atcomsystems.ca/forum Forums VOIP Phones & IP PBX Networking Gigabit to Desktop: Is it a hype or real demand?

I only see benefits for gigabit when used on office networks that have file servers. Most networks that use browser-based applications or other simple network protocols usually don't get much benefit from it.

Usually I see gigabit deployed into improperly wired/designed networks to fix the flawed layout. While it helps, something as simple as rewiring between switches and consolidating servers to a single switch would have yielded better results.

My colo is all based on 10/100 switches with gigabit uplinks between switches. Never have an issue. But in all fairness, there is no windows running at my colo either. Just by existing Windows seems to spam networks more then pretty much anything else.

Plus layer-3 switches for 10/100 are VERY inexpensive. I'm a big HP ProCurve 2600 fan. The Cisco Catalysts (3500/2900) have come down in price as well. I only use full gigabit switches for the core.

tito1411, an IP phone usually sits in between the switch and computer, which usually has a built-in 3-port switch, one port connects to the switch and one port connects to the PC. If your switch port and PC port are of Gigabit but your IP phone is of 10/100, they will all be auto-negotiated down to 100M. That means you pay for Gigabit, but you don't get gigabit. Cisco may forget to tell you this in their "Gigabit to Desktop" whitepaper. :-))

Clinton, first I should say I started looking at this "gigabit to desktop" issue when I was studying the marketability of our cableshare switch product. but I don't think I raised this question here on this forum was from the perspective of trying to educate the market not to use gigabit Ethernet. I have reached my own conclusion, and my intention is to validate the conclusion by getting feedback or comments from people on this forum who I can tell are senior professionals and gurus with first-hand experiences and insights from their daily involvement in voice/data network cabling and set up. We also sell gigabit switch product (USB powered), so don't just think I raised this question just to promote people's interest in our 10/100 products.

Now I want to get back to ROI from "gigabit to desktop" you mentioned. I can understand improved productivity when you go from 10M to 100M, but I bet most people don't feel this when you go from 100M to 1000M. Just take an example of how long it takes to move a 10MB file over network. At 10M, it is 10 seconds (estimated with 20% overhead), 1 second at 100M and 0.1 second at 1G. You see, from 10M to 100M, you save 9 seconds, but from 100M to 1G, you save only 0.9 second. Is this saving of 0.9 second the ROI you talked about? Cisco may say that ROI is where "Gigabit to Desktop" just costs twice as much as 10/100 Ethernet would cost but you get 10 times faster speed, but they may not tell you that "Gigabit to Desktop" would just give you a 0.9 sec saving in productivity per downloading of 10MB file. In other words, the benefits of "gigabit to desktop" is beyond what most human beings can see or feel, and probably there is not much there after you spent what you could saved otherwise. Agree? This may be really a good instance of how we consumers got "brain washed" to pay comfortably for what we don't need.

As for uplinks. For a 24x port 10/100 switch with 1x 1G port, the blocking ratio is about 1:2. But for a 24x port gigabit switch with 1x gigbit port as uplink, the ration is 1:24. See the difference?

Link aggregation? then you convert 10x 1G ports to a 10G pipe, and much less ports you can use now, and increase the port price 10 times higher, let alone the hassles of configuring the switch to do so. Also, you need to find a 10-port NIC to do so. Bad idea and not practical, I should say.

My own conclusion is that 100M (Fast Ethernet) will always be the "sweet spot" in term of data rate to desktop, for at least next 10 years from now, if not forever. May be I am crazy, but remember I said so if no one else has ever said so.

.
.
.
.


“There is no need for any individual to have a computer in their home.”
[Ken Olson, President of Digital Equipment Corp, in 1977]

“No one will need more than 637 kb of memory for a personal computer. 640K ought to be enough for anybody.”
[Bill Gates, founder of Microsoft, in 1981]

John,
Yes when an IP phone sits in between the switch and computer it will negotiate to 10/100 or 1000 depending on the phone's capabilities. I think thats pretty basic knowledge. I dont have any ip phones that use the internal switch to feed a computer. I prefer to home run if possible and that is also the recommended method of cabling so that the phone and the pc have their own cable. 100 meg for the next 10 years?? Ill buy ya a beer if your right. :toast:

It would be cool either way if anyone would quote the following 10 years later.

"100M (Fast Ethernet) will always be the "sweet spot" in term of data rate to desktop, for at least next 10 years from now, if not forever"

[John, Someone of Dualcomm Technology, in 2009]

tito1411: Running two separated cabling for two separate LANs will be the best and most expensive approach for IP phones, which are not unusual to be seen. The link below is a white paper ("VoIP without Hype") by Fonality that talks about different approaches in terms of performance vs. cost (page 4).

https://www.trixbox.com/files/VoIP-without-Hype.pdf

Another approach is to build two separated LANs on the single set cables of an existing LAN by using the two unused twisted pairs in each cable. This approach is applicable for 10/100M fast Ethernet though.

John, if you are looking for the opinion of senior professionals and gurus, I think you will find that most are somewhat horrified by a product that encourages people to ignore standards and best practice. Splitting one cable into two Ethernet connections will usually work, but anyone concerned with quality workmanship will not do this. Period. I have been asked to do things like this for clients in the past and I refused. I have a feeling most members of this board would do the same.

If you focus on moving little 10MB files over the network, then yes, gigabit seems like overkill. When I deploy a >40GB image to a workstation, I'm pretty happy to have it. When any IT department has to roll software out to 500 workstations, gigabit can save quite a few man hours. I deal with .VHD files and DVD .ISO files all the time that are >4GB, and even gigabit seems slow when I need to copy them to my lab machines. Does this represent all networks? No it doesn't. Some people do less with their network and some do more, but why would you close the door on higher speeds? We have standards which allow for scalability and compatibility with future technologies. What reasonable person would do the opposite?

The scenario you depict for link aggregation is not at all realistic. All you are doing here is giving examples of how to do things completely wrong. This doesn't prove that gigabit is pointless, it just proves that you are really good at inventing silly scenarios where things don't work properly. Your 10 port link aggregation is a wonderful straw man argument, but it completely misses the point.

You have a product that isn't capable of gigabit, so gigabit isn't necessary. You can't do link aggregation, so that's a useless technology as well. Your switch is unmanaged, so anything that requires configuring is just overly complicated and difficult. Does your cable splitter support POE, or is that another pointless technology that nobody needs?

Long live standards based cable plants and well designed networks. Gigabit capability will be there whether you need it or not.

Hi Clinton. First I would like to thank you and all other people who expressed their opinion to my question, either negative or positive and also who read my these posts. The last thing I want to see is no one cares and no one responds.

Now, I would like to explain a little bit more.

1) 10MB vs 4GB. The reason I used 10MB as the file size, not 4GB, is because I believe 10MB is probably already much larger than the average file size of files (including email, documents, pictures, audio and video) that are being daily transferred over a typical company's network. I remember I read somewhere that the average file size traveling over the Internet was much less than 100KB.

2) In a scenario that you install DVD-size software on a remote station over network, I guess the bottleneck might not be 100M vs. 1G. The whole thing may be just more slow down by CPU reading and writing to HDD.

3) "Gigabit to Desktop" does have its attractiveness for some special or niche applications such as doing movie special effects, just as the original article I referenced in my fist post indicated. But to majority of users, it is an over-kill, and it will be such for many years to come until daily average file size over network becomes larger, say 100MB, and I don't know when that would happen.

4)The 10-port link aggregation was indeed exaggerated, but it was partially used to point to the fact that there are no 10G uplink ports available on most gigabit Ethernet switches that you can buy today. This is a weakness that has been "ignored" by those who hype "gigabit to desktop".

5) It seems that you have some mis-understanding of our cableshare Ethernet switch. The cableshare switch product family can be a managed switch, can do link aggregation, and can have uplink ports of gigabit rate. It can support PoE too. It is also in full compliance with IEEE 802.3 Ethernet standards. Also, the cabling is the same as you cables regular Ethernet switches, meaning that there is no cable splitting along the cabling from patch panel to each drop of RJ45 wall plates. In other words, you get the same workmanship as you do for normal switches.

6) I guess you may be not comfortable with running two Ethernet signals within one network cable, which may be what you considered as a non-standard approach. Cisco actually has their 6500 switches that do cable sharing ( see Cisco Cable Sharing Switch ). In the past,people concerned about cross-talk degrading signal quality, this concern was valid for Cat3 and Cat5, but CAT5e and CAT6 are much better constructed in reducing the cross-talk.

Hope the above may be helpful.

John, your "RJ45/RJ11 Splitters" would have come in handy for me once or twice where I had no other choice but to split an existing phone or data cable due to damage in a totally inaccessible place to the other (repulling cable was not an option). The product is interesting, though I'd like to see something done with the fourth pair in the splitter (a second telephone line would be my first choice, PoE second). As for using it in new installs or in place of two cables, no way I'd ever consider it. Ditto the RJ45 dual splitter. Any data cable I pull must be able to be certified (whether or not it is), and splitters of any type do not allow for that.

As to your topic question, it really depends on whose desktop we're talking about. As Clinton already mentioned, some intra-office cabling is already using 1GB, and 10GB would be nice in some applications. For those who are just trying to 'get on the internet', 100MB hardware is massive overkill since the choke point is the service coming into the building. I see this all the time in hotels: they want cat6 everywhere (including POTS :bang: ) and gigabit switches, yet they buy two T1s to service the entire building's data needs. :confused:

I don't by any stretch consider myself a "guru", just pointing out what I have observed in my brief time in this field.

I have to say though, that you will learn more about data (and telecom) by looking through other sections of this forum. I can't help but notice that all of your posts have been in this thread, and that comes off as a bit one-sided and self-promoting to me. Not that that is always a bad thing, but imho you'd be better off explaining specific applications in which your product(s) excels.

Jack