Has anybody come across or know of bandwidth intensive applications that require end users to have gigabit ethernet network connections.
I've read the 'Deploying Gigabit Ethernet to the Desktop: Drivers
and Applications' Cisco white paper but
I'm not clear how it will benefit users at a remote site who access most of their applications from a datacenter over a 10 Mbps WAN link.
I worked in the oil industry for a while back in the late 90's and even then they were looking for gigabit to the desktop as the amount of data they were shifting was huge. As a side note gigabit ethernet wasn't ratified then so we ended up with ATM to the desktop running 155/622Mbps. All i can say is just be thankful gigabit ethernet came along :-)
But in answer to your specific question if all or the vast majority of traffic that the remote site users access is via a WAN link of 10Mbs then you are right there is little point in having gigabit to the desktop. The bottleneck will always be the WAN link.
Gigabit to the desktop is useful in situations where you have applications that shift huge amounts of data and the servers are either local to the LAN or in a gigabit/10 Gigabit campus environment.
If you don't have that setup by all means you can upgrade to gigabit / 10gigabit but you would be better off upgrading your WAN link to be honest, or of course rethinking your server placement.
Customer has an IT guy who thinks that 10/100 switches will become obsolete/eos within the next seven years.
AFAIA 10/100 port switches still account for most new LAN deployments so I think this is highly unlikely....I hope I'm right!
When i was answering i was primarily thinking about a current setup where you have a 10Mbs WAN link and someone was proposing to upgrade to gigabit to the desktop as a way to increase throughput. So what i said still stands.
But if you are looking at a new deployment it may well be worth considering buying gigabit capable switches even if you don't intend to use the gigabit functionality straight away.
Two main reasons
1) Bandwidth is getting cheaper. Admittedly we are all going through a difficult time financially at present but that won't be the case forever. All you need is a link above 100Mbps to the WAN and then your LAN infrastructure "could" become a bottleneck. I say could because you would still have multiple clients accessing the WAN at any one time so it is more likely that 100Mbps link will still be the limiting factor
2) Flexibility - it may be that further down the line that remote site needs to use apps that move large amounts of data and the business is suffering because of the WAN links. There may then be a case to move the servers locally.
Personally having everything in a data centre is in my opinion a better solution for many reasons but if the business is unable to function properly because of this setup then it's not really fit for purpose.
As with all things design it comes down to many factors not least of which is budget. If deploying gigabit capable switches made little difference to the overall cost then why not. Most people deploy Cat 5e/6 now even though they are not necessarily running gigabit speeds over it.
But if you need to cut costs and choosing 10/100 capable switches only gets you within the budget then with a 10Mbps WAN link i would still say this is a perfectly reasonable thing to do.
As for 10/100 switches being obsolete. Well it depends what he meant by that exactly. I think the cost of hardware will continue to drop the point where buying a gigabit switch is like buying a 10/100 switch now. But I still think there will be a place for 10/100 within the Enterprise. Your initial scenario is a good case in point.
As already discussed by Jon, perhaps gig connections would lend themselves for future developments.
You note "most of their applications" run across the WAN, but you didn't note all. So, although the WAN link would be your primary bottleneck, there could be immediate benefit to the remote site's LAN performance.
One side note, if the clients are pre-Vista Windows hosts, believe the RWIN default is different for gig vs. 100 Mbps. The former (gig) will often increase throughput on high (for a WAN) bandwidth WAN links (due to larger RWIN for gig and default being too small for the WAN BDP).
Most people don't need gig to the desktop for end users . Servers and such could benefit . I think I have seen maybe 2 or 3 instances where they were even able to fill a 100 meg pipe so for most end users gig is not a neccesity, 100 meg is a pretty big pipe. Gig is really only needed in server switches for the most part but it makes for a big marketing target .
Glen and Leo bring up a couple of common objections to desktop gig, either users really don't need it or workstations don't have the processing capability to utilize it. (This is a bit of dÃ©jÃ vu, for those of us who experienced the 10 to 100 transition.)
On the "need it" issue, unless there's absoluting no benefit to the additional bandwidth, it might be hard to argue there's no need, although the cost vs. benefit issue is a different issue. (One of the biggest cost factors is the cabling plant if it doesn't support the higher bandwidth.)
As to host performance, providing better than 100 Mbps thoughput shouldn't be impractical for most recent vintage PCs. Any PC that can support USB 2.0 (480 Mbps), Firewire (400 Mbps), ATA PATA 133 (about gig) or SATA-150 (about 1.2 gig) with PCI 32-bit/33 MHz (about gig) or later bus architecture should have sufficient performance for a gig NIC. (BTW, a PC that has a disk drive that averages "only" 50 MB/s [400 Mbps], a 100 Mbps NIC is somewhat slow compared to it. A gig NIC would permit network I/O to match disk I/O performance.)
BTW, I'm not suggesting everyone jump into gig to the desktop, just that it should be carefully analyzed for your network, both now and perhaps for a couple years out; whether it makes sense to start to provide it. Also, just as with 10 to 100 transition, not all users are alike. I.e., some, depending on their work, are likely to benefit more from gig than other users.
I just recently read the Gartner report regarding GigabitEthernet-to-the-Desktop.
Servers, for obvious reasons, would benefit with GigabitEthernet but desktop/laptops? Desktop/laptop CPUs don't have the oomph! to deliver full 100 mbps.
I agree with everyone's post, the WAN is going to be a bottleneck. Purchasing dual speed (10/100) or triple speed will depend entirely on your company's budget. I'd prioritize getting PoE switches than triple speed switches and if there is still adequate fundings, 10Gig uplink switches with triple speed.
Gartner analyst: Resist Gig Ethernet (25 May 2006)
Hope this helps.
Nice post! (+5) I didn't realize how a simple question could result in a very complex answer. He he he ...