Active Ethernet vs. GPON

[rrh]

I am on a resident's association committee that is tasked with selecting a fibre supplier for the suburb. The suburb is comprised of a mix of residential and small business properties.

Vumatel have proposed a point-to-point active Ethernet solution; all other suppliers have proposed a GPON solution, with splitting ratios of up to 64:1.

From my research it would appear that the benefit of GPON is based mainly on the fact that it is significantly cheaper to install, but that active Ethernet is a better long-term solution. New Zealand (for example) originally mandated an active Ethernet solution; that was, until the supplier's prices/complaints forced them to permit GPON .

Whilst I am sure that all solutions will work today I more concerned with the future, when at 7PM every night all households in the suburb will be streaming to multiple 8K (or higher) TVs whilst the businesses are uploading to offsite backups.

Any comments / experiences greatly appreciated.

 

[itareanlnotani]

Both run fibre to the premises, although one is active, and the other is passive.

GPON uses passive splitters to a OLT with up to 64 users per OLT. OLT -> Telkom or similar for backbone.
Active Ethernet uses fibre to a switch. Switch -> Telkom or similar for backbone.

You could always compromise, and have them split to 32 users or even 16 users.
This isn't much more cost for them, and gives you some headroom. I think thats the best solution if you don't go with AE..

To be honest, you can't future proof too far, as equipment is outdated fast. It looks like GPON has a fair future ahead of it though, as its in use globally. Only real issue with GPON is upload, but that can be solved fairly easily at the cabinet by giving users a separate connection. Doesn't affect cabling.

This is a good read regarding that - http://www.geekzone.co.nz/forums.asp?forumid=49&topicid=58040

 

[itareanlnotani]

Snarfed from the thread I linked above

The main difference is underlined.
If thats important to you, go for AE.

Thinking about it, just future proof it, and go AE..


---

GPON - Ethernet comparison

Terms:

Central Office is the location where access nodes are hosted and connect to backhaul. Might be a traditional phone exchange, data centre or large enough outdoor cabinet to host several access node chassis, cooling, backhaul switches, and fibre trays. Always has some form of battery and/or diesel backup.

Backhaul is the bandwidth (usually redundant fibre links) connecting the access nodes to the provider or ISP, and from there to internet.

ONT is the optical terminal located at the users place, and converts the optical data into useful signals on copper, eg phone, ethernet, powerline, TV/coax.

OLT is the device that terminates PON optics at the CO, normally a card in a chassis of several OLTs along with any ADSL or Ethernet subscribers sharing the service and redundant backhaul, which together form the access node.

PON means Passive Optical Network, ie fibre that doesnt require any data switch to split the connection among many users. PON splitters can be hosted in the street or at the CO but lines are limited to around 5km total distance.


Homerun fibre means the customer gets a dedicated strand back to the CO. Passive splitters can be located at the CO (to allow individual upgrade to Ethernet) or in street cabinets (sharing part of link to central office).

GPON


Many choices of ONTs designed for home and small/medium business connection;



Works with either homerun fibre or using passive splitters in the street;



Supports dynamic bandwidth allocation that prevents upstream congestion/collisions;



Avoid the complexities involved in keeping electronic equipment operating outdoors;



Several splitters can be colocated in a single cabinet, allowing a combined PON or multiple PONs, up to 64 users each;



Possibly more practical in scenarios where overhead lines are used;



ONTs are not compatible with 10GEPON but use same fibre and splitters if ONTs are to be upgraded, although more efficient with colocated splitters;



Competitive issues to be managed include competitor hosting at central office and choice of bypassing splitters.


Active Ethernet


The access node does not use passive splitters, so most efficient using homerun fibre to avoid needing the number of distributed nodes that Telecom have been installing to overcome copper limitations;



Provides synchronous bandwidth so ideal for commercial users, ie same speed both upstream and downstream;



Supports longer distance;



Very standard technology, so fibre can connect to an access node (some have ethernet subscriber cards plus redundant uplink cards) or a variety of Ethernet switches available on the market.



Greater potential for congestion of backhaul and unreliability, so Metro Ethernet switches need to be designed for this type of application.



Optical modules available for 100Mbps (100BASE-BX) or Gigabit (1000BASE-BX10) will depend on equipment being connected to;



Because efficiency dictates hosting at a Central Office, competitors can (and expensively) host their own equipment at the CO and move users between competing providers and/or types of service;



Coexisting services and equipment that Active Ethernet encourages has the effect of improving competition/service levels and reducing economic efficiency (prices may be higher).



Access to hosting and backhaul at CO would still be issues, but regulation may be required to ensure that minimum standard of bitstream services remain available to smaller competitors along with QoS for voice services.



Competitive hosting at CO without the user's ONT having POTS lines raises the possibility of 111 emergency voice calls being unavailable on some services due to convergence of voice and data. This will probably result in regulatory intervention, eg from Commerce Commission.

 

[Roman4604]

The most practical difference between activeE vs xPON is not related to data, but rather electricity reticulation/distribution, which could make a large difference to the cost/effort of deployment.

 

[rrh]

Both run fibre to the premises, although one is active, and the other is passive.

GPON uses passive splitters to a OLT with up to 64 users per OLT. OLT -> Telkom or similar for backbone.
Active Ethernet uses fibre to a switch. Switch -> Telkom or similar for backbone.

You could always compromise, and have them split to 32 users or even 16 users.
This isn't much more cost for them, and gives you some headroom. I think thats the best solution if you don't go with AE..

To be honest, you can't future proof too far, as equipment is outdated fast. It looks like GPON has a fair future ahead of it though, as its in use globally. Only real issue with GPON is upload, but that can be solved fairly easily at the cabinet by giving users a separate connection. Doesn't affect cabling.

This is a good read regarding that - http://www.geekzone.co.nz/forums.asp?forumid=49&topicid=58040

Looking ahead only two or three years:

• GPON currently runs at 2.5GB.
• UHDTV (4K) TVs are already available on the market. UHDTV requires around 25Mb/sec to stream.
• A split of 64:1 therefore allows a maximum of 40MB/household. Two UHDTVs in one household - very much a reality in this area - would therefore probably exceed the maximum.
• Reducing the split to 32:1 (much the norm) would allow a maximum of 80MB/household - which is in itself close to the 50Mb/sec required by two UHDTVs.
• Double up if 8K TVs become consumer items

OTOH active Ethernet allows a 1GB connection.

My feeling is that active Ethernet can currently cope with tomorrow, whereas GPON's future depends on in-development hardware. And, given that it took ten years for GPON to become a commercial reality, I am loath to recommend we base our future on hardware that has an unknown delivery date.

Am I wrong ?

 

[rrh]

The most practical difference between activeE vs xPON is not related to data, but rather electricity reticulation/distribution, which could make a large difference to the cost/effort of deployment.

True, but in this case the P2P connections run from the fibre supplier's point-of-presence to the household.

So no reticulation - or protection - required.

 

[rrh]

Snarfed from the thread I linked above

Provides synchronous bandwidth so ideal for commercial users, ie same speed both upstream and downstream;

I had read the NZ discussion that you quoted. Very interesting.

I have subsequently found two discussions that explain the advantages of AE. That said, I fully expect that the major factor in the decision will be the disturbance (or otherwise) of the pavement flower-beds

http://www.sympatika.cz/wp-content/uploads/2012/12/pon_vs_active_ethernet.pdf

http://www.sympatika.cz/wp-content/uploads/2012/12/pon_vs_active_ethernet.pdf

 

[dotVIBE]

If you are thinking long term, active ethernet is the obvious choice. GPON is fine with current home broadband data requirements, but will it be ok 5-10 years from now? metro-ethernet already supports 40GB symmetrical links to each connection. The only reason we install 1GB switches is that the 40GB switches are very expensive. However, if the fibre network is installed for point-to-point, upgrading the electronic equipment (Switches, CPE's SFP's etc) is easy enough when prices come down (Moore's Law). If you install a PON configuration fibre network, you are stuck with G-PON technology. That's why we (ATEC) always propose point to point.

 

[jackshiels]

Remember:

1. Not everyone uses the bandwidth at once. Vumatel locations are running something like 20% of total capacity even on active ethernet. The backend from the suburb to the net will also have a set limit on bandwidth, so remember that you will only go as fast as the backhaul.

2. GPON can be resplit. If a certain area is getting overloaded, it is quite easy for a provider to change the split from 64:1 -> 32:1, 16:1 etc.

3. GPON can be upgraded - 10Gbps XPON is backwards compatible, meaning that the system can be replaced if need be and congestion becomes an issue.

 

[HavocXphere]

How sure are you the fibre project as a whole will fly?

If its borderline & might be rejected then I say go for the cheapest option available just to get any kind of fibre in there.

If its a sure thing then you have the luxury of pushing for active ethernet.

You could also try a hybrid thing & point out to the SMEs that they'd likely suffer the most if its GPON so it might be in their interest to push for active ethernet.

Realistically though in the short to medium term I'd be much more concerned about the quality of the data that backs it. If you've got a couple of gigabit links on the consumer side that very quickly adds up to some serious capacity that the supplier needs to cater for. So thats the area where I'd like to see evidence of forward thinking / suitable planning. especially since.

Not everyone uses the bandwidth at once.

You'd be surprised. I'm in a building compound that has ~400 fibre links and you can certainly feel that evening peak time.

I wouldn't think about 8K etc though...if you can get a system in place that delivers real world 50mbps with a reasonably data cost then its a good win imo.

If they want a capped product then try to push them for a night-time uncapped type deal. Uncapped is ideal of course...but letting people push heavy data during off-peak hours is a decent compromise.

 

[Chris.Geerdts]

Funnily enough I've just completed research for the SA market on this exact issue

For residential, GPON is so widely used worldwide that the attention to the issue seems a bit over-done. There are operators using GPON who also use active on their own telco core networks and are no stranger to high capacity design.

I don't see the roadmap as being in much doubt, The same fibre will remain in the ground but the active components have a roadmap, jumping to 10Gbps (ITU spec'd in 2010 already) and 40Gbps (NG-PON2 is being standardised now). ITU-T Study Group 15 is looking at 80Gbps and symmetical services (for business). Any solution involves a hardware refresh from time to time but the GPON roadmap is geared to maximise re-usability and co-existence.

I would say a residential area rollout can easily be mixed with the typical small and medium business too. Perhaps any exceptions which require major bandwidth, particularly uplink, could be accommodated by prior consultation.

For larger businesses (which you don't seem to have), or those with specific or critical requirements, consultation may be required in any event. The issues I would raise are wider than just speed, but include security, uptime guarantees, QoS guarantees and (on rare occasions) latency.

 

[Level7]

Hi All,

We are deploying GPON for our FTTH networks.

There are many valid points in this thread. Firstly GPON has a total down capacity of 2.5gbps and upload of 1.25gbps.

That is a capacity per PON port, so most standard slot cards will have 8 ports. So lets say you go at a 1:32 split, this will yield a maximum download throughput of 70mbps. Now, the nature of services means that not everyone will be consuming 70mbps at the same time. Most likely your backhaul wont even have that much total capacity regardless.

Active ethernet requires expensive switches and power in street cabinets just raising the prices even more. Its also a logistical problem getting power to these cabinets specially if you need to take power from a resident.

In a PON setup the splitters are passive and dont require power, just a small enclosure.

Also, take a look at Calix and some of the Hauwei OLT's are coming out with 10gbps downlink speeds.

 

[itareanlnotani]

Huawei （华为）, not Hauwei.

 

[rrh]

Remember:

1. Not everyone uses the bandwidth at once. Vumatel locations are running something like 20% of total capacity even on active ethernet. The backend from the suburb to the net will also have a set limit on bandwidth, so remember that you will only go as fast as the backhaul.

2. GPON can be resplit. If a certain area is getting overloaded, it is quite easy for a provider to change the split from 64:1 -> 32:1, 16:1 etc.

3. GPON can be upgraded - 10Gbps XPON is backwards compatible, meaning that the system can be replaced if need be and congestion becomes an issue.

1. The problems associated with backhaul bandwidth apply equally to both GPON & active Ethernet networks, so (for the purposes of this discussion) I believe can be safely ignored.
2. True, GPON can be re-split to lower ratios. However there are at least two problems associated with a re-split: cost and convenience:
   1. Cost: who will pay for the new splitters and the cost of installation ?
   2. Convenience: it may require digging up he existing splitters to gain access.
3. The next iteration is, as you say, 10Gb XGPON. This is available, but currently is not very popular. XGPON would require completely new equipment - at a cost. Who pays ?

 

[koeksGHT]

Who pays ?

The consumer, fibre is still going to be expensive for a good couple years.

 

[rrh]

How sure are you the fibre project as a whole will fly?

If its borderline & might be rejected then I say go for the cheapest option available just to get any kind of fibre in there.

If its a sure thing then you have the luxury of pushing for active ethernet.

You could also try a hybrid thing & point out to the SMEs that they'd likely suffer the most if its GPON so it might be in their interest to push for active ethernet.

Realistically though in the short to medium term I'd be much more concerned about the quality of the data that backs it. If you've got a couple of gigabit links on the consumer side that very quickly adds up to some serious capacity that the supplier needs to cater for. So thats the area where I'd like to see evidence of forward thinking / suitable planning. especially since.


You'd be surprised. I'm in a building compound that has ~400 fibre links and you can certainly feel that evening peak time.

I wouldn't think about 8K etc though...if you can get a system in place that delivers real world 50mbps with a reasonably data cost then its a good win imo.

If they want a capped product then try to push them for a night-time uncapped type deal. Uncapped is ideal of course...but letting people push heavy data during off-peak hours is a decent compromise.

The project is already "flying"

The cost to the suburb is the same irrespective of the architecture used, i.e. no charge. Furthermore the backhaul stays pretty much the same, so can (I believe) be ignored for the purposes of this discussion.

That said, I am attempting to evaluate the suppliers by thinking our requirements ten years into the future, The fibre suppliers say that their time-scale is twenty-five years.

The suburb is considered affluent (in my case: I wish !) so multiple TVs per home are pretty much standard. And so multiple 4K streams per household @ 25Mb per stream - when available in South Africa e.g. from Netflix next year - is a given.

And that's by 2017/8.

Incidentally, a story from one of the ISPs: they provided their new fibre users with a free 100Mb/sec connection on trial for three months. At the end of the trial 60% of the users signed for the 100Mb service ... !

 

[rrh]

If you are thinking long term, active ethernet is the obvious choice. GPON is fine with current home broadband data requirements, but will it be ok 5-10 years from now? metro-ethernet already supports 40GB symmetrical links to each connection. The only reason we install 1GB switches is that the 40GB switches are very expensive. However, if the fibre network is installed for point-to-point, upgrading the electronic equipment (Switches, CPE's SFP's etc) is easy enough when prices come down (Moore's Law). If you install a PON configuration fibre network, you are stuck with G-PON technology. That's why we (ATEC) always propose point to point.

My thoughts exactly.

 

[rrh]

Funnily enough I've just completed research for the SA market on this exact issue

For residential, GPON is so widely used worldwide that the attention to the issue seems a bit over-done. There are operators using GPON who also use active on their own telco core networks and are no stranger to high capacity design.

I don't see the roadmap as being in much doubt, The same fibre will remain in the ground but the active components have a roadmap, jumping to 10Gbps (ITU spec'd in 2010 already) and 40Gbps (NG-PON2 is being standardised now). ITU-T Study Group 15 is looking at 80Gbps and symmetical services (for business). Any solution involves a hardware refresh from time to time but the GPON roadmap is geared to maximise re-usability and co-existence.

I would say a residential area rollout can easily be mixed with the typical small and medium business too. Perhaps any exceptions which require major bandwidth, particularly uplink, could be accommodated by prior consultation.

For larger businesses (which you don't seem to have), or those with specific or critical requirements, consultation may be required in any event. The issues I would raise are wider than just speed, but include security, uptime guarantees, QoS guarantees and (on rare occasions) latency.

All of the research that I have read recommends GPON solely on the basis of cost: one can install as many as seven GPON connections for every one active Ethernet connection.

Future bandwidth shortfalls can possibly be met by switching to XG-PON and/or NXG-PON, but at what price and whose account ?

Whereas active Ethernet allows - today - a one GB connection. And since we aren't paying for it ...

 

[rrh]

Hi All,

We are deploying GPON for our FTTH networks.

There are many valid points in this thread. Firstly GPON has a total down capacity of 2.5gbps and upload of 1.25gbps.

That is a capacity per PON port, so most standard slot cards will have 8 ports. So lets say you go at a 1:32 split, this will yield a maximum download throughput of 70mbps. Now, the nature of services means that not everyone will be consuming 70mbps at the same time. Most likely your backhaul wont even have that much total capacity regardless.

Active ethernet requires expensive switches and power in street cabinets just raising the prices even more. Its also a logistical problem getting power to these cabinets specially if you need to take power from a resident.

In a PON setup the splitters are passive and dont require power, just a small enclosure.

Also, take a look at Calix and some of the Hauwei OLT's are coming out with 10gbps downlink speeds.

Backhaul bandwidth is common to both GPON and active Ethernet, and so for the purposes of this discussion can IMHO be ignored.

All of the research that I have read thus far recommends GPON (over active Ethernet) for essentially major one reason: the lower cost to the installer. I would be happy if someone would point me to a document that recommends GPON for reasons other than cost.

GPON solutions result in the users being 'locked in' to GPON. So a move to XG-PON or NXG-PON would be relatively expensive and might require trenching. Who will pay ?

OTOH active Ethernet provides a 1GB connection per household (vs. 70Mb for GPON) today.

If upgrading - or repair - is required Ethernet switches are relatively common and thus relatively inexpensive.

And, in our particular case, the active Ethernet solution will have fibres running from the supplier's point-of-presence to each household, i.e. no street cabinets, power etc. required.

 

[rrh]

The consumer, fibre is still going to be expensive for a good couple years.

My point exactly. And the reason why I'm fighting for a solution that won't require an upgrade in the next five-to-ten years.