Advice needed on how to go about cheap home 10GBE

[thiggins]

Follow the money. If there were significant demand for 10GbE, manufacturers would produce the products, but there is not.

The use cases cited are niche, not mainstream. The average consumer is mostly wireless and going moreso.

 

[System Error Message]

I am a computer / networking guy and I would love to see 10G over Copper for the home and I am sure we will see it eventually, it is pretty common place these days in the Data Center along with 40G but it will be sometime before we see it come down in price enough to be even remotely affordable.

I do a lot of photo / video editing and with today's SATA 3 Drives and SSDs it is simple to saturate a single gig link and while Link Aggregation is very common and supported on even some relatively low end switches it doesn't work how most would like it do. Even with the new SMB Version 3 which supports Multi-Path you aren't going to get more than 1G over a your multi-link connection / file copy unless you breakup your file copies in to multiple file copies which will then create different connections.

So since I don't see this changing anytime soon the only way to get faster network file copies is to get faster links and 10G would be awesome. I looked in to it before buying my Synology but realized we just aren't there yet and probably won't be for the foreseeable future. So I opted for a regular version with 4 ports in a LAG Group, sure it load balances but as mentioned above only a single TCP Connection so only 1G Max.

That said I do agree that 10G is way overkill for the "Average" user. Most people are thrilled with their Wireless G Wireless and that is fine for them but there are a lot of power users out there that wireless will just never be an option even with the newer standards out / coming out you still need the wired uplink to support it or at least other devices that will support it.

The cheapest way to get 10G is by using a managed switch with a motherboard with 2 gigabit ethernet + 2 Intel quad port server NICs. I once got a 2nd hand intel quad port server NIC for £25 and it was low profile. You will need 10 cables though and another 10 cables somewhere else. So for more devices you'll need more quad port NICs and larger switches or perhaps switches with stacking capability. Really shows how the prices havent improved. By the time we need 10G inflation would've taken over and 10G would be worth the same as 1G is now.

 

[forlotto]

I am a computer / networking guy and I would love to see 10G over Copper for the home and I am sure we will see it eventually, it is pretty common place these days in the Data Center along with 40G but it will be sometime before we see it come down in price enough to be even remotely affordable.

I do a lot of photo / video editing and with today's SATA 3 Drives and SSDs it is simple to saturate a single gig link and while Link Aggregation is very common and supported on even some relatively low end switches it doesn't work how most would like it do. Even with the new SMB Version 3 which supports Multi-Path you aren't going to get more than 1G over a your multi-link connection / file copy unless you breakup your file copies in to multiple file copies which will then create different connections.

So since I don't see this changing anytime soon the only way to get faster network file copies is to get faster links and 10G would be awesome. I looked in to it before buying my Synology but realized we just aren't there yet and probably won't be for the foreseeable future. So I opted for a regular version with 4 ports in a LAG Group, sure it load balances but as mentioned above only a single TCP Connection so only 1G Max.

That said I do agree that 10G is way overkill for the "Average" user. Most people are thrilled with their Wireless G Wireless and that is fine for them but there are a lot of power users out there that wireless will just never be an option even with the newer standards out / coming out you still need the wired uplink to support it or at least other devices that will support it.

I like your statement here.
But really I think we need to define average user.

Average is one of many different kinds of people:
- A trendy someone who needs to keep up with the joneses and stay trendy.
- A follower and slower adopter but they still like to keep up with the times.
- A power user who can't afford power user pricing so they wait for the price to be fair.
- A power user in the making they are learning new things and eventually will land upon logic or personal need to make them a power user.

To me these bottom three do not even equal what I would call a user their hand is basically forced by technology in order to stay in touch and on the level with social norms but they try with every moral fiber of their being to reject computers and internet most of them will not upgrade anything they will not buy anything at any price the only way they adopt it is through a free upgrade through some kind of phone plan or if a friend or relative buys a device for them so these ARE NOT the people buying anything anyways not to even be included because they contribute little or nothing to the market no matter what is or is not on the market.
- A person that doesn't have or does not want to have a clue until their machine no longer supports stuff or it lets the magic smoke out and they expect another free one from somewhere.
- A person that strictly uses their mobile device to check on facebook and thinks computers are still the devil.
- A person that has their family or friend come over to help the access facebook or their email on their device or their computer normally elderly and they have no real want to learn much they just want to see some pictures or read a message and respond.

These are the types of people we all know them we all have seen them they are not fiction they exist.

So in understanding what is an average consumer of these types of goods I would say that weather they need them or not at this moment many average consumers would adopt affordable 10GBE equipment no matter if the need it or not. It is not much different than what we see in other computer markets do we need directx 13 graphics ? Do we need SSD's that can go faster than 1GB or do we need a heck of a lot of things really NO we don't but we want them even if just for novelty at the time. The consumer even the average consumer if you do some research you will find that it is more the idea of something that sells than the product. Which is why we have all of these routers using hacks to cut through NAT to achieve speed unless you want to use the features of your consumer router than your speed gets heavily limited as the hardware is not designed correctly to give you the speed advertised as it does not properly deal with overhead this is a huge false advertisement yet people love the idea that they have this great 1GB router or this wireless AC3200 router which is really not capable of providing anywhere near that speed it is a false label. It is the idea of stuff that sells all marketing professionals know it. Part of the reason for the holding back of 10GB equipment is because it will prevent another idea from being thought as home users get home user class 10GBE probably yet aonther hack job that will get about 2-4Gigs of real useable speed the idea that home users have this ability will make their business class customers upset because instead of explaining to them that we are essentially ripping of consumers and small businesses by selling them an inferior product that does not really achieve the speeds advertised they will just say no no no trust me these things are way better than anything you can buy in the store.

Follow the money. If there were significant demand for 10GbE, manufacturers would produce the products, but there is not.

The use cases cited are niche, not mainstream. The average consumer is mostly wireless and going moreso.

No. This is not a money issue nor is it a demand issue once you sell them at a reasonable price there will be a demand for it! Yes they can be produced at the same rates that 100MB nics were some 10 years ago or even at the rates that 1GB nics are now they might not be as profitable but, they would still be profitable! Tell me that a 10GB nic should cost more than a motherboard with all of its interconnects and so forth and I will call you a liar! It does not cost more to make period end of discussion. The solution is really simplistic do a run of reasonable 10GB equipment for consumers at a reasonable price and then stop see what happens instead of saturating the market with 10GB equipment marking it way up until the price falls I assure you it will fly off store shelves once people catch wind of this affordable device.

The cheapest way to get 10G is by using a managed switch with a motherboard with 2 gigabit ethernet + 2 Intel quad port server NICs. I once got a 2nd hand intel quad port server NIC for £25 and it was low profile. You will need 10 cables though and another 10 cables somewhere else. So for more devices you'll need more quad port NICs and larger switches or perhaps switches with stacking capability. Really shows how the prices havent improved. By the time we need 10G inflation would've taken over and 10G would be worth the same as 1G is now.

Correct they haven't improved and the tech market is not the market that suffers from inflated prices so much actually they are more victims of inflation than the commodities market because when the chips are down you are only going to buy what you absolutely need you are not going to buy the latest and greatest cellphone if yours can already do everything but tie your shoes for you unless they force you to do it like apple when they update their IOS and say ok now the old programs won't work with any of new IOS and the new IOS won't support the old programs either! This is cool until your consumers jump ship and star

 

[Cloud200]

Honestly I see the fractional implementation of 10Gbe getting traction in the home before 10Gbe itself.
IE. 2.5 and 5Gbe.

There are four reasons they are going to be pushed as "needed" instead of 10gbe
First off is the Wave II AC devices will be pumping more data than a 1gbe line can feed but not so much it will saturate 10Gbe.
Second NBaseT uses Cat5e and Cat6 cables which are far cheaper and easier to work with compared to Cat6a.
Third, NBaseT is not a competing standard to 10Gbe. It is designed to be backwards compatible all the way to 10BaseT
Fourth, the hardware is cheaper if the manufacturer designs it to negotiate at only the lower rates.
Side point; Cat7 and up are not ratified standards and thus a manufacturer can sell you Cat3 cables at Cat7 and it will be legit. It is just their definition of Cat7.

As an interim solution being used in businesses today that is applicable to power users at home;
Etherchannel/Port Trunking/LACP/SMB3 Multichannel
There are a ton of names and implementations but it comes down to using more than 1 cable to link a device to the network.
2 cables = 2Gbe
4 cables = 4Gbe

 

[thiggins]

Quantenna Rolls Out 10 Gigabit Wi-Fi

Note Freescale's reference design

 

[forlotto]

Interesting .... It sounds to me like it is using more chips to accomplish the task more or less a distribution of load ... I still think this is very useful for home wifi really especially for a decent sized family home. The other good thing is it will likely cost less electricity to balance the load than using a couple of wifi access points or routers although the benefit of using multiple access points is simple you can cover more area put one at each end of your home with them mounted up higher and you'll have strong signal in your home anywhere instantly and have the load balanced.

I could be wrong but it sounds like more chips to achieve the goal and more antennas due to the more chips can't say for certain though didn't thoroughly look into it..

Forgot to mention I do love the copper port
Will likely buy one at a reasonable rate !

 

[sfx2000]

Quantenna Rolls Out 10 Gigabit Wi-Fi

Note Freescale's reference design

I don't think Quantenna has a lot of Credibilty, based on comments observed here on their QSR1000 platform deployments...

But I agree what is interesting is the Freescale and Cavium 10GigE reference platform hints...

Freescale - PowerPC based
Cavium - MIPS if I recall correctly

I think we're run into the wall the current SW/HW, mostly SW, on current designs as we look at what's becoming more common on the WAN side..

 

[sfx2000]

The cheapest way to get 10G is by using a managed switch with a motherboard with 2 gigabit ethernet + 2 Intel quad port server NICs.

Not just the cheapest, but probably the best at the moment - use cases for 10G are fairly focused, and that would be in the LAN side of the house, so really, no need to have it in a SOHO gateway other than for stickers on the box, as most folks wouldn't be able to use it anyways.

I'd rather have the SOHO vendors focus on being more efficient with the 1G they already have... performance and power, rather than piling on features that actually impact performance. In other words, what is a router, what is a small home/office server, rather than making the toaster-fridge that many OEM's insist on doing these days with DLNA, Samba, Torrents, UTM, DPI, blah, blah, blah...

 

[forlotto]

Anyone tested the speed through smoothwall wonder if it handles any 10GBE NICs someone could build a fairly solid router using smoothwall I know funny you hardly ever see it mentioned here ....

Been a while since I looked at it as my main train of thought was damn that thing is going to suck power quicker than a thirsty vampire but I'm still on the fence I have a couple of spare desktops I could easily adapt I've only ever really took a look at the setup in a VM personally but haven't owned anything real great hardware wise to throw at it surely got me thinking about it though. The dell 8024 is interesting and the Linksys that was mentioned here seemed rather nice as well but both are overpriced just calling it like it is the prices for this 13 year old tech is highway robbery granted it is only like a 15th of the price it was back then.

Well we will have to see about the LAN... It is my personal hope that they don't go 10GB copper just for the WAN and they extend it to the LAN as well! I didn't really see any references to the LAN ports or if there will even be any?

Interesting so it is not just me who sees the wall and the issue we have about a 50/50 split on the poll.

I thank you all for participating in this thread weather you agreed with me or not it was a good healthy discussion lots of good info and views really!

I'm not saying this is an end by all means post away if you dig anything more up! Seems like some sweet news for the small network builder community.

 

[funhouse69]

Honestly I see the fractional implementation of 10Gbe getting traction in the home before 10Gbe itself.
IE. 2.5 and 5Gbe.

There are four reasons they are going to be pushed as "needed" instead of 10gbe
First off is the Wave II AC devices will be pumping more data than a 1gbe line can feed but not so much it will saturate 10Gbe.
Second NBaseT uses Cat5e and Cat6 cables which are far cheaper and easier to work with compared to Cat6a.
Third, NBaseT is not a competing standard to 10Gbe. It is designed to be backwards compatible all the way to 10BaseT
Fourth, the hardware is cheaper if the manufacturer designs it to negotiate at only the lower rates.
Side point; Cat7 and up are not ratified standards and thus a manufacturer can sell you Cat3 cables at Cat7 and it will be legit. It is just their definition of Cat7.

As an interim solution being used in businesses today that is applicable to power users at home;
Etherchannel/Port Trunking/LACP/SMB3 Multichannel
There are a ton of names and implementations but it comes down to using more than 1 cable to link a device to the network.
2 cables = 2Gbe
4 cables = 4Gbe

Etherchannel/Port Trunking/LACP/SMB3 Multichannel

I have this implemented in my network from my various systems to my Synology NAS and all it does / will do it load balance. You will not / can not get more than a single Gig of bandwidth per TCP Connection. This is only good for multiple TCP Sessions and even SMB3 / Multi Channel isn't working for a single file copy as of yet. This is one of the main reasons I would love 10ge as it is a single port and a single tcp connection would be able to use the full bandwidth.

 

[Cloud200]

Etherchannel/Port Trunking/LACP/SMB3 Multichannel

I have this implemented in my network from my various systems to my Synology NAS and all it does / will do it load balance. You will not / can not get more than a single Gig of bandwidth per TCP Connection. This is only good for multiple TCP Sessions and even SMB3 / Multi Channel isn't working for a single file copy as of yet. This is one of the main reasons I would love 10ge as it is a single port and a single tcp connection would be able to use the full bandwidth.

Correct, best you can get is 1gb/s per device.
Eg. 4gbe ports on a NAS means 4 clients can connect at 1gbe each.

 

[KenZ71]

While the demand is not here now and probably won't be tomorrow consider the improvements seen in some areas.

Google fiber is pushing the old slow cable companies to step up their game. Municipal Internet providers are offering gigabit service now for similar prices vs the cable co. How soon until the real demand increases?

In 18 months my home connection has gone from 10mbps via Uverse to 150 Mbps via comcast with no real change in cost.

 

[azazel1024]

Sorry, lazy and I don't feel like reading over the extremely lengthy posts (not that they might not be worth reading).

A 10GbE ROUTER, is going to be massive money. Up in the many thousands of dollars range. A 10GbE internal network with a 1GbE router that can actually handle 2Gbps of full duplex routing is a better option and not terribly pricey.

For a 10GbE internal network, what kind of distance and how do you want to set it up? You can use SFP+ and Twinaux SFP+ modules which tend to run about $75 for a pair of cards with 10 meters to Twinaux.

On SMB3+, you are mistaken on the performance. It does NOT work like Link Aggregation. It spawns multiple SMB sessions for the SAME file. So, if you have two GbE links between machines, you get 2Gbps of through put EACH direction. I currently have my server and my desktop running Win8.1 with a pair of Intel Gigabit CT network cards in each one. I regularly get 235MB/sec of throughput copying files between the machines. For single large files.

If you set up the cards in teaming/LAG, then you lose this ability. I've played with a lot of networking with multiple Intel NICs and Windows 8 and 8.1 machines (and a bit with 10).

You absolutely can get a couple of quad port NICs and get 4Gbps of throughput each direction on large file copies if your disks can keep up.

So honestly, no, I see no reason for the expense of 10GbE right now. That doesn't mean I don't want it, but 2.5GbE and 5GbE are likely to come first before 10GbE is really affordable for a prosumer user who doesn't want to shell out $1500+ for a small 10GbE LAN (other than Twinaux, which you could do for maybe $250 for a pair of machines and $600-1000 for two machines connected through a switch that has a pair of SFP+ slots. Otherwise you are looking at around $1500-2000 for 10GBASE-T cards and switch for just two machines and not direct connecting them).

Personally since most of my bandwidth needs are with big ole files, they are going to spinning disks. My current 6TB array in server and desktop has plenty of space for me and to grow, I have 3.21GiB free right now (5.4GiB formatted space). Empty, the 2x3TB RAID0 array in both machines can manage about 360MB/sec read/write performance. At it's currently level, the free space read/writes are more in the 300-320MB/sec range. Still above what my 2Gbps link can do, but not super far ahead of it (235MB/sec).

 

[azazel1024]

To add since you mention a NAS, SMB3 itself does not support SMB Multichannel, ONLY Windows 8, 8.1 and Windows 10 implementations of SMB3 and 3.01 support SMB Multichannel.

No other vendor supports SMB Multichannel in SMB3 (and no one else is using 3.01 yet). So you can only do it between Windows 8, 8.1, 10 and server 2012 machines. If you are connecting to a NAS, you will only get 1Gbps of throughput.

 

[Cloud200]

To add since you mention a NAS, SMB3 itself does not support SMB Multichannel, ONLY Windows 8, 8.1 and Windows 10 implementations of SMB3 and 3.01 support SMB Multichannel.

No other vendor supports SMB Multichannel in SMB3 (and no one else is using 3.01 yet). So you can only do it between Windows 8, 8.1, 10 and server 2012 machines. If you are connecting to a NAS, you will only get 1Gbps of throughput.

Well, to add to that, Thecus does sell NAS like devices running windows storage server 2012
http://wss.thecus.com/product_W5000.php

So if you do have that series of NAS, then SMB3 Multichannel will work to speed up transfers.

 

[sfx2000]

Quantenna Rolls Out 10 Gigabit Wi-Fi

Note Freescale's reference design

Some more details... this is not consumer focused, so maybe not a worry, lol...

The reference design is a Freescale QorIQ LS1043A SoC ARM Cortex-A53 design, with the Quantenna QSR10g chip providing 4*4:4 2.4GHz and 8*8:? 5GHz support.

The Freescale SoC has 10GigE from the WAN, and interfaces with the QTN over PCIe - reference platform is running OpenWRT, not sure if this is BB or CC, or a private fork...

Couple of block diagrams...

 

[sfx2000]

Looking at the HLD Block Diagram, there's an imbalance of resources if you ask me...

I'll put something together soon on what I think would be useful for most SNB members...

 

[sfx2000]

Here is Freescale's 10G Reference Design for Residential GW's - note, no QTN on it, but I suspect the QTN ref board would be on that PCIe slot...

The heatsink is for the 10G NIC I'm guessing as an informed/knowledgable

engineer...

 

[System Error Message]

Sorry, lazy and I don't feel like reading over the extremely lengthy posts (not that they might not be worth reading).

A 10GbE ROUTER, is going to be massive money. Up in the many thousands of dollars range. A 10GbE internal network with a 1GbE router that can actually handle 2Gbps of full duplex routing is a better option and not terribly pricey.

For a 10GbE internal network, what kind of distance and how do you want to set it up? You can use SFP+ and Twinaux SFP+ modules which tend to run about $75 for a pair of cards with 10 meters to Twinaux.

The mikrotik CCR1036 can already do 10Gb/s NAT easily or 20Gb/s NAT if you include full duplex SFP+. It has 2 SFP+ slots so you need a switch with SFP+ to distribute to the 1G ethernet that many devices use and the many wireless AC APs that you may have. Although SFP+ fibre optic module may cost money but it will still be about the same price as a 10G fibre optic modem and SFP+ to SFP+ direct is very cheap. The CCR1036 now costs less than $1000 so saying that it cost thousands for that much throughput is false. The CRS226 with 2 SFP+ ports cost around $300 and than the rest just depends on whether you want your PCs using 10G or multiple gigabit ethernet ports.

The freescale reference board looks interesting, Where can it be bought from and how much?. Another thing interesting about speeds between 1G and 10G is that usb3 is 5G which will comfortably support a 2.5G full duplex NIC. The ASUS AC5300 may claim 5Gb/s of wifi throughput but practically i would expect at most 2Gb/s which would require some LAGG so if they're going to release it with the usual 4 switched gigabit ethernet ports and 1 gigabit ethernet WAN it will be a very dumb thing to do if it is marketed as a router instead of AP since there isnt enough ports to properly utilise it.

 

[azazel1024]

Okay, I stand corrected. Near a thousand bucks is a rather large cost increment over a router capable of ~1Gbps full duplex NAT. I'll grant you it isn't that much for many who have a moderately large budget and pretty tiny if you are actually paying for a 10Gbps internet connection, but none of that is what I would consider "residential" or "consumer" levels of affordability anytime soon.

A note, USB3 will not support a 2.5GbE NIC comfortably at full duplex. USB3 still has a fair amount of overhead and has additional latency over PCIe. At best you are looking at maybe 4Gbps out of USB3. Of course, you could still do a 2.5GbE NIC for USB3 and in most cases, use it very healthily, but it cannot support full duplex at maximum performance. USB3.1 could, but that will also fall short of a 5GbE at full duplex.