Search /
Docfinder:
Advanced search  |  Help  |  Site map
RESEARCH CENTERS
SITE RESOURCES
Click for Layer 8! No, really, click NOW!
Networking for Small Business
TODAY'S NEWS
Google's Android did not infringe Oracle patents, jury finds
VMware acquires desktop management company Wanova
Groups launch gigabit-per-second broadband project
Windows 8 touchscreen devices to be priced higher, Dell says
Google warns users infected with DNSChanger as Web outage nears
SAP buying e-commerce vendor Ariba for $4.3 billion
Brocade outlines its SDN approach
Managing Mobile Mania
Jailbreak of Apple iOS 5.1.1 due 'in days'
Google has finally closed Motorola Mobility acquisition
Privacy advocates fear CISPA
10 years later, Alcatel-Lucent revisits Cisco and Juniper in the core
Avaya lays out roadmap for unified software management
Sidecar app for iOS, Android seeks to give smartphone voice calls overdue respect
Medical firm avoids Exchange nightmare with outside help
Service Provider Networks / (none) /

Juniper's next-gen revealed?

Competitive document says 320G system with terabit scale, packet sequencing fix.

Related linksToday's breaking news
Send to a friendFeedback

Juniper's next-generation core router is a 320G bit/sec eight-slot, half-rack chassis that can be linked with others via separate switch matrix to achieve terabit-level scale, according to a competitive assessment memo from a Juniper rival.

The Juniper router, which was referred to as "Gibson" in the competitor's memo, will also attempt to alleviate the packet misordering situation that occurs on OC-192 interfaces in the company's current high-end offering, the M160. The competitor requested anonymity.

A Juniper spokesman would neither confirm nor deny the contents of the memo.

Gibson would bring Juniper up to par with rival Cisco, which shipped a 320G bit/sec router earlier this year.

Gibson will support 40G bit/sec of bandwidth per slot and 32 interfaces at 10G bit/sec per chassis, the memo states. Twelve Gibsons will be able to interconnect through a common switching matrix.

"They (Juniper) are telling customers that they will be able to perform demonstrations of this functionality at their headquarters during Q4 '01," the memo states.

Gibson uses a multi-stage network for its switch fabric architecture. Each line-card chassis has five switch fabrics as part of the standard system configuration.

Juniper is claiming that six line-card chassis, for example, can be connected using three switch-card chassis to provide 1.44 terabit/sec of SONET bandwidth, the memo states.

"They are telling customers that 'the (switch chassis) can be thought of as a cheap POP's inter-connect device replacing some of the expensive ATM and Gigabit Ethernet switches found in the middle of a router mesh,'" the memo states.

Approximately 3.2 switch fabrics of capacity are required for full, non-blocking, system operation, the memo states. Fabrics are hot-swappable in the case of failures, and in normal operation four fabrics are used and the fifth is in a standby mode.

"Juniper claims that no upgrades are necessary to the line card chassis to allow participation in a multi-chassis system other than new switch fabrics and the addition of a central switch card chassis," the memo states.

The initial multi-chassis system, planned for the second half of 2002, will connect four Gibson line-card chassis, the memo states. Future versions of the switch-card chassis will enable larger multi-chassis systems, the memo states.

The competitor was not surprised with Juniper's Gibson strategy.

"The only thing I see new is Juniper will be the first of the two viable core router vendors to bring out this (multi-stage) capability," a vice president at the company says. "Anyone that goes beyond 160G bit/sec or 320G bit/sec, they've got to go to a multi-stage distributed architecture. I don't see anything new to the industry in terms of what they're doing."

To alleviate packet misordering, Juniper is adding "per-flow traffic distribution" across Gibson's aggregated SONET links. Per-flow link assignment prevents packet reordering within a flow, the memo states.

"This prevents issues of circuits within a bond group having different propagation delay characteristics, due to divergent fiber paths, from arising," the memo states.

Gibson will support up to 16 trunks of aggregated links with up to eight equal bandwidth links per trunk, the memo states.

If a link within a trunk fails, flows mapped to that link are automatically assigned to another link.

Gibson will also feature 20G bit/sec distributed packet-forwarding ASICs called "Gimlets," according to the memo. Each slot in a Gibson chassis can contain up to two Gimlet ASICs.

Each Gimlet ASIC has its own hardware route lookup engine and delay bandwidth buffer. Incoming unicast packets are handled by two Gimlets -- one on ingress to the router and one on egress -- and multicast packets are sent to as many Gimlets as necessary, based on the number of multicast destinations, according to the memo.

The forwarding table is derived from the routing table and contains an index of IP prefixes and MPLS labels that are actively used to associate a prefix or label with an outgoing interface. Gimlet uses the contents of the forwarding table, not the routing table, to make its ultimate forwarding decision, the memo states.

Line-rate performance is not possible for all cases, as filters have varying impacts on performance, the memo states. Performance drops gradually from line-rate as filters become more complex.

Egress interface congestion is handled via outbound queuing. Each Gimlet has 100 milliseconds of delay bandwidth buffer that is allocated among interfaces.

In the event of congestion, packets are dropped using a weighted random early discard algorithm, according to drop profiles that have been configured for that interface. By default, there are four queues associated with each outbound interface.

More queues can be associated with a particular interface and drop profiles can be associated with each queue, with differing drop profiles associated with various types of traffic, the memo states.

Gibson will run current versions of Juniper's operating system, JUNOS 5.X, the memo states. Routing protocols, interface management, chassis management, SNMP management, system security, and the user interface all interact with the operating system as subsystems.

Each program executes as an independent process, complete with its own memory protection, according to the memo.

"This removes most opportunities for runaway applications to corrupt each other or the kernel," the memo states.

Target first customer ship (FCS) with JUNOS 5.1 and production hardware as a standalone router is slated for this quarter, the memo states. At FCS, Juniper will support a single-port OC-192c module, and "shortly thereafter," Juniper will release a four-port OC-48c module.

Channelized OC-192 packet-over-SONET modules will support one level of de-multiplexing down to 4 x OC-48c, the memo states.

Support of existing M160 single-port OC-48c modules is also planned at Gibson FCS, as is a four-port OC-12c packet-over-SONET module and a dual-port Gigabit Ethernet card.

An OC-192c long-reach optics module is slated for the first quarter of 2002, the memo states. An OC-192c card featuring very short-reach optics is planned for the second quarter of 2002.

Also on tap for the second quarter of 2002, according to the memo: a 10G bit/sec Ethernet module; a four-port Gigabit Ethernet module; and a packet-over-SONET OC-48c module with long-reach optics.

A four-slot OC-768c module is also planned, the memo states. It will support one level of de-multiplexing, down to OC-192.

Channelized OC-768 packet-over-SONET will support one level of demuxing down to 4 x OC-192c, the memo states.

RELATED LINKS

NWFusion offers more than 40 FREE technology-specific email newsletters in key network technology areas such as NSM, VPNs, Convergence, Security and more.
Click here to sign up!
New Event - WANs: Optimizing Your Network Now.
Hear from the experts about the innovations that are already starting to shake up the WAN world. Free Network World Technology Tour and Expo in Dallas, San Francisco, Washington DC, and New York.
Attend FREE
Your FREE Network World subscription will also include breaking news and information on wireless, storage, infrastructure, carriers and SPs, enterprise applications, videoconferencing, plus product reviews, technology insiders, management surveys and technology updates - GET IT NOW.