Cisco Prepping Monster IOS Upgrade

Cisco Systems Inc. (Nasdaq: CSCO) is gearing up to announce some major upgrades to its internetworking operating system (IOS) software in the next couple of weeks, according to sources who have been briefed on the project.

Cisco refused to comment, saying that they do not discuss unannounced products or speculate on rumors.

While new versions of IOS are released almost daily for one Cisco product or another, this across-the-board update is particularly noteworthy because it attempts to tackle the problem of IP resiliency across all networks, ranging from enterprise networks to the carrier core.

Although large-scale improvements to IOS would be a major step for Cisco, it still does not represent a completely new version of the software -- and the company has apparently put off building such a product. For years, network engineers have urged Cisco to start from scratch and build a new IOS that is easier to implement and update. But such a project is daunting, especially in the face of an industry downturn and budget cuts.

“IOS is difficult to overhaul, because it is a moving target,” says Frank Dzubeck, president of Communications Network Architects (no Web site). Dzubeck declined to comment on the specifics of forthcoming IOS upgrades, but he noted that Cisco must adapt its software. “The industry is moving so quickly that they need to build something that is flexible. And that isn’t so easy.”

Instead, Cisco is trying to make the existing IOS software so resilient that packets will never be dropped from Cisco routers, even when hardware fails or lines go down, according to sources familiar with the plans. This represents the first major upgrade to IOS in a quite some time.

Three of the new resiliency upgrades will be available in June of this year. The first enhancement is designed to make routing more resilient by offering non-stop forwarding of BGP (border gateway protocol), a major routing protocol. This means that if a failure occurs on the hardware or in the routing protocol, packets will continue to be forwarded while the problem is being resolved.

The second piece will enhance the resiliency of the link layer. This means the routing device will remain stateful, or maintain persistent connections, allowing security features such as a VPN (virtual private networking) tunnel to continue to operate during network failures.

The last major piece to be released in June has to do with improving the reroute time of Multiprotocol Label Switching (MPLS). When a routing connection is disrupted and MPLS tunnels are affected, this new software feature will allow for MPLS to recover much more quickly, according to Cisco. Other resiliency improvements will be added in the second half of this year, say sources.

The net effect of these improvements is more uptime for service providers, less downtime for customers. Vendors can claim their products have 99.999 percent reliability, but hardware fails, fibers can be cut, and software bugs exist. In networking, 100 percent reliability isn’t possible. But by improving resiliency, end users never have to notice when a failure occurs, because traffic will be quickly rerouted and statefulness will be maintained.

Other routing vendors have already begun to touch on this notion of zero packet loss. Alcatel SA (NYSE: ALA; Paris: CGEP:PA) recently announced nonstop routing for its 7770 core router (see Alcatel Unveils New Routing Technology). Juniper Networks Inc. (Nasdaq: JNPR) mentioned improved resiliency of its new ASIC set when it introduced the T640 (see Juniper Goes Terabit With the T640). And Unisphere Networks Inc. honed in on statefulness when it announced the MRX platform in January of this year (see Unisphere Cutting the Edge). While other companies have implemented some resiliency, analysts say Cisco is the first to address it so comprehensively, reinforcing its leadership in IP networking.

But the big question is: When will service providers actually implement the upgrade? While enterprise users tend to implement new revolutions of software quickly, service providers usually lag behind, because they are often dealing with many more routers and much more complicated networks. Service providers often have to extensively test new software versions before they are able to deploy them.

This lag-time in upgrades among service providers was thrown into relief last week, when UUNet, owned by WorldCom Inc. (Nasdaq: WCOM), suffered a major outage due to a bug affecting its Cisco routers (see WorldCom's IP Outages: Whodunnit?). Cisco claimed that the bug had been discovered and fixed in the latest version of its IOS, but Uunet was running an older version of the software that did not have the patch.

— Marguerite Reardon, Senior Editor, Light Reading
http://www.lightreading.com