Giving EPON the Vote

9:00 AM Comcast VP Doug Jones explains why EPON was picked over GPON when cable looked to tie in Docsis provisioning for biz-class services

spc_markl 12/5/2012 | 4:16:12 PM
re: Giving EPON the Vote


As you know, I did not want to get sucked into this debate, but when you get a chance, please ask him about the total networking cost comparing GPON with EPON.


Jeff Baumgartner 12/5/2012 | 4:16:08 PM
re: Giving EPON the Vote

I'll see if he cares to respond directly to me, or perhaps here on the boards. JB

paolo.franzoi 12/5/2012 | 4:16:05 PM
re: Giving EPON the Vote


Also, while we are going to ask about PON for business services let's talk about redundancy.  Not a lot of great options.  There are some other trickier issues with PON reliability for businesses of any magnitude (let's say with a component failure), but let's start simple.




Duh! 12/5/2012 | 4:15:58 PM
re: Giving EPON the Vote


I think you'd agree with me that G.983.5 was not a resounding commercial success.  Some bits and pieces of it did make their way from BPON to GPON, although not really enough to be implementable directly from the standard. I'm not aware of anybody who implemented any of it.

I think you might also agree that business services requirements are not monolithic.  Heck, there are lots of enterprises that are ok with availability objectives for SDSL, which has a lot more fault modes than any flavor of PON.   Many (in fact most) enterprises will likely find that availability objectives for unprotected GPON are acceptable.  Others might set their expectations  lower if they were presented with pricing for, say, a 1+1 protected GPON, including fiber route diversity.    And for the ones that can't tolerate any risk of outage, there are still SONET/SDH rings (and hefty bills to go along with them).



paolo.franzoi 12/5/2012 | 4:15:57 PM
re: Giving EPON the Vote


Here is my point.  The need for fiber would go with higher data rates (more than you can get from a cable modem).  If you are buying a 100Mb/s service - you want redundancy.

Right now, I work for a small (25 person) SaaS vendor with a 100Mb/s connection to our data center and WE have redundancy.

So, the space for EPON business services in this case are those customers who need more bandwidth than a cable modem and no redundancy.  I think this market will approach 0 (it won't be 0 but be close).

I have to disagree that many enterprises will find unprotected PON acceptable.


paolo.franzoi 12/5/2012 | 4:15:55 PM
re: Giving EPON the Vote



I agree with your segmentation.  And I agree that we disagree.  The thing is that you would have to uncover how many SOHO businesses are willing to pay the price of 100Mb/s service (and I know what that is) versus the price of standard cable modem service.

My thinking goes something like this.  Okay today I can buy cable modems with VoIP.  If I want to (with no new infrastructure), I could dedicate a QAM group to small businesses with DOCSIS.  So, now to plow in PON I have to justify it to go after bigger customers.  Okay great.  What are they buying today and what would I have to offer to get them to move.  Is it worth a construction project?

Would they run a Fiber into small business parks or office buildings?  I know that they told me they would have to do construction to get a fiber to me (and I am right next door to a good sized Comcast facility).  AT&T did not have to construct fiber and I use a completely independent backup service from a 3rd local carrier (would have loved to have it from Comcast but I could not wait for them).

I can not imagine a 1000 person office willing to live with no voice or data connectivity for an extended time.  Even in my little company our Customer Service guys have their Cell Phones connected as a backup to a cloud voice provider so that incoming calls can be routed to their Cells if we lose our voice connections.



Duh! 12/5/2012 | 4:15:55 PM
re: Giving EPON the Vote


We may have to agree to disagree.  In order to resolve it, we'd have to do a lot of microsegmentation to understand the business case for serving enterprise customers with a 100Mb/s service with availability targets consistent with PON.  My hypothesis would be something like this.

A rational business decision between 100 Mb/s services with- or without- protection, presumably at much different price points, is going to be based on an actuarial risk analysis.  It's a tradeoff between the loss to the business per hour of an outage in the worst case, and the availability objectives for the service, versus the monthly cost of higher availability and SLAs.   Clearly, that analysis would fall on the side of very high availability for your business, which I presume has SLAs to meet and thus would suffer catastrophic business loss in the event of an outage.  Others, not so much.  For example:

<li>an office with 1000 employees, each giving rise to 100 kB/s of traffic at busy hour,&nbsp; none of it mission critical</li>
<li>an engineering, video, architecture or graphic design firm that needs to transfer very large files with reasonable latency (and can, in the rare worst case, put a file on a USB stick and drive it to its destination)</li>
<li>a web-based enterprise that use CDNs for customer-facing services that can operate on a degraded basis without the backend with modest loss of revenue</li>
<li>a user of a remote backup service</li>
<li>some vertical segments, such as higher ed or perhaps insurance or maybe some sub-segments in retail</li>
<li>an enterprise that finds that it can get business disruption insurance for lower monthly cost than the difference between high availability and ordinary availability service.</li>

In any event,&nbsp; there appears to be a significant segment between the high end enterprise services you'd be using, and SOHO.&nbsp; My impression is that the MSOs are going after that segment, and leaving the SLAs to the AT&amp;Ts, Verizon Businesses and L3s of this world.

Duh! 12/5/2012 | 4:15:54 PM
re: Giving EPON the Vote


This would be a great conversation to continue over a beer.&nbsp;

Just to be clear, we are talking mostly about middle market companies, not many SOHO.&nbsp; Possibly large branch offices of large companies as well.&nbsp;

The opportunity as it was explained to me (by another MSO) was in opportunistic scenarios where an office park or a cluster of offices was passed by unlit fiber owned by the MSO.&nbsp; So right-of-way, engineering, permiting, and construction cost is going to be pretty reasonable, relative to service revenue.&nbsp; Also, opportunistically, greenfields&nbsp; builds.&nbsp;

As far as losses to "normal" businesses from outages, I think your example of cell phone backup through a cloud service provider is a pretty good indicator that there are alternatives to expensive protection switching.&nbsp; I'm also reminded of the time that a massive MS Exchange outage took out email and calendar at the big multinational enterprise where I worked,&nbsp; for two or three days.&nbsp; It sucked, and most of the troops wanted the severed head of the CIO delivered to us on a stake, but the incident didn't even rate a footnote in the 10Q, and nobody even got fired over it.&nbsp;

Also as you also point out, enterprises with high availability needs will often dual-home, through two different access providers, over disjoint fiber routes.&nbsp; Do both access providers need strict SLAs?&nbsp; Probably not.

So, yeah, it's about segmentation, and you and I aren't in a position to settle here whether the relevant segments add up to a decent business case.&nbsp;

Now, about that beer...

Sign In