I feel like there isn't a large disparity between Gigabit Ethernet and 10 Gigabit Ethernet. Gigabit Ethernet has moved along quite quickly in its maturity curve in the modular switching area, representing a large number of the switches shipping today. I'd say there may be a lack of knowledge about how far along 10 GigE technology has come, and how ready it is for deployment today. As GigE deployments increased in the past few years, companies have moved multiple Gigabit Ethernet links from the wiring close to the core. By moving to 10 GigE, they can reduce those trunks back to a single connection. Are there really significant benefits in doing that?
Sure -- easier manageability, since you're better off running virtual links to a single link, simplifying network operations. It also increases the throughput from the access layer to the core distribution. Many networking pros still believe Gigabit Ethernet deployments are about "future-proofing" networks when the opportunity arises, as opposed to upgrading specifically for performance benefits. Is it a constant challenge to offer counter-evangelism on that front?
With chassis-based systems, we've seen a lot of the interest in those systems from customers who are specifically interested in future-proofing. So I wouldn't see that as a disadvantage but as an advantage. The architecture for the Catalyst 4500 products has all the intelligence centralized, so the line card can be used generation after generation. So that's something we're actually promoting -- since the cost of labor in terms of reconnecting and disconnecting these switches is pretty tough -- so in that regard future-proofing is a good idea.
I think what you'll see first is 10 GigE moving from the core to the distribution layer, and from distribution to the wiring closet. That will largely be a fiber-based transition, from fiber to 10 GigE fiber. There are technologies emerging to run 10 GigE over copper, and that I think where you'll see that type of port deployed first is the data center. We've seen the transition beginning, starting several years back with campus interconnect efforts, and the copper 10 GigE core distribution connections are probably a few years away from mass adoption. Do you see 10 GigE over copper being driven by need, cost or a combination?
Ten GigE cost curve for copper-based connections over twisted pair will follow a slower timeline than we saw with GigE. There are power consumption challenges, such as the power needed to drive copper interconnect over 10 GigE. It will take several generations of silicon to achieve the density that'll be appropriate for the desktop, but prior to that you'll see strong interest in 10 GigE the data center. What's Cisco's stance on cabling regarding Category 5 vs. Category 5e? Obviously, Cat5 is widely deployed and companies have an investment in it, yet there are performance gains to be had with Cat5e.
For new deployments, customers are future-proofing their cable plans, moving to Cat5e. Our equipment for GigE will run fine forever with Cat5 cable. As new capabilities emerge, such as 10 GigE, there emerges the prospect of operating over longer distance, perhaps longer than the standard cable, so I think that for new installations, Cat5e or higher makes sense. But I don't think there's a need to retrofit existing cable to make that migration now. At this point, is there any excuse for hanging onto ATM systems in the enterprise, or does it still have advantages over Ethernet implementations?
I think the only reason for hanging on would be the time it takes to execute on a migration. Organizations have for the most part been upgrading to and deploying GigE, and moving forward. Though I think one of the areas where ATM had advantages over early Ethernet implementations was in the area of quality of service, the administration the Layer 3 intelligence Ethernet switching technology that has been incorporated into Ethernet switches. That was one of the last remaining barriers. Even in service provider networks -- where ATM is very strong -- there's been a large movement to Ethernet switches, and with our Catalyst 4500 midrange switch product line we've seen dramatic growth in service providers using its for DSLAM aggregation and triple-play networks. Obviously, over the years you've seen many changes in the Ethernet arena with the growing popularity of network convergence, most recently IP voice services. How is that technology changing or redefining the need for high-speed Ethernet?
The demand for IP services has dramatically impacted the intelligence that we're putting into the switches in a number of areas. The first one is availability and resiliency. Real-time telephony is very demanding; you can't afford any downtime or glitches or hits in the audio quality of a call.
So we've seen increased demands for that, and have developed features like seamless failover in Layer 3. In our chassis-based systems, we have the ability to have two supervisors, or two basic switch engines. One is the primary and one is the backup, and we can execute a failover -- take one out of operation or failover to the redundant system in the event of a component failure -- without losing any phone calls, so there's no perceived glitch in conversations from any users attached to those. Power over Ethernet shipments tripled last year in the Ethernet switch space, according to Gartner. How is the rapid emergence of PoE influencing what you're doing at Cisco?
One of the things that's influencing PoE is that people are adopting powered endpoint devices, phones and powered access points and the like, and those require a centralization of power in the wiring close not just for the switch, but also for all the aggregate power required by the endpoints. One of the hallmarks of our chassis-based system has been the capability for a primary power supply and a second power supply, since the switch now has to provide all that power. So we see an interest in these chassis-based systems so the power can be centralized, and each can be backed up by independent UPS units that can be connected to different circuits. Your products offer IPv6 capabilities, and the IPv6 vulnerabilities in Cisco products revealed at the recent Black Hat conference have caught a lot of attention. Is there reason for companies considering Cisco GigE products to be scared off by the security implications, especially now that malicious hackers are allegedly working actively to exploit the flaw?
Security is an area that we continue to develop, not only addressing issues in terms of vulnerabilities, but security capabilities built into the switch to protect it from known potential attacks. One capability built into all the catalyst switches is that where we built our tables, we bind the IP addresses to the MAC addresses so we can tell if someone is trying to execute a man-in-the-middle attack. So while we always work on specific vulnerabilities, we really want to prove that we can deploy our products and use their security capabilities proactively to make the whole infrastructure secure. Finally, what are the top GigE trends you see over the course of the next few years?
I think we'll continue to see the desktops migrate. I think the biggest trend is that there hasn't been a matching of the technology on the backhaul side, so as you're increasing your desktop performance, people are still running those links to the core network. But with the introduction of 10 GigE technology that will run over existing multimode fiber in the backbone, we'll see that pent-up demand for 10 GigE connections start to come to fruition.