Why No 10GbE For Home Users?

[azazel1024]

I wonder if thunderbolt2 could be switched for more than direct pc to pc connections.

Simple answer, no.

 

[zakoh]

I can't imagine 10GbE showing up in any consumer or SOHO type of products any time soon. What is the killer application that needs 10GbE some time soon? The most obvious candidate is Ultra HD streaming, but apparently with modern compression, you need 20-50mbps of throughput for it. So a gigabit pipe can support multiple UltraHD streams. The next most obvious application is working with very large media files, and many of them, but here we're already in the realm of an extreme media nerd or producer. At best, I think we might see a 10GbE WAN port on the higher end home routers because 2gbps internet is going to become a reality in some places.

 

[azazel1024]

Unlikely. 2.5 and 5GbE are supposedly to be ratified this year. So more likely than not we will start to see 2.5GbE WAN/LAN ports on routers, switches, NICs, etc. within the next 1-2 years for higher end consumer products. Maybe in 4-5 years we will start to see 5GbE ports on those products for the higher end ones.

If 2.5 and 5GbE really do get ratified (seems very likely) and ever get deployed in a meaningful way (also seems likely), I would be 10GbE is at least 5 or closer to 10 years away from the consumer market.

 

[Cloud200]

I just got a few Cisco Catalyst 3850 switches with multi-gigabit ports at a client site.
The main differences that I can see are two-fold
1. Power. 10gbe burns a helluva lot more energy than 1gbe. 2.5 and 5 are in between but even 5gbe is less than half of 10gbe. When you have 48 ports in a switch all using 15-30 watts per port . . . the heat and energy cost add up very quickly.

2. Compatability. This client did not need to recable and was able to upgrade 70-100 meter links from 1gbe to 5gbe. The time and cost for running new cable was more than the cost of the switch itself. They had Cat5e already in place.

 

[Terry Kennedy]

Power. 10gbe burns a helluva lot more energy than 1gbe. 2.5 and 5 are in between but even 5gbe is less than half of 10gbe. When you have 48 ports in a switch all using 15-30 watts per port . . . the heat and energy cost add up very quickly.

Where does the 15-30W/port value come from? The Dell PowerConnect 8024 I have (24 copper ports w/ 4 SFP+ shared) specs at 237.77W maximum, all 24 ports operating in copper mode, for < 10W/port. The 8024F (24 SFP+ w/ 4 copper shared) specs at 160.78W maximum, or < 7W/port. Obviously, there's some fixed power consumption even with all ports shut down, so the per-port power is even lower than that calculated above. Both of those switches have a fair amount of processing power / memory / redundant power supplies. The semi-managed XS708E consumes 4W/port according to this benchmark. Even ancient XENPAK / X2 optics are only 4W max, and that's the whole SERDES in there (the switch side is essentially a parallel port).

Cabling cost is definitely a good point, but many 10GbE deployments will happen within a single rack (or a small number of racks) where re-cabling cost is less of an issue.

IMHO, the industry would be better served by getting prices (mostly) and power consumption (somewhat) down on 10GbE products rather than further fragmenting the market into additional speed niches below 10GbE - higher speeds are another matter, but should still be limited to a reasonable number of choices provided by a few base speeds (perhaps 10GbE and 100Gbe) with varying numbers of lanes (as 40GbE is today).

 

[azazel1024]

Terry, I almost agree, except for the issue of cabling. Unless a new 10GbE protocol comes about, you cannot run 10GbE to 100 meters on cat5e or cat6 cabling, which is what most buildings have installed in them (or cat5). You can run 5GbE up to 100 meters on cat5e. That means that a lot of businesses could PnP 2.5/5GbE right in to their existing infrastructure just having to replace switches and routers (of course NICs too).

Since all of this stuff is backwards compatible, I don't see the risk or issue. Deploy some 5GbE, some 2.5GbE and some 1GbE...they'll all work together. That is why IEEE is looking at standardizing the specifications from the handful of proprietary 2.5/5Gbps ethernet stuff. So that everyone's equipment will work together and wide scale deployment can happen.

10GbE is still massively higher power than 1GbE. Stuff from the last ~4 years is roughly half a watt per port, maybe as high as a full watt for long runs (probably not that much). I haven't looked at 2.5 or 5Gbps ethernet gear, but I'd assume it is somewhere between 1GbE and 10GbE. Which means that it is likely to be half or less the power of 10GbE, which can add up. Especially if you are starting to deploy this on the consumer level, as you have more than just the port costs in power consumption, you also have the NIC processor itself, which might be a few watts (so you might have only 4-6 watts per port, but you probably also have a relatively fixed 3-8w load from the NIC processor, obv depending on load). So you want to deploy this in a consumer router? Now you need to account for possibly 8-12w of extra power. As tiny as consumer router chasis are, you'd likely have to move to active ventilation. Want to deploy it in a laptop, you might be looking at 5-6w of extra load on a rather power constrained mobile device.

Power consumptions are constantly getting driven down through newer, better silicon on smaller processing nodes. It isn't a miracle though and realistically from what I have seen of the pace of things. To get 10GbE to a more "acceptable" level for wide spread consumer and business deployment (not core networks) we are probably at least 4-8 years away (driving ~4w per port levels down to 1-1.5w). Most businesses don't need 10GbE on workstations. Heck a lot of businesses are still using 100Mbps just fine. Some do need higher speeds or wouldn't mind being able to cheaply convert backbones across campus from 1Gbps to 2.5/5Gbps.

I personally see this as a good thing. Just like 100Mbps is almost always cheaper and lower power than 1Gbps, even if only marginally these days, 2.5GbE and 5GbE are probably always going to be cheaper and lower power than 10GbE. This also means they are more likely to get dropped in to a design that otherwise would have been 1GbE.

I do, in a lot of ways, also hope we see something like 25GbE some day and able to run on Cat5e/6, even if it is only for very short runs, like 10 meters or something. That might never happen, but it would be cool if it did.

 

[wis]

As "PrivateJoker" said, the first problem is price. If you know what you are doing, and don't mind buying used, you can get clear to 40Gb in the home for an OK price. If you get a good deal on Ebay, you'll only pay about $50 or $100 per card. Of course, the cable runs about $50/foot, which might add up to a bit.
...

In Apr 2015, I got 1000' of Belden 10GX12 Augmented for less than $550.00 (~$0.55 / foot) including shipping. At this point it cable costs don't seem to be the limiting issue. Installation, equipment, and power dissipation/removal...well that's another kettle 'o fish or two, isn't it?

 

[awstey]

Despite the devices needed for 10GbE network is being cheaper, such as the SFP+ transceiver, but the age of 10GbE for Home use still has not yet come.

 

[Cloud200]

I wouldn't say it has come for home users . . . but power users can easily take advantage of it.
I have a 48 port switch in my equipment closet with 2x SFP+ ports.
One is attached by a DAC cable to a server that if not for the need for me to run VMs at home, I could easily replace with a QNAP TS-563.

If you want the absolute cheapest switch that would have worked for this setup . . . a Mikrotik CRS226-24G-2S+RM.
24x 1gbe ports, 1x SFP/+ port and 1x SFP+ port.

Switch is $260
NAS is $560
10gbe Card for NAS is $247
10gbe DAC cable is $23
5TB HDD is $145

Total cost without drives: ~$1100
Total cost with drives: ~$1825

For a home with a dedicated media room and multiple users that will access content simultaneously this is well within the realm of reason.
Don't think of users with a $500 50" TV
Think of users with a $2000 70" 4k TV or even an $8000 curved 80" model (which btw are so much better if you can recess into the wall vs surface mount).

 

[sfx2000]

In the warmer months, powering and cooling it runs some $750/month (thanks to our ruinous Northeast electricity rates). I just switched over to outdoor air cooling (6" duct and fan) so bills should drop shortly.

Well, I suppose if you have a hot-side/cold-side on the home data center, you could offset the winter heating costs

Nice setup..

 

[azazel1024]

Well, I suppose if you have a hot-side/cold-side on the home data center, you could offset the winter heating costs

Nice setup..

Oh dear Christ. I just saw that bit. I can't imagine paying out that much a month in electrical. I'll grant you, I am not running serious equipment, but my entire network, router, WAPs, switches and server eats something like $10 a month MAX in electrical, not considering additional cooling cost in the summer (but since I have expensive oil heat, it probably offsets more oil heating costs in the winter than it incurs in summer cooling costs).

Your electrical costs for running your equipment is about 7-8x as much as my entire home uses in a typical month (for a family of 5).

 

[Terry Kennedy]

Your electrical costs for running your equipment is about 7-8x as much as my entire home uses in a typical month (for a family of 5).

The electrical rates here are simply ruinous - I have a friend who lives in a 1-bedroom apartment with gas kitchen appliances and aside from the normal lights, tiny A/C, etc. all he has is a laptop PC and a flat screen TV, and his bill is nearly $100 a month.

 

[L&LD]

The electrical rates here are simply ruinous - I have a friend who lives in a 1-bedroom apartment with gas kitchen appliances and aside from the normal lights, tiny A/C, etc. all he has is a laptop PC and a flat screen TV, and his bill is nearly $100 a month.

Interesting. What is the actual cost per KWH? Does he not turn anything off (at all)?

 

[azazel1024]

I know plenty of people who are ridiculously wasteful. My rates are $.14 per kwh. The best I can find for Con Edison was a story from a few months ago and it looks like around $.28 per kwh, which is pretty darned high.

That said, it would still be the equivalent of me more than tripling my electrical consumption, and that just being for electrical equipment itself. If that $750 a month is really just computers, electrical gear and the cooling system for it (not including any residential use in those figures) that works out to almost 2700kw-hr a month. Which is a LOT higher than the US average, for an entire house (which is around 1000kw-hr IIRC). I mean, that is about 3750 watts continuous in computer gear and necessary cooling for it all.

 

[azazel1024]

Oh, the math on the apartment is around 380kw-hr, which actually isn't too terrible, but considering it is gas for appliances and I assume water heating, that is actually pretty bad (maybe change out some lights for CFL or LED as a start). Back when I was in an apartment on my own, I used about 180kw-hr a month with gas appliances and water heating. The summer time it might have come close to 300kw-hr.

In a 2 bedroom apartment with my wife, I think we used on average about 350-400kw-hr, but that was with all electric everything. We currently use around 600kw-hr in the spring and fall, about 700kw-hr in the summer and 650kw-hr in the winter (oil/pellet stove in the winter and in the summer our house just doesn't use much AC to keep a 1-story rancher with a basement cool...and the reduced hot water use makes up for some of the increased AC use). Of course that 600-700kw-hr range is with a family of FIVE now, and a stand alone structure heating and cooling it. Then again, I work to keep the place energy efficient and tend not to have the AC turned up much or the heat up high (generally 67F in the winter and 78F in the summer).

 

[Chrysalis]

this is down to "is there a market" as well as costs.

currently all the momentum is on wireless.

We will see 1gbit+ speeds on wireless before any faster ethernet ports.

 

[azazel1024]

this is down to "is there a market" as well as costs.

currently all the momentum is on wireless.

We will see 1gbit+ speeds on wireless before any faster ethernet ports.

Except that we ARE rapidly approaching wireless being actually faster than the wired ports that can connect it all together. Wired will have to be improved before wireless gets there, otherwise wired is the bottleneck.

I doubt we will see link aggregation with consumer routers to fix that, so I'd expect to see 2.5/5Gbps Ethernet sooner rather than later.

 

[Chrysalis]

Which is why I said wireless will overtake ethernet on consumer routers.

 

[eddiez]

SOHO (retail) market is all about volume. Since 10Gbps has very limited added value in 95% of all homes (they just don't require the speed), they won't spend the money on it. It's just not a Ferrari that, besides the performance, also has quite some visible reputational value.

 

[Terry Kennedy]

SOHO (retail) market is all about volume. Since 10Gbps has very limited added value in 95% of all homes (they just don't require the speed), they won't spend the money on it. It's just not a Ferrari that, besides the performance, also has quite some visible reputational value.

True, but I would have expected to see more 10GbE offerings in the "prosumer" and SMB areas. Netgear has shown that a reasonable 10GbE switch can be profitable at under $100/port. But as soon as you look at the next tier up (Dell, Supermicro, etc.) you're looking at $350+/port (for example, Dell N4032).