Mark Fabbi, vice president distinguished analyst at Gartner, leads the firms Enterprise Network Infrastructure research, and regularly has the ears of the top CIOs and network executives in the Fortune 500. At Gartner's Symposium/ITxpo event last year in San Francisco, he laid out this argument against over-overbuilding corporate LANs with Gigabit Ethernet: installing huge pipes to the desktop is a waste, as more users are working from remote offices and from home. Fabbi expands on this idea in this Q&A with Phil Hochmuth.
Why is Gigabit Ethernet spending being wasted in the enterprise? Is it a lack of applications that require 1000Mbps?
There are no typical business apps that need [Gigabit Ethernet] to the desktop. We have identified a number of apps where Gigabit Ethernet is warranted to increase productivity. Examples are video production house, GIS, scientific research, CAD, medical imaging & PACS.
A common criteria across these is that they all involve large files [a minimum of 100GB, but more likely 500MB plus]. For the other 90 percent of us, extra bandwidth is completely unwarranted. If you think about it, an increasing number of employees are working remotely, from home, on the road, using Wi-Fi, and they are generally not inconvenienced by less bandwidth. If we can work with little (or no) problems at the end of a DSL link, why do you need 1000x the bandwidth if you work in an office?
The why do you think network managers are buying GigE if they don't need it?
Companies have continued to follow their thought processes/design process of the past. Speeds and feeds used to be important at the edge of the network. We now have enough and need to turn to some other areas of investment. Vendors have been pushing GbE very strongly and too many organizations just follow the recommendations of their vendors.
Pricing for Gigabit Ethernet -- copper-based GigE, especially -- has dropped dramatically over the last several years. Could this be a factor as to why users are spending on the technology? In other words, if Gigabit Ethernet is only a few dollars more, why not get?
Gigabit Ethernet prices are falling, however the delta is still pretty high across many product lines. We factored this into our calculations (and even assumed that, despite are strong recommendations to avoid buying Gigabit Ethernet, the market adoption would continue). We looked at the delta between copper 10/100 and copper GbE for 206-2008 and yes the delta declines, but it's still material even in 2008.
We are seeing some vendors attempt to make GbE more economically viable. I'd point to HP, Foundry and Nortel all of which have 10/100/1000 switches with PoE at roughly half the price of Cisco's equivalent (and equal or less that Cisco's 10/100 PoE products). While this is a good move (and I encourage this approach) I would still tend to recommend looking at 10/100 PoE products are part of the evaluation.
If users are over-engineering the LAN, where should they be spending there IT dollars? Are you saying dollars spent on Gigabit Ethernet LANs be better-spent on application acceleration, WAN optimization and other technologies?
On the LAN, enterprises really need a reason NOT to buy PoE, so buying 10/100 + PoE is a much better investment than 10/100/1000. Other areas of spend would be in improving application performance - so looking at improving the performance of the 80 percent of users that don't work in larger offices. So better WAN solutions, Application delivery controllers improved access solutions and then outside the network org - improving the deployment time of new applications that will have direct impact on user productivity. Finally, some of the savings can be returned to the bottom line for a private sector organization or to the tax payers in the public sector.