMs forward packets at up to OC192(1) speeds (9.6 Gbits/sec) in next-generation multi-gigabit and terabit routers is available now from Alliance Semiconductor (NASDAQ:ALSC). Router makers are under pressure from customers, including large ISPs, to move higher volumes of data more quickly over the Internet backbone. Alliance's Internet Protocol Routing Processor (IPRP(TM)) family of embedded-memory routing-engine Integrated Circuits (ICs) breaks the bottleneck that has been widely reported as a major impediment to wire speed routing of IP packets onto the Internet at OC192 speeds. Samples of Alliance's IPRP chips are available now, cost less than alternative solutions, can shrink OEM development costs and timelines, and will scale easily to future generations.
The new chips leverage Alliance's embedded memory expertise to automate the packet-forwarding "next-hop" search in silicon. Routing engines maintain and periodically update routing-table databases that they use to determine where to forward each data packet that they process. To take full advantage of the bandwidth of the Internet optical-fiber backbone, router designers need to increase the throughput bandwidth of next-generation core routers so that they can accept packet inputs from multiple edge routers, each with output bandwidths as high as 2.4 Gb/s (OC48), and place the resulting aggregation of packets onto the Internet backbone at a bandwidth as high as 9.6 Gb/s (OC192). Alliance's chips make it possible to design routing engines that can reliably accomplish packet forwarding fast enough to accommodate these rates with less engineering effort and silicon cost than it would take to design equivalent packet-forwarding circuitry in-house.
The first IC in the IPRP family, the IPRP-V4(TM), enables bandwidth capabilities in excess of 66 million packets per second, and the ability to support routing table entries from 16K to 384K. A family for IP version 6 will be announced at a later date. Samples of the IPRP-V4 family are available now with full production expected for the third quarter of 1999.
More About High-Speed Routing Issues
Increases in Internet traffic driven by the popularity of the Web in general and especially by time-critical Web applications such as audio and video streaming, conferencing and multicast, and real-time stock trading make it imperative that packets be routed with minimal delay.
One of the bottlenecks in any router is the packet-forwarding portion of the routing engine - the decision-making subsystem that determines each packet's next destination and forwards the packet to an output that corresponds to the destination address. All of this has to be accomplished along with the routine updates to the routing table that occur without stalling the routing of packets.
Routing engines currently pre-sort packet sub-net masks according to length, a process that becomes a major bottleneck as whole tables need to be pre-sorted even when the change is minor. Alliance's IPRP routing engine IC eliminates the need for pre-sorting and allows individual entries, rather than entire tables, to be updated, reducing the time spent on non-forwarding activities.
The other part of the bottleneck arises from the time it takes to match incoming packets' destination addresses with entries in the routing table. Table entries are not all the same length, and the routing engine is required to select the longest table entry that matches the packet's destination address. These operations typically require several clock cycle times to complete. Using Alliance's search algorithm, the search and forwarding operation requires only a single clock cycle, yielding 66 million packets (representing 30 Gbits for a central look-up engine-based architecture) per second.
"Other networking IC vendors are talking about achieving OC-192 data rates. Alliance is one of the first standard solution providers delivering an IP forwarding engine that enables OC-192 today," said Bhanu Nanduri, director of marketing, embedded products, for Alliance Semiconductor. "By deploying the IPRP-V4 devices to determine next-hop addresses in high-speed routing engines, system designers can greatly simplify router design complexity, resulting in fewer components per system, less overall power consumption and more reliability, providing our customers with a high-performance solution that offers significant cost and time-to-market benefits."
According to Ryan Hankin Kent (RHK), a leading market research and consulting firm specializing in the analysis of advanced technologies for the public telecommunications network, the market for next-generation multi-gigabit and terabit routers is expected to grow to $5.53 billion by 2003, with substantial growth expected in the next millennium as the Internet infrastructure continues to upgrade to address increased data-handling demands.
"Higher core and edge bandwidth requirements demand a more efficient approach to router table management and memory searches," said Tracey Vanik, director of edge switching and routing for RHK. "Off the shelf routing components that reduce the number of memory accesses required to resolve address lookups are attractive alternatives to costly in-house design. These components will be of interest to companies involved in developing gigabit and terabit routers."
Embedded memory networking products such as the IPRP-V4(TM) are application specific standard products (ASSPs) and represent a new strategic business thrust for Alliance Semiconductor. The strategy takes advantage of the company's traditional core competencies in memory design and manufacturing, and joins that with carefully acquired expertise in networking to offer an array of solutions for new generations of high-end data communications equipment.
"Alliance's entry into the networking IC market is a further extension of our established expertise in the development of embedded memory technologies," said N.D. Reddy, chairman, president, and chief executive officer for Alliance Semiconductor. "The launch of the IPRP family leverages our industry-recognized expertise in memory technology, design talent in networking products, and our strategic joint venture manufacturing relationships, to deliver a world-class product that exceeds customer demands for quality and value."
About the IPRP-V4 Family
Alliance Semiconductor's IPRP-V4 is an Internet Protocol Routing Processor family that enables wire speed routing of IPV4 packets. The ASN64132, ASN32132 and ASN16132 are pin and function compatible; they differ in the size of the database they support. The ASN64132 has a built-in 64K-entry database; the ASN32132 and the ASN16132 support 32K and 16K entries, respectively.
When operating at a frequency of 66 MHz, the ASNxx132 can provide system architects with sufficient bandwidth to design wire-speed multi-port gigabit and terabit routers. The ASNxx132 is able to search its entire database and provide the Longest Prefix Match (CIDR Protocol) search result once every 15 ns, implying that this device can enable a router to route up to 66 million packets every second. Up to six ASNxx132 devices can be depth cascaded with no glue logic, resulting in a database capable of supporting up to 384K IPV4 addresses. Up to eight devices can be cascaded when operating at lower frequencies.
Pricing and Delivery
The ASN64132 will be priced from $150, the ASN32132 from $75, and the ASN16132 from $50, all for 10,000-piece quantities. Samples of the ASN64132 are available now, with production volumes available in the third quarter of 1999. Samples and production of the ASN32132 and ASN16132 will be in the third quarter of 1999.
|
