|
Posted on 10.02.06
After recently putting a hold on a city-funded FTTH proposal, Palo Alto, California (in the heart of Silicon Valley) has issued a new request for proposals for FTTH. Filed under: New FTTH Rollouts and Regulatory Environment Comments: 17 Comments »RSS feed for comments on this post. TrackBack URI Leave a commentLine and paragraph breaks automatic, e-mail address never displayed, HTML allowed: |
This RFP is a victory for Active Ethernet. The RFP specifies… “Such a network must provide minimum symmetrical bandwidth of 100 megabits per second per end user connection.” My compliments to the City of Palo Alto for setting such a minimum requirement. They will get a fully future proofed network and may influence other municipalities when they move forward with their FTTP plans.
Comment by Frederic Feit — October 4, 2006 @ 9:43 am
PONS can also provide symmetrical bandwidth of 100megbits per second per end user connection.
Get rid of the marketing bulls**t. As we all know Active Ethernet oversubscribes the upstream from the switch at rates similar to that which is oversubscribed on the single Gigabit PON.
E.G
Active Ethernet have 24 or 48 00baseBX ports on a switch (such as the world wide packets, packetfront, cisco 3750 etc) and typically only a 1Gbit uplink, therefore you have a maximum gauranteed rate of between 20 to 40MBps.
A GEPON has 32 subscribers on a 1Gb PON (aka Uplink) therefore a maximum guaranteed subscriber rate of 35Mbps.
A GPON has 64 subscribers on a 2.5Gb PON (aka Uplink) therefore a maximum guaranteed subscriber 39Mbps per sub.
Basically the same numbers all round.
The tender does not say 100megabit per second per end user GUARANTEED, so PON (GPON, BPON or GEPON) can all deliver that functionality.
Comment by Stephen Davies — October 5, 2006 @ 3:43 pm
No Fibre to the HOME technology today can deliver symmetrical 100Mbps GUARANTEED. Residential home access networks since the internet was invented have always worked on massive oversubscription. Dialup, DSL, Cable networks and now FTTP (be it Active Ethernet or PON) all work on this principle.
Comment by Stephen Davies — October 5, 2006 @ 3:50 pm
In your description you use the atypical scanario of an active ethernet system sharing a 1Gbps uplink among 24 users, even 48. That does get the numbers down a bit but it is an unfair comparison. An active system can and typically does use a 2Gbps uplink among 24 users. One vendor offers a 10Gbps uplink for 22 users with 1Gbps links to the home. You do the math.
Comment by Frederic Feit — October 5, 2006 @ 9:51 pm
Your arguement is still flawed.
If an active ethernet does use 2Gbps (assuming unprotected) for 24 users the rate is 83Mbps per sub; 48 users the rate is 42MBps; GPON is 2.5Gbps with 32 users the rate is 78Mbps per sub.
As for the 10Gbps uplink for 22 users, what happens once this is agregated back at the central office. Are you infering there is no contention there.
You can make the numbers work any way you want, the reality is the access network is one small component of the end to end network.
The tender asks for FTTH solutions, and you cannot tell me that you are going to deliver uncontented capacity of 100Mbps to each subscriber in the network. No carrier in the world would deliver this.
Comment by Stephen Davies — October 8, 2006 @ 11:54 pm
“You can make the numbers work any way you want.”
But it is you who seems to insist on making assumptions and tweaking specifications to benefit PON in a comparison to Active. Yes, you can choose to deploy a GPON system with 32 users rather than 64 per 2.5 Gbps, but that doesn’t do wonders to the price per port and the basic financial modeling for PON in general. If you want to stay entirely inside the thoretical world I suppose you can keep reducing the number of users on a PON system until the numbers work the way you want them to, but you can’t sell it that way.
The appropriate numbers to compare are the 40Mbps guaranteed per user (64 of ‘em) for PON compared to 80Mbps guaranteed (24 of ‘em)for Active. That is the only fair comparison given the predominance of current equiptment availability.
I admitt, you make a very powerful argument for the equlaity of PON vs. Active, as long as you halve the capacity of Active and double that of PON.
As for the 10Gbps for 22 users, no I am not infering there is no contention, I am infering that, all things being equal back at the CO, with this Active hardware the performance for the end user is superior to GPON.
“…the reality is the access network is one small component of the end to end network.” With this it sounds as if you are stepping back from promoting or defending PON, instead saying there are many potential weak links and PON is just one of them. What if the satellite TV headend is in the development, is at the literal edge of the access network, for that part of the triple play the access network is no longer “one small component of the end to end network.”
Specific to your comments regarding the Palo Alto RFP, which I am not involved with in any way, well… of course not (100 Mbps uncontended). You are putting up straw man arguments then knocking them down yourself. It seems you’ve read the RFP, at least the pertinant parts, and if so you know Palo Alto did not specify their preference for a particular architecture. But you tell me, given the language they use, which I quoted in my initial post, to which architecture are they leaning. “Such a network must provide minimum symmetrical bandwidth of 100 megabits per second per end user connection.”
Comment by Frederic Feit — October 9, 2006 @ 3:08 pm
“But it is you who seems to insist on making assumptions and tweaking specifications to benefit PON in a comparison to Active. Yes, you can choose to deploy a GPON system with 32 users rather than 64 per 2.5 Gbps, but that doesn’t do wonders to the price per port and the basic financial modeling for PON in general.”
Clearly if you knew anything about the FTTH market and technology you would know that not one of the current deployments of GPON use a 64 way split. In fact it is impossible to use 64 way split to deliver RF video because the optical loss is too high.
All the business models have been worked out on 32 way split.
“you want to stay entirely inside the thoretical world ”
It is in fact you who are in the theoretical world. You fail to consider what is actually being deployed (32 split); you have not considered the whole of network and bandwidth engineering; and you fail to consider the cost of network ownership. The cost of deploying outside plant for Active Ethernet is some 250% more expensive than any form of PON. So I think making claims of adjusting PON counts is just showing your ignorance. In fact in calculating financial models for PON in brownfield deployments like Palo Alto you never consider 100% utilisation of the PON. In the past I have used a figure of 75% utilisation because it is difficult to predict where take is going to occur. Therefore the most accurate representation of bandwidth availability based on the costing model is 24 subscribers per PON. In Japan they actually have it as low as 16 subscribers per PON.
“…the reality is the access network is one small component of the end to end network. With this it sounds as if you are stepping back from promoting or defending PON, instead saying there are many potential weak links and PON is just one of them”
Not stepping back at all. I am pointing out that your assumption of 80Mbps per sub is based only on the access network, one small part of the total network. You have to aggregate that network back at the CO. If you are inferring a single 22 port active ethernet switch is going to have a 10G uplink, to service a small deployment of say 5000 homes you will require 228 10GE ports on the aggregation router. Please tell me which router are you thinking of that supports 228 10GE ports and a 2.28Tbps backplane.
“Such a network must provide minimum symmetrical bandwidth of 100 megabits per second per end user connection.”
And as I said in my orginal post, PON technology can just as easily provide 100Mbps Symmetrical bandwidth today as Active Ethernet.
Get off you soap box and look at reality rather than marketing hype and theory of networks. Look at what is actually deployed.
I notice, not once did you argue against my issue of network contention because you its right. And that is my whole point and the major flaw in the active Ethernet arguement.
Comment by Stephen Davies — October 9, 2006 @ 4:59 pm
You should go into politics as you are a master of knocking down straw man argumants I never made. Regardless, I guess Ben Lomand Telephone co-op and VolkerWessels Telecom just made major mistakes as “the cost of deploying outside plant for Active Ethernet is some 250% more expensive than any form of PON. ” Man, talk about ignorance. I’m done with you, go deploy your RF video over fiber.
Comment by Frederic Feit — October 12, 2006 @ 6:22 am
You quote” You should go into politics as you are a master of knocking down straw man argumants I never made”.
I am certainly no Politian. I prefer my life as a consulting engineer who has spent the last 7 years designing, building and analysing FTTH and IPTV networks throughout the world. I have researched and written many white papers on these subjects and presented at the several FTTH conferences.
“I guess Ben Lomand Telephone co-op and VolkerWessels Telecom just made major mistakes”
Again, you do dont know anything about the market - particularly in Europe. Approximately 90% of all FTTH deployments in Europe are currently using Active Ethernet because the EU (partly) funds most of the MUNI projects in Europe. One requirement of this funding is the fibre must be deployed using an open access method aka PTP. Hence the reason they tend to use Active Ethernet.
However this is likely to change over the coming years, because France Telecom (GPON), Telefonica (GEPON), British Telecom (BPON), Deutcher Telecom (PON) have all annouced FTTH rollouts.
“Ethernet is some 250% more expensive than any form of PON”. I have on many occations come across ignorant people who fail to work out what I term the “Total cost of Network ownership”.
Frequently people look just the cost of the ONT/ONU, sometimes all the equipment, but rarely the Out Side Plant and particularly the network maintenance cost. e.g. If you have a lot more fibre, it costs a lot more to maintain it. Only yesterday I came across this issue with one of my new carrier clients here in Australia. After presenting my analysis on this they have changed their mind.
Costing of a network should be analysed for a “Life Cycle” of deployment that encompasses the initial capital cost, deployment cost and operating cost of all components.
If you would like to see my detailed cost analysis on PON v PTP happy to send it to you.
“Go deploy your RF video over fiber”. Never said I deploy RF video over fibre. In fact every FTTH deployment I have been involved in has been IPTV. I have been deploying IPTV since 2000.
Comment by Stephen Davies — October 12, 2006 @ 5:00 pm
Stephen how quickly could i get a copy of your cost analysis on PON v PTP?
Comment by Rod Collingwood — October 31, 2006 @ 7:42 pm
Mr. Davies,
I will take you up on your kind offer to send me your cost analysis, I’m sure it contains valuable information and insight. Please send it to frederic.feit@gmail.com
I’m curious though…since the other gentleman seems eager to see your cost analysis as well, do you think he may be one of those people you were describing when you said “I have on many occations [sp] come across ignorant people who fail to work out what I term the ‘Total cost of Network ownership’.”? I’m sure when you come across these people you are certain to first call them ignorant, and then proceed to share your obvious expertise in the field. Your comments regarding the European market are a good example. I am certainly not an expert and I did learn something from your remarks, but only after wallowing through your insults. The two companies I cited (the Dutch company as well as the American one) were the lead stories on this blog that day, if you care about little things such as context. Your point that the Dutch company is in a position where it is required to build out with PTP due to EU funding participation was very interesting, germane to the subject of OSP costs and delivered with the eloquence of a ten year old.
Your comments regarding the EU open access requirements (for funding) raise another interesting issue, which has to do with a comparison of the capabilities of PON vs. PTP in the context of a open access system without the requirement for a PTP infrastructure. I would love to hear your thoughts about the capabilities of PON for open access systems compared to PTP, if you think you can manage to do so without hurling insults at me.
Cheers
Comment by Frederic Feit — November 1, 2006 @ 9:10 am
Stephen,
Any thoughts on the operational complexity of PON vs Ethernet. I think of…
- fiber management issues (more fiber terminations/ bigger ODFs in PTP compared to PON)
- troubleshooting of PON (OTDR won’t help much thru a spitter, right)
any insights are much appreciated!
Erik
Comment by Erik — November 2, 2006 @ 6:11 am
For those wanting a copy of the cost comparison, send me an email to stephend@titan.net.au
Comment by Stephen Davies — November 5, 2006 @ 4:38 am
Erik writes “fiber management issues”
There is significantly more fibre in a PTP solution than in a PON solution. To consider a typical 10000 home deployment (very typical for a CLEC) with 25% take up, a PTP would require 5000 fibres (2500 is using 100baseBX but few vendors offer low cost Active Ethernet CPE with this interface). A PON solution with 32 way split about 80-90 fibres (depending on PON efficency).
About 576 fibres can be “effectively” managed in a single fibre cabinet. While you will get larger fibre cabinets (such as the Corning cabinet - 1728 fibres) the reality with these cabinets is the fibre management fills long before you can populate the ports. So this equates to about 9 racks for fibre management. For a CLEC space is a premium.
Another issue with deployments of this scale is duct space. The about of duct space to bring this size of fibre back to the exchange is huge. Even if you pick the perfect location where you can send fibre into four different directions, this is still 1250 fibres per conduit, that 9 144-fibre cables for each conduit.
Next is the issue of terminating that size of fibre. One of the contractors I work with once said to my client that if they went with this technology they would not be involved in the deployment. When asked why he said to terminate that amount of fibre would take some 3-4 months just at the exchange end, let alone other parts of the trunk routes.
So ignoring the cost of PTP, the issues of fibre management of PTP are significant and complex, hence why the majority of FTTH deployments world wide are PON based.
“troubleshooting of PON (OTDR won’t help much thru a spitter, right)”
Actually the new OTDRs do go through splitters and some have been specifically designed for this type of work. You can even get an OTDR with 1625nm ports and 1310 and 1490 filters so you can test for fibre faults on a live PON without impacting any customer.
I frequently design networks where in the headend there is a two way combiner with a 95%/5% asymmetrical split - one port for the PON and one for a test point. This enables the test point to be used for testing of the PON using modern PON enabled OTDRs without impact on the customers with only minimal impact on the PON budget. The 95% port is used for the PON, the 5% port is used for the test point.
Happy to send you more insights, just send an email to stephend@titan.net.au
Comment by Stephen Davies — November 5, 2006 @ 5:01 am
Frederic wrote “I would love to hear your thoughts about the capabilities of PON for open access systems compared to PTP”
PON is very limited in its ability to provide open access of the local loop due to its sharing of the fibre, hence why Active Ethernet is more popular for open access style networks in Europe at the moment.
However, several of the recent DSL forum (although DSL has nothing to do with Fibre the standards are sometime relevant and frequently used by FTTH operators) standards (particularly TR-101), talks about the deployment of open access networks at layer 2 as opposed to layer 1 (the cable network).
Over the past 18 months I have transformed one of my clients in Australia to being a completely open access network using PON technology. They sell layer-2 wholesale services over FTTH exactly in the same way that encumbant operators offer layer-2 services over DSL networks.
Several FTTH operators in the United States are now using PON based solutions to deliver open access networks. Connexions and Zoomy communications build and operate PON based FTTH networks for developers and then sell access to those networks to any service provider wishing to deliver their products to home owners in the estate.
Both companies have been very successful in this style of deployment.
Comment by Stephen Davies — November 5, 2006 @ 5:13 am
Stephen,
Thanks for your insights in fiber management issues. I agree that terminating 10.000 connections in a POP will mean a large nr of fiber racks. In Europe (where I’m based) I see a POP size of 2.500 to 5.000 connections as more probable. This puts more of a burden on the hunt for suitable POP locations in urban areas, but the ducting etc is far more manageable.
Comment by Erik — November 8, 2006 @ 2:43 am
This is consistant with my studies. The average local loop length is Europe (UK excluded) is about 800m. The average loop length in the US and Australia is about 3.5kms.
The shorter loop lengths makes for lower density for the POP. The problem is as you indicate finding those POP sites. For an ILEC that easy, use the existing exchange, for a CLEC this is a much harder problem to deal with.
Comment by Stephen Davies — November 8, 2006 @ 8:16 am