Why 2.5 / 5Gbps Ethernet? And Not 10Gbps?
- Thread starter iwod
- Start date
joegreat
Very Senior Member
Hi,
Ethernet cable speeds are currently 0.01, 0.1, 1 and 10 Gbps (better know as 10 MBit, 100 MBit, 1 GBit and 10 GBit).
The 1 GBit speed is becoming quite common and is availiable at a low price (e.g. I have full CAT6 cabling and run most of my cable connected devices on 1 GBit speed).
The 10 GBit speed is getting more and more common in the data-center area, but for home usage it's simply far to expensive (as of now)! The reason for the high cost is simple: Low number of controller/devices sold, leads to high cost in production (and slow amortization of the vendors investment). But time will come soon where 10 GBit will the the standard in home devices (like it was from 100 MBit to 1 GBit).
The mentioned 2.5 and 5 Gbps speeds are not available, but there is/was a discussion to define them as intermediate step towards the 10 Gbps cable speed - but there is no real reason (anymore) as 10 Gbps is standardized and in us!
Outlining my answer leads to the this conclusion: What is actually your question all about?
With kind regards
Joe
The issue is not just with the cost of the controller.
The real benefit is on the supported cable.
10gbe requires CAT6a cable for lengths greater than 55meters.
NBase-T will run 100m at 2.5g over CAT5e and 100m at 5g over CAT6.
Ask any installer. CAT6a is a very difficult cable to work with compared to even CAT6 and in larger installs is a costly upgrade over the existing lines. Making it more difficult is when you have large bundles over long distances, to keep the lines fully 10g capable they need to be shielded. This is due to alien crosstalk between other lines in the bundle. That should in turn must be grounded on one end else the performance will drop worse than if it was just UTP cable.
I have 1 client out in the field with the new Cisco nbase-t ports. They are not using them as intended but getting great benefit out of them as used. All it took was 1 line card and a replacement of 6 access switches. The existing cable was installed during construction in 1998. To replace the cable was quoted as more than the switch upgrade. Why the contractor chose to pour concrete over the bare cable instead of using conduit . . . I don't know. All I know is, the client is very happy about the results.
The real benefit is on the supported cable.
10gbe requires CAT6a cable for lengths greater than 55meters.
NBase-T will run 100m at 2.5g over CAT5e and 100m at 5g over CAT6.
Ask any installer. CAT6a is a very difficult cable to work with compared to even CAT6 and in larger installs is a costly upgrade over the existing lines. Making it more difficult is when you have large bundles over long distances, to keep the lines fully 10g capable they need to be shielded. This is due to alien crosstalk between other lines in the bundle. That should in turn must be grounded on one end else the performance will drop worse than if it was just UTP cable.
I have 1 client out in the field with the new Cisco nbase-t ports. They are not using them as intended but getting great benefit out of them as used. All it took was 1 line card and a replacement of 6 access switches. The existing cable was installed during construction in 1998. To replace the cable was quoted as more than the switch upgrade. Why the contractor chose to pour concrete over the bare cable instead of using conduit . . . I don't know. All I know is, the client is very happy about the results.
System Error Message
Part of the Furniture
There are also other alternatives such as HDMI networking and fibre optics. HDMI is a lot faster than 10Gb/s. Some of the cables used in connecting networks may also be at 100Gb/s but those are custom made and dont use the ethernet specification. Ethernet has been common because of backward compatibility. If you included HDMI network on a motherboard instead of the ethernet NIC than a consumer would most likely mistake it for a display port and plug a monitor into it.
Cisco edgerouters have a blade configuration with very high speed interconnects so while they take 10Gb/s fibre optics they are connected to one another by a much much faster CPU bus.
Cisco edgerouters have a blade configuration with very high speed interconnects so while they take 10Gb/s fibre optics they are connected to one another by a much much faster CPU bus.
sfx2000
Part of the Furniture
Client adapters are getting pretty reasonable on the ebay - Mellonox single port cards are about $20 bucks for an SFP+ adapter...
http://www.ebay.com/itm/671798-001-...e-10GBe-ETHERNET-NIC-/181853151588?rmvSB=true
http://www.ebay.com/itm/671798-001-...e-10GBe-ETHERNET-NIC-/181853151588?rmvSB=true
System Error Message
Part of the Furniture
I went on ebay and saw some NICs too, but the cheapest ones were from US. Perhaps this would help my testbed quite a lot but ubiquiti doesnt have any SFP+. Ofcourse is anyone here who wants to sell me an SFP+ NIC or even 10Gbe ethernet NIC cheaply do PM me. 10Gbe NICs will greatly reduce the time needed to get my test bed setup since i wont have to rely on having 20 gigabit NICs just to test if a router can do 20Gb/s of NAT. I thought it was still very expensive got get SFP+ NICs. I plan to use SFP+ direct since it is the cheapest and fasters but i suspect a lot of the SFP+ cables and modules on ebay may be fake.
Thx I think this pretty much answer my question. The reason is cabling, or specifically cabling Type.
Hi,
Outlining my answer leads to the this conclusion: What is actually your question all about?
Speed. Or the Lack of thereof. Basically Wireless Speed and Devices has improved and replaced at a much quicker rate we are now / going to be bottle necked by our wired connection.
We recently have 802.11ac 5300, that is now perfectly capable of saturating the LAN / WAN in a Multi devices environment even in real world usage.
We will have first draft of 802.11ax by late 2016 that bumps up to AX26000, if we follow the current marketing numbers. ( That is 26Gbps )
Thanks to the Smartphone SoC development, the CPU on Router and NAS are now much much more capable and at a cheaper price. Low Cost Home NAS used to be CPU bound when doing Network transfer, can now all do 100MB/s.
HDD has been hitting over 100MB/s transfer for a long time. Although no one cares about its speed anymore ( We use SSD for that now ), the NAS could help with some more transfer speed improvement without the need of using Link Aggregation.
On the server side there is Kinetic Open Storage, which is basically a HDD using Ethernet instead of SATA.
There are regions which has FTTB and CAT6 cable to their home.
So it sounds like with Nbase-t, may be we could see 5G speed in short length even with Cat5e cable.
So are there any plan for Nbase-t on the consumer side?
System Error Message
Part of the Furniture
It is unlikely that we will see 26Gb/s wifi because look at the range of 5Ghz wifi now and how many channels the AC5300 covers just for it. Increase the frequency and you decrease the range. All they can is try to squeeze out more from the 5 Ghz spectrum.
azazel1024
Very Senior Member
5GbE is actually supported on Cat5e to 100 meters, it isn't Cat6 that is needed.
Businesses are the reason why, LOTS of already installed cabling. What would a medium sized office want to do. Even if per port cost was the same for 2.5/5GbE compared to 10GbE, if they have a Cat5e/6 installed already, it might cost something like $30-40k to roll out 10GbE today to a 100 seat office for ONLY NIC/switch costs. However, unless you've got Cat6a, or the runs are short, you'll need to upgrade the wiring, which might easily be an additional $20-40k in installation and wiring costs.
Also power. 10GbE is ALWAYS going to use more power than something slower. Sheer physics. There can be power savings with race to sleep for other components, but generally not something that is a huge worry. What is a concern are things like chassis power dissipation. If you've got a 48 port switch with each port drawing 1w, you might only need to dissipate 55-60w for the entire thing (controller and internal PSU using a little power as well). That is probably doable with a good design for passive dissipation. However, 10GbE is typically using more like 2-3w per port and there is no way to cram 150-200w worth of heat dissipation design in to a 1u chassis.
Same thing for consumer routers and enterprise APs.
As System Error Message mentioned, keep in mind with 802.11ax there are nice increases in spectral efficiency, but only so much of that is going to be useable. From the bits and pieces I have seen, with similar noise floors and signal strength, you are only at best looking at around 1.3-1.7x increase in actual throughput compared to 802.11ac. Since some of it, IIRC, involves moving up the QAM set to either 1024 or 2048, to get those true "3x faster" speeds, you need very, very, very low noise floors. So expect that you MIGHT see those "3x faster" speeds when you happen to be standing a few feet from the router with nothing in the way. Probably even same room you'll be lucky to get 1.5x faster than 802.11ac.
The 26Gbps also will require 4 stream and 160MHz. The former is never going to happen on a client, so call it 13Gbps to be a little more realistic (2:2 client) and we still haven't seen 160MHz 802.11ac clients and routers shipping...because 160MHz is very, very hard, especially in 5GHz, because you have a huge frequency gap. You can't do 160MHz frequency contiguous. Your amps, radios, etc. have to be able to do spread spectrum across a very large range at the same time. That or you have to double a lot of components to do two 80MHz segments at the same time and bond them. Which increases costs a lot.
I am sure we will eventually move towards real 160MHz products, some day. I just don't think we are actually that close. I also doubt we will see real world speeds for 802.11ax hit anything like laboratory testing or future box claims. With 802.11ac the best of the best between client and base station (not base station to base station, as they generally have much better radio budgets because of higher antenna gain and more powerful radios) is currently around 60-65% efficiency. So a "1.3Gbps" stream is netting an actual ~840Mbps at absolute best with a tail wind. Not shabby at all I'll say.
I'd be shocked if you saw 802.11ax hit much better than 1200-1400Mbps under the best circumstances (also leveraging 80MHz). Of course it'll need 2.5/5GbE to do it, as you cannot do LAG to support that, at least not from a single connection.
On that note, I forget who just announced it (Netgear?) but latest "super fast router" is supporting 2 port LAG out of the box. About time. Wireless still isn't likely to push past 1Gbps of actual throughput and thus be port limited, but if you do have multiple clients, we ARE getting close to that point or past it, especially if you have 2.4GHz and 5GHz clients all going at once.
Businesses are the reason why, LOTS of already installed cabling. What would a medium sized office want to do. Even if per port cost was the same for 2.5/5GbE compared to 10GbE, if they have a Cat5e/6 installed already, it might cost something like $30-40k to roll out 10GbE today to a 100 seat office for ONLY NIC/switch costs. However, unless you've got Cat6a, or the runs are short, you'll need to upgrade the wiring, which might easily be an additional $20-40k in installation and wiring costs.
Also power. 10GbE is ALWAYS going to use more power than something slower. Sheer physics. There can be power savings with race to sleep for other components, but generally not something that is a huge worry. What is a concern are things like chassis power dissipation. If you've got a 48 port switch with each port drawing 1w, you might only need to dissipate 55-60w for the entire thing (controller and internal PSU using a little power as well). That is probably doable with a good design for passive dissipation. However, 10GbE is typically using more like 2-3w per port and there is no way to cram 150-200w worth of heat dissipation design in to a 1u chassis.
Same thing for consumer routers and enterprise APs.
As System Error Message mentioned, keep in mind with 802.11ax there are nice increases in spectral efficiency, but only so much of that is going to be useable. From the bits and pieces I have seen, with similar noise floors and signal strength, you are only at best looking at around 1.3-1.7x increase in actual throughput compared to 802.11ac. Since some of it, IIRC, involves moving up the QAM set to either 1024 or 2048, to get those true "3x faster" speeds, you need very, very, very low noise floors. So expect that you MIGHT see those "3x faster" speeds when you happen to be standing a few feet from the router with nothing in the way. Probably even same room you'll be lucky to get 1.5x faster than 802.11ac.
The 26Gbps also will require 4 stream and 160MHz. The former is never going to happen on a client, so call it 13Gbps to be a little more realistic (2:2 client) and we still haven't seen 160MHz 802.11ac clients and routers shipping...because 160MHz is very, very hard, especially in 5GHz, because you have a huge frequency gap. You can't do 160MHz frequency contiguous. Your amps, radios, etc. have to be able to do spread spectrum across a very large range at the same time. That or you have to double a lot of components to do two 80MHz segments at the same time and bond them. Which increases costs a lot.
I am sure we will eventually move towards real 160MHz products, some day. I just don't think we are actually that close. I also doubt we will see real world speeds for 802.11ax hit anything like laboratory testing or future box claims. With 802.11ac the best of the best between client and base station (not base station to base station, as they generally have much better radio budgets because of higher antenna gain and more powerful radios) is currently around 60-65% efficiency. So a "1.3Gbps" stream is netting an actual ~840Mbps at absolute best with a tail wind. Not shabby at all I'll say.
I'd be shocked if you saw 802.11ax hit much better than 1200-1400Mbps under the best circumstances (also leveraging 80MHz). Of course it'll need 2.5/5GbE to do it, as you cannot do LAG to support that, at least not from a single connection.
On that note, I forget who just announced it (Netgear?) but latest "super fast router" is supporting 2 port LAG out of the box. About time. Wireless still isn't likely to push past 1Gbps of actual throughput and thus be port limited, but if you do have multiple clients, we ARE getting close to that point or past it, especially if you have 2.4GHz and 5GHz clients all going at once.
azazel1024
Very Senior Member
It is possible, but since pretty much no one is touching DFS, unless the FCC changes the regulations surrounding the DFS channels, I don't see it happening anytime soon.
It would be nice if the wireless companies grew up and implemented DFS on their products across the board. In probably 98% of the US you could use most or all the DFS channels indoors and probably 90+% of the US you could use them outside as well. It is pretty much only very close to airports (not sure the distance, but from the bit I do know, something like 1-2 mile radius on indoor use from 5GHz radar and probably 3-5 miles for outdoor use) and some military bases where some DFS channels might be unavailable. Oh, airports which use 5GHz radar, which excludes just about all local and most regional airports.
IIRC there is 190MHz of DFS sitting out there, generally not being used, which would give 370MHz of total 5GHz that could be used in most of the US (180MHz for UNI-II and ISM that is currently non-DFS and the 190MHz that is DFS). You could fit two 160MHz non-overlapping networks, or 4 80MHz bad boys.
It would be nice if the wireless companies grew up and implemented DFS on their products across the board. In probably 98% of the US you could use most or all the DFS channels indoors and probably 90+% of the US you could use them outside as well. It is pretty much only very close to airports (not sure the distance, but from the bit I do know, something like 1-2 mile radius on indoor use from 5GHz radar and probably 3-5 miles for outdoor use) and some military bases where some DFS channels might be unavailable. Oh, airports which use 5GHz radar, which excludes just about all local and most regional airports.
IIRC there is 190MHz of DFS sitting out there, generally not being used, which would give 370MHz of total 5GHz that could be used in most of the US (180MHz for UNI-II and ISM that is currently non-DFS and the 190MHz that is DFS). You could fit two 160MHz non-overlapping networks, or 4 80MHz bad boys.
sfx2000
Part of the Furniture
You're really hitting two areas...
1) DFS in general - there are routers that support this already - just that many vendors/clients choose to not deploy in the consumer realm - Much of this has to do with network discovery and that clients cannot actively probe to search for the AP's in those channels - just ends up being a terrible user experience - oddly enough, I have seen AP's that when in Auto Channel mode select those channels, and there, some clients can eventually find their way their - either thru better service rescan (look for beacons), or thru AP channel lists, which some AP's support
2) 160MHz channels - 80MHz (VHT80) was a pretty significant challenge at the RF/Baseband realm - going to 160MHz channels (whether bonded or contiguous) is even more of a challenge. And even then, 160MHz channels in some areas will have to consider DFS... so there's both the engineering challenges (radio) and regulatory aspects...
And we still have headroom in the non-DFS channel space along with VHT80 as it is...
What's interesting is that in enterprise space, vendors are deploying VHT20/VHT40 modes (yes, this is allowed in 11ac), as there are still obvious benefits to running 11ac in those modes compared to 11n
NBASE-T technology is designed for the deployment of 2.5GbE and 5GbE data rates over existing Cat 5e and Cat 6 cablessansthe expensive and disruptive upgrades to cabling infrastructure. 10 Gigabit Ethernet is not something that can be deployed over Cat 5e cables. A majority of existing cables connecting wireless access points to access switches are Category 5e. Therefore, this means that 2.5GbE and 5GbE data rates would work with most of the existing cabling infrastructure. 10 GbE is more prevalent in data center networks and campus backbone.
Similar threads
Similar threads
Similar threads
-
-
-
-
Default Windows Network/Ethernet Settings- Started by anotherengineer
- Replies: 2
Support SNBForums w/ Amazon
If you'd like to support SNBForums, just use this link and buy anything on Amazon. Thanks!
Sign Up For SNBForums Daily Digest
Get an update of what's new every day delivered to your mailbox. Sign up here!