Hello all,
For those of you with optical last mile networks that are familiar with both GPON and GEPON, would you mind sharing experiences of the differences between GPON and GEPON, especially from an operative perspective?
For arguments sake let's assume bitrate, split ratio, cross vendor compatibility and purchase price differences aren't of major interest.
Thanks,
Jared
If you take out "bitrate, split ratio, cross vendor compatibility and
purchase price differences" then what else would you like to compare or
know? Those would be the major differences I would say. We only deploy GPON
here. I would say in a system like GEPON or GPON where a port is shared
between users more bandwidth is better, and GPON has more capacity than
GEPON. I am not sure which region you are in, but in the USA GEPON is
almost non-existent from the larger players. Meaning that most GEPON
equipment won't be ANSI certified, and might not have FFC certs.
Huawei used to have a couple of slides.
I looked on some other list and found the following:
We considered EPON, and there are some inexpensive solutions from off shore
that are worth considering.
In the end, we went for GPON for two reasons:
One, you can deliver a true 1Gbps service where more than one customer on a
PON segment can actually get 1Gbps at a time, because the GPON supports
2.4Gbps of total usage on the segment.
Two we like our current vendor, Adtran, and we wanted to put OLT cards into
the same chassis and manage them using the same systems. The cost premium
versus a new vendor for EPON wasn't huge. The CPE is the bigger cost, and
we didn't see a real cost difference between EPON ONT and GPON ONT.
In the end, the price difference for GPON versus EPON wasn't great - and
while GPON is a bit "designed by committee" and there are some valid
criticisms there, they're academic in light of the other factors.
The solution for selling 1G internet with EPON could be 10GEPON. This is
still cheaper than GPON. The idea is that the ONU has a cheap standard 1G
transmitter. Apparently you can make a 10G receiver very cheap, it is the
transmitter that is expensive. So it is 10G downstream and 1G upstream.
With the option to deliver 10G upstream per ONU. It is about reusing
standard ethernet components that are dirt cheap - you can buy ethernet SFP
modules for peanuts after all and 10G SFP+ modules are not that expensive
either.
However when we asked some vendors about this, nobody wanted to sell to us
because Europe (and USA I assume) is GPON while China is GEPON. They did
offer to sell us GPON at 10GEPON pricing instead...
Something fishy is going on. It is not about EC compliance as it is just a
matter of buying a 10GEPON card instead of GPON card to the same chassis
switch. Maybe patents?
Regards,
Baldur
At this point if you haven't deployed any of these system, make sure you know the road map of your vendor for N-GPON2 that is going to be the next wave of deployed pon systems.
https://www.calix.com/solutions/next-generation-pon.html
Carlos Alcantar
Race Communications / Race Team Member
1325 Howard Ave. #604, Burlingame, CA. 94010
Phone: +1 415 376 3314 / carlos@race.com / http://www.race.com
If you take out "bitrate, split ratio, cross vendor compatibility and purchase price differences" then what else would you like to compare or know?
All the interesting bits obviously 
Anybody can read the bitrates, split ratios, compatibility and price of a spec sheet/quote. That however leaves out all the interesting operative aspects such as auxiliary network requirement, service turn up and software tool differences between the two standards.
The hard facts only cover the CAPEX part of the TCO equation and the differences between GPON and GEPON are small. Controlling for any parameter roughly equal or if any different within a constant factor of less than two.
I'm more interested in the OPEX part, to find out if there are any (significant) differences between the two.
I welcome all insight into the operative aspects of GPON and/or GEPON, regardless if you have used one or both.
One, you can deliver a true 1Gbps service where more than one customer on a PON segment can actually get 1Gbps at a time, because the GPON supports 2.4Gbps of
total usage on the segment.
I know this is a quote of a quote, whose origin I do not know, but I would not feel comfortable offering "a true 1Gbps service" on any PON system with less than 10G of capacity. Plain GPON/GEPON is meant to be split vigorously to achieve cost savings in the OSP and as such aren't suitable for gigabit speeds. It's more like a 100M kind of technology.
Jared
It all depends on how it is designed as well.
Take a Calix E7-2. You could do a pretty high split per gpon port, I think
either 32 or 64 is the max for them, but you're really just shooting
yourself in the foot IMO if you're advertising and selling a lot of gig
service.
A 8-16 way split per gpon is more reasonable. I think the current cards are
4-10 gpon ports per, and 2 cards per E7-2. I know they have 2x10Gbps LAG
working for uplink, can't remember if 4x10Gbps LAG works yet or not.
Count in oversubscription rates for residential, and consider that most
people, despite what they say or think, will end up on 2.4GHz wireless in
the home due to 5GHz sucking more than a room away - that ends up being a
very scalable solution for residential service.
For SMB, they end up on a different split, or with SLA end up on an active
port on the chassis or on the Juniper access/transport switch.
Count in oversubscription rates for residential, and consider that most
people, despite what they say or think, will end up on 2.4GHz wireless in
the home due to 5GHz sucking more than a room away - that ends up being a
very scalable solution for residential service.
Um… 5GHz works a lot better from one end of my house to the other than 2.4Ghz
due (in large part) to this fact… Almost every one of my neighbors is using
various 2.4GHz devices including about 45 external SSIDs visible from the
center of my house using the on-board antenna of an ESP8266 board from Adafruit.
The noise floor and congestion on 2.4GHz in many urban settings, especially here
in Silicon Valley makes 5Ghz a much better option in any home where people are
smart enough to pay attention to the difference.
OTOH, since the WiFi consortium took away the ability for consumers to easily
differentiate (it’s all “n” or “ac” now regardless of frequency) and you have
to really read the fine print on the side of the box to find a 5Ghz capable
WAP at your local big box store, most consumers end up on 2.4Ghz because those
are the least expensive routers on the shelf.
Personally, I don’t mind this, but I think the 2.4Ghz prevalence has more to do
with consumers not knowing what they are buying than it does with performance.
Owen
I think that was Josh's point, that 5 GHz will likely deliver better RF performance than 2.4 (despite physics) due to the amount of interference in 2.4.
Customer devices will see the higher signal on the 2.4GHz AP and simply
connect to that, especially as they roam through the house. Most don't pay
attention to SNR at all.
Only if the 5Ghz and 2.4Ghz networks are on the same SSID.
I don’t do that… I maintain separate 5Ghz and 2.4Ghz SSIDs. This allows me to know
which one I am on and force when desirable (usually forcing 5Ghz is desirable).
Owen
You are not the average user.
Most reputable enterprise wireless solutions employ band-steering which helps to "force" users onto 5ghz, but still allows clients to connect to 2.4 if it's the only SSID strong enough or if the client only supports 2.4ghz. Band steering largely negates the need to run two SSIDs for optimal band selection.
Chris
True. I know a number of average users that also do what I am doing, however.
Owen
True, but most households are not using a reputable enterprise wireless solution.
Owen
ac = 5 GHz.
The only dual frequency standard is 802.11n. But that has resolved itself
by now: any router only advertizing "n" is going to be a 2,4GHz only router
and even if you find a rare old model that was 5 GHz "n" it still sucks for
lacking "ac".
In our market everyone delivers "ac" routers by now. One reason for that is
that DSL now needs VDSL2 with vectoring and channel bonding, and this
brings you to a price point where you also want to get "ac" for little or
no extra. Or you are selling high speed internet and the user experience is
simply lacking without "ac".
But 5 GHz usage is still low because people have a ton of devices that are
2,4 GHz only. Even brand new laptops are sold without a 5 GHz radio. People
don't know that they have to check - it is oh but it has wifi and it is
brand new, therefore it must have support for the new standard you are
talking about! Sometimes we have to send someone out to the customer to
demonstrate how crappy his new purchase is.
Regards,
Baldur
That is rubbish. We are using 128 optical splits and 64 users per OLT and a
mix of users buying either 100 or 1000 Mbit/s service. This just works. The
system is very far from being overloaded. We would put even more users on
the OLT if our vendor would allow this (they only support a max of 64 users
per OLT).
Remember the very first thing users do when you sell 1000 Mbit/s internet
is to run a speedtest. Our users do that too and they do get the expected
940-950 Mbit/s (=gigabit ethernet wire speed) speedtest result at all time
of day, also at peak usage.
Regards,
Baldur
Unfortunately almost all of the Internet of Things (IoT) client devices I have come across or purchased lately are 2.4GHz only:
- Belkin Wemo
- Airconsole
- Sense Sleep Tracker
- LIFX
- Ninjasphere (now defunct, but this was interesting because these appear to have a 5GHz radio in them but don't have the antenna to support it)
The explanation I have been given a few times is that the antenna requirements for 5GHz are just too difficult to achieve in what are often small and low powered devices.
We're mostly there with phones and PCs though.
Reuben
We do not sell TV but that means our customers are cable cutters that do a
ton of Netflix, HBO Nordic, ViaSat, SBS, DR TV etc streaming. Our traffic
level per customer is about the double of what others report.
VoIP is not very popular, but people do that too. In either case traffic
levels from VoIP is so low that it is below the noise floor. When you can
get 940 Mbit/s transfer rates with 1 ms latency and no jitter, a single 64
Kbit/s voice stream is not going to be a problem. We point customers to
third party SIP providers and everyone are happy with that.
Do the math: a Netflix HD stream is about 5 Mbit/s. How many such stream
can you have with 2,4 Gbit/s capacity on a GPON OLT? Yes a lot. You might
say but every home has at least 5 TVs now, so with 64 users you need to be
able to do 5 times 64 times 5 Mbit/s (*). But it simply does not work that
way. We are very far from a situation where it works that way. Instead we
monitor the traffic levels, and if sometime in the future the peak traffic
becomes a problem, we are ready to either lower the split ratio or invest
in the next technology (probably some kind of x*10 Gbit/s PON). Until then
we take the cost savings of using a split ratio that works in the real
world.
(*) nobody has a backbone that can cope with that kind of traffic either.
Regards,
Baldur
You might be surprised...
Our upstreams want to simply bypass 40Gbps waves and want us to move
straight to 100Gbps. The cost difference is minimal.
We are set up where each customer can DVR or watch up to 6 shows at once,
per household.
There's a reason Google did 16 way splits, and yes, we have two paths we
are looking at for NG-PON2. One with Calix, another with another vendor.
It is hard to be surprised when you have hard numbers. I run a network and
unsurprisingly know exactly how much traffic my users cause. That number is
currently about 2 Mbit/s peak aggregated per household. Do you need 100
Gbit/s instead of 40 Gbit/s? Yes you do if you carry traffic from more than
20,000 users or perhaps you have 10,000 users but want to plan for expected
traffic increase over the next two years.
But nobody plans their backbone so it can carry 20-30 Mbit/s aggregated per
household. Well if you do, you have no competition, because otherwise
someone else will figure out how to run a network at 1/10 the cost of what
you do, and you will go out of business.
Before someone points out the obvious: That math does not carry over to
GPON OLT planning (too few users for the aggregation). You will have higher
peak than 64x 2 Mbit/s on your OLT. But still, 2.4 Gbit/s shared among 64
users is currently more than sufficient that nobody is going to see any
limits on their download rate, even during peak. And that is with users on
1000 Mbit/s plans.
I have no idea what Google did or why. I have a feeling that my own hard
earned experiences overrides any hear say on that matter... Of course what
I am telling you might also be hear say (although directly from a primary
source) so do what you think is best. I am just sharing our experiences in
the spirit of this forum.
Regards,
Baldur