Joint Alliance Pushing 2.5 Gbps and 5 Gbps Into The Mainstream

[Julio Urquidi]

In an effort to help spread the word about the soon to be ratified 2.5 Gbps and 5 Gbps Ethernet standard, the Ethernet Alliance and N-BaseT Alliance will be holding a plugfest event at the University of New Hampshire InterOperability Laboratory in Durham, New Hampshire. During the week-long event, the two consortiums will demonstrate the vendor-driven, interoperable potential of IEEE P802.3bz, the foundation for 2.5GBASE-T and 5GBASE-T networking.

Ease of deployment for the standards' new supporting hardware was also mentioned in the announcement. Since 1000BASE-T’s introduction twenty years ago, deployers of IEEE P802.3bz will be able to take advantage of the already existing 70 billion meters of Category 5e and 6 Ethernet cable and the 1.3 billion network outlets already existing out in the wild.

Next steps for the Alliances, after the standard definition and deployment, will be to spread the word about the IEEE P803.3bz standard, and the faster speeds it brings to the table. by working with vendors that are rolling out supporting 2.5 Gbps and 5 Gbps products.

For more information about the Base-T Plugfest, scheduled from October 10th to October 14th, go check out the Ethernet Alliance website.

 

[Ois]

When do you think one could realistically expect 2.5/5GBe switches and NICs to be on the market?
Will it still take 2 or more years?

I have heard that some existing hardware has comparability built in or only needs firmware upgrades, not sure how true this is.

 

[TheLostSwede]

When do you think one could realistically expect 2.5/5GBe switches and NICs to be on the market?
Will it still take 2 or more years?

I have heard that some existing hardware has comparability built in or only needs firmware upgrades, not sure how true this is.

No need for firmware upgrades, several 10Gbps cards supports line rates of 5/2.5Gbps as well, such as Intel X550-AT based cards and Tehuti Networks TN9510. Currently, the only switch I've seen that supports 5Gbps is a Netgear product and it's still $1,000+...

 

[Razor512]

Seems like a good step in the right direction, though 2.5 GbE will still be very slow for the average home network. Hopefully router makers just to at least 5GbE.

 

[thiggins]

Seems like a good step in the right direction, though 2.5 GbE will still be very slow for the average home network. Hopefully router makers just to at least 5GbE.

I don't know about the "average" home network. But data shows the number of Ethernet devices on home networks continues to decline.

 

[RussellInCincinnati]

I don't know about the "average" home network. But data shows the number of Ethernet devices on home networks continues to decline.

Yes, even the use of any devices at all bigger than tablet computers or cellphones is steadily decreasing among the ordinary users am encountering.

Am even losing interest in highest speed networking for ordinary office computing, in proportion to how many of my clients' tasks involve work with cloud storage "drives" and web apps in general.

 

[sfx2000]

I don't know about the "average" home network. But data shows the number of Ethernet devices on home networks continues to decline.

I would agree - if we were to do an informal poll looking at number of devices wired vs. wireless, it would prove this out... I'm about 50 percent wired myself, but most of that is infrastructure inside what is a non-typical setup, but the primary means of access for the wife and my work laptop is wireless, along with tablets/phones... The entertainment end-points are "wired" via MOCA and Powerline...

The NAS and Linux dev box are wired for obvious reasons, and my primary desktop is wired because it's two feet away from the main switch..

 

[lifespeed]

The geek-to-normal ratio here is skewed, but an informal survey among acquaintances indicates the wired network is not dead yet. I also find it is more desktop and infrastructure with phones, tablets and laptops roaming free.

It may not be growing much in installed base, but I think there are plenty of people who would upgrade their wired infrastructure when 5Gbps comes down to consumer pricing.

 

[mtganzer]

I don't know about the "average" home network. But data shows the number of Ethernet devices on home networks continues to decline.

Pretty much the whole reason for 2.5 and 5Gbe is to be able to support Wave 2 AP's over existing Cat5e/Cat6 cable plants without having to configure LAG on the switch. On the enterprise switch side, we see these ports already on POE LAN switches such as the Brocade ICX7450. Since the home market is still predominantly integrated AP/router with a few switched LAN ports, there really isn't a lot of need to put 2.5 or 5Gbe ports on these devices

 

[W4RH34D]

So are we going to have cat5f for this or what?

This seems completely asinine and 10g is the obvious technology to go with. I understand that wiring changes could be expensive but that's the cost of moving forward.

"Oh I want the bleeding edge wireless but I'll be damned if I invest in the infrastructure that supports it!"

 

[thiggins]

This seems completely asinine and 10g is the obvious technology to go with. I understand that wiring changes could be expensive but that's the cost of moving forward.

You significantly underestimate the cost of replacing all the CAT5e that is connected to access points in businesses throughout the world.

This is similar to DSL. Phone companies' largest investment was in all the copper wire connecting homes to their COs. Technology found a way to get lots of bits pushed across those wires.

If 10GbE were the answer to all the bandwidth needs above 1GbE, do you think folks would be mucking with developing lower bandwidth technologies?

 

[sfx2000]

One of the things that was not mentioned with NBase-T is that it sets the baseline for upgrades to the MII interfaces, not just external ports - and this is a big deal from a capability and cost perspective with the consumer Router/AP market.

Right now, a majority of the AC1900 and higher classes - the WiFi NIC's use PCI-e to send data to/from the SoC - and in the case of the tri-radio/dual band devices we see in AC3200/AC5300, this will typically include a PCI-e switch as well.

PCI-e is fast, but it adds more circuit traces for each interface, and typically one might need more layers in the CCA, and this drives up cost - with NBaseT, we simplify that connection, and we see a net cost reduction. Additionally, some SoC's break out to a switch IC, many times over MII, and up to now, this has been a sometimes bottle neck - with 2.5GB/5GB across that interface, the bottle neck goes away for the most part.

The SOC inside my router has four ethernet interfaces that are already capable of 2.5GB, but run through a Marvell 1GB combo PHY (4 to 4), so it's almost there...

So yes, it's good news, and something that is directly applicable to the SNB community...

 

[W4RH34D]

You significantly underestimate the cost of replacing all the CAT5e that is connected to access points in businesses throughout the world.

This is similar to DSL. Phone companies' largest investment was in all the copper wire connecting homes to their COs. Technology found a way to get lots of bits pushed across those wires.

If 10GbE were the answer to all the bandwidth needs above 1GbE, do you think folks would be mucking with developing lower bandwidth technologies?

Siemon - Why use Category 7A for Enterprise Networks?

This was over a decade ago. There are many, many reasons to go this route. It's time to move away from DSP correction.

 

[sfx2000]

Siemon - Why use Category 7A for Enterprise Networks?

This was over a decade ago. There are many, many reasons to go this route. It's time to move away from DSP correction.

In most data centers - 10GB and higher is on Fiber, not copper, and they buy enough that the cost delta isn't significant - as such, with Fiber, moving to a 40G or 100G link, it's usually just changing out the interface adapter itself, the same fiber is still used.

1GB links are usually CAT5e, with a mix of CAT6/CAT6a depending on the length of the runs - and that NBaseT can use the same media is a plus there.

 

[W4RH34D]

In most data centers - 10GB and higher is on Fiber, not copper, and they buy enough that the cost delta isn't significant - as such, with Fiber, moving to a 40G or 100G link, it's usually just changing out the interface adapter itself, the same fiber is still used.

1GB links are usually CAT5e, with a mix of CAT6/CAT6a depending on the length of the runs - and that NBaseT can use the same media is a plus there.

That's my issue. 1000T on cat5e was what I considered a hack to begin with, now they want to cram more bandwidth on it? How big is the heat sink going to be on the dsp processor? ;-)

Anyway, my next rewire will be for 7a (if I can find the darn installation components).
Unless computer manufacturers decide to integrate fibre nics into them, then yes, we all know fibre is the end game for cabling.

 

[crashnburn]

That's my issue. 1000T on cat5e was what I considered a hack to begin with, now they want to cram more bandwidth on it? How big is the heat sink going to be on the dsp processor? ;-)

Anyway, my next rewire will be for 7a (if I can find the darn installation components).
Unless computer manufacturers decide to integrate fibre nics into them, then yes, we all know fibre is the end game for cabling.

Hmm.. so who's keeping the fibre cabling expensive? Everything goes down in price after a point in time.. Something/ someone's holding the price up?

 

[TheLostSwede]

Hmm.. so who's keeping the fibre cabling expensive? Everything goes down in price after a point in time.. Something/ someone's holding the price up?

It's not a matter of the cabling, it's everything else that goes with it. It's hard making optical to electrical converters, especially at these speeds. $20 gets you 10m of optical cable, but it's hard to work with, comes pre-terminated in connectors that aren't easy to pull through a wall pipe and you have to be careful with how you handle the cables, as if they snag, you might as well throw it away. It's just not a consumer friendly technology, unlike copper wires.

 

[RMerlin]

Another reason for 2.5G is home consumer ISPs starting to roll out 2 Gbps in some areas. I find 2 Gbps totally overkill for a home Internet connection, but marketing departments don't care - they need bigger numbers to beat the competition...

Broadcom's next generation of HND devices will support 2.5G.

 

[TheLostSwede]

Broadcom's next generation of HND devices will support 2.5G.

Any idea which models?

 

[RMerlin]

Any idea which models?

No manufacturer has announced any model based on the BCM4908 yet.