Local OM3/4/5 installers...

[Don Y]

How accessible are OM3/4/5 installers in \"typical\" communities?
Yeah, I don\'t expect to find someone local in East Bumphuque, IA
who can do it. But, perhaps a nearby metro area that\'s not too
inconveniently located?

Said another way, how much does using OM3/4/5 over something like
copper inconvenience installations?

(Aren\'t there larger munis that run direct to home? So, the
skills to install have to be available somewhat nearby)

 

[Dan Purgert]

On 2023-07-25, Don Y wrote:

How accessible are OM3/4/5 installers in \"typical\" communities?
Yeah, I don\'t expect to find someone local in East Bumphuque, IA
who can do it. But, perhaps a nearby metro area that\'s not too
inconveniently located?

Said another way, how much does using OM3/4/5 over something like
copper inconvenience installations?

It\'s more of a pain (sometimes considerably). What\'s your end-goal
here?


--
|_|O|_|
|_|_|O| Github: https://github.com/dpurgert
|O|O|O| PGP: DDAB 23FB 19FA 7D85 1CC1 E067 6D65 70E5 4CE7 2860

 

[John Walliker]

On Wednesday, 26 July 2023 at 11:10:53 UTC+1, Dan Purgert wrote:

On 2023-07-25, Don Y wrote:
How accessible are OM3/4/5 installers in \"typical\" communities?
Yeah, I don\'t expect to find someone local in East Bumphuque, IA
who can do it. But, perhaps a nearby metro area that\'s not too
inconveniently located?

Said another way, how much does using OM3/4/5 over something like
copper inconvenience installations?
It\'s more of a pain (sometimes considerably). What\'s your end-goal
here?

Why use OM-anything? Most installations are single-mode nowadays
and there is no significant length restriction unlike multi-mode.
Multi-mode is often cheaper as well. There are connectors that don\'t
need a fusion splicer but involve pushing the fibre end into a gel
interface medium.

John

 

[Don Y]

On 7/26/2023 3:10 AM, Dan Purgert wrote:

On 2023-07-25, Don Y wrote:
How accessible are OM3/4/5 installers in \"typical\" communities?
Yeah, I don\'t expect to find someone local in East Bumphuque, IA
who can do it. But, perhaps a nearby metro area that\'s not too
inconveniently located?

Said another way, how much does using OM3/4/5 over something like
copper inconvenience installations?

It\'s more of a pain (sometimes considerably). What\'s your end-goal
here?

I need some fat pipes to span buildings and campuses
(the latter requiring more isolation than copper typically
affords).

I\'d like NOT to have to come up with several different solutions
to different problems.

I suspect that, in most metro areas, *someone* will have the
skillsets to subcontract this work -- expecting others to have
the ability to handle the shorter (more numerous) copper runs.

But, I\'m not sure how universal that expectation can be; at
what \"settlement density\" do you start to lose easy access
to \"specialists\"?

E.g., my hometown was tiny; you\'d not expect to find anyone
with these skills, there. However, it abutted other more
populous towns that *would* have such skills -- a short
drive away (if \"the next town\" is 20+ minutes away on
an interstate, you probably have slimmer pickings!)

[Note that it\'s not just finding someone who *can* do the work,
locally, but someone who can do it *well*; a piss poor job
is worse than no job at all!]

 

[Don Y]

On 7/26/2023 5:02 AM, John Walliker wrote:

Why use OM-anything? Most installations are single-mode nowadays
and there is no significant length restriction unlike multi-mode.
Multi-mode is often cheaper as well. There are connectors that don\'t
need a fusion splicer but involve pushing the fibre end into a gel
interface medium.

At higher data rates, the single-mode transceivers are pretty pricey.
I\'m not looking to run miles of cable but, rather, get across a building
(use a concentrator to accumulate traffic in one area and move it
to another -- think mesh instead of bus)

 

[John Walliker]

On Wednesday, 26 July 2023 at 15:24:46 UTC+1, Don Y wrote:

On 7/26/2023 5:02 AM, John Walliker wrote:
Why use OM-anything? Most installations are single-mode nowadays
and there is no significant length restriction unlike multi-mode.
Multi-mode is often cheaper as well. There are connectors that don\'t
need a fusion splicer but involve pushing the fibre end into a gel
interface medium.
At higher data rates, the single-mode transceivers are pretty pricey.

That was once true, but now there is hardly any difference at speeds of
up to 10Gbit/s and ranges of up to a few km. (Except of course that
multi-mode fibre is only good for a few hundred metres at 10Gbit/s
due to dispersion.)
Most, if not all, large data centres in the UK use exclusively single mode fibre,
even for short connections. There are good reasons for sticking to
single-mode. It is made in much larger quantities, so it is usually cheaper.
Mixing fibre types in a single link is bad news if there are multiple connectors,
so having only one type in stock avoids mistakes. There is a much better
upgrade path to higher speeds.

I\'m not looking to run miles of cable but, rather, get across a building
(use a concentrator to accumulate traffic in one area and move it
to another -- think mesh instead of bus)

It is possible to buy pre-terminated lengths of single or multi-mode fibre
with LC (or other) connectors. LC are nice because they plug directly
into SFP transceivers.These cables often come with a polymer braid
tube fitted to one end so that the cable can be pulled through ducts
without damaging the connectors. Plenty of suppliers will make these up
to your specified length for a modest extra cost.
Such cables can also be ordered with stainless steel tape armour if you
want to discourage rats from chewing them. There are plenty of
Chinese suppliers of such things who really push the costs down.
If you want to minimise the number of terminations, then it is now economical
to use dual wavelength transceivers so that only one fibre is needed
for a bidirectional link. Ubiquiti sell these at very reasonable prices for
both 1Gbit/s and 10Gbit/s. They are only rated to work on single-mode
fibre as far as I know, although for short links haviong all the same type of fibre
from end to end it is remarkable what you can get away with. I have a pair
of 1Gbit/s Ubiquiti bidirectional transceivers which only cost GBP20 + tax for the pair.
If you are buying switches with SFP sockets, check carefully about compatibility.
Some manufacturers will let you plug any make of SFP module in, and
they will invariably work. Others, like HP, restrict you to using only their own
modules and charge much higher prices. Some low cost switches and media
converters do not even implement the i2c interface, so they have no way of
knowing what make of SFP you are using. Others implement it and tell you what
you have but don\'t get in your way.

John

 

[Don Y]

On 7/26/2023 9:37 AM, John Walliker wrote:

On Wednesday, 26 July 2023 at 15:24:46 UTC+1, Don Y wrote:
On 7/26/2023 5:02 AM, John Walliker wrote:
Why use OM-anything? Most installations are single-mode nowadays
and there is no significant length restriction unlike multi-mode.
Multi-mode is often cheaper as well. There are connectors that don\'t
need a fusion splicer but involve pushing the fibre end into a gel
interface medium.
At higher data rates, the single-mode transceivers are pretty pricey.

That was once true, but now there is hardly any difference at speeds of
up to 10Gbit/s and ranges of up to a few km. (Except of course that
multi-mode fibre is only good for a few hundred metres at 10Gbit/s
due to dispersion.)

I\'m looking at relatively short distances (~100-300m) but much higher
data rates (40-100GBase).

Most, if not all, large data centres in the UK use exclusively single mode fibre,
even for short connections. There are good reasons for sticking to
single-mode. It is made in much larger quantities, so it is usually cheaper.
Mixing fibre types in a single link is bad news if there are multiple connectors,
so having only one type in stock avoids mistakes. There is a much better
upgrade path to higher speeds.

Data centers have endpoints on expensive pieces of kit. Even if I put
*100* devices on a single fiber, that\'s only ~$1500 of kit supported by
that fiber. (if I put 40, then $600 or $375 for 25). The cost of the
transceiver becomes a large fraction of the materials cost.

(The labor to run cables/fiber is essentially fixed, regardless -- and paid
to a subcontractor so more under the customer\'s control and not part of my
invoicing)

I\'m not looking to run miles of cable but, rather, get across a building
(use a concentrator to accumulate traffic in one area and move it
to another -- think mesh instead of bus)

It is possible to buy pre-terminated lengths of single or multi-mode fibre
with LC (or other) connectors. LC are nice because they plug directly
into SFP transceivers.These cables often come with a polymer braid
tube fitted to one end so that the cable can be pulled through ducts
without damaging the connectors. Plenty of suppliers will make these up
to your specified length for a modest extra cost.

But I\'d either have to do a site-survey to determine exact lengths
or standardize on some (number of) specific lengths (?)

I don\'t have \"equipment closets\" in which to hide cable service loops
(imagine if a data center used all the same length cable to connect
each server to it\'s switch, regardless of location in rack)

Such cables can also be ordered with stainless steel tape armour if you
want to discourage rats from chewing them. There are plenty of
Chinese suppliers of such things who really push the costs down.
If you want to minimise the number of terminations, then it is now economical
to use dual wavelength transceivers so that only one fibre is needed
for a bidirectional link. Ubiquiti sell these at very reasonable prices for
both 1Gbit/s and 10Gbit/s. They are only rated to work on single-mode
fibre as far as I know, although for short links haviong all the same type of fibre
from end to end it is remarkable what you can get away with. I have a pair
of 1Gbit/s Ubiquiti bidirectional transceivers which only cost GBP20 + tax for the pair.
If you are buying switches with SFP sockets, check carefully about compatibility.
Some manufacturers will let you plug any make of SFP module in, and
they will invariably work. Others, like HP, restrict you to using only their own
modules and charge much higher prices. Some low cost switches and media
converters do not even implement the i2c interface, so they have no way of
knowing what make of SFP you are using. Others implement it and tell you what
you have but don\'t get in your way.

The endpoints are custom kit. I route based on process ID (which can
migrate -- i.e., the host that originated a transaction may not be
the same host that receives the *response* from that transaction) so
\"connections\" aren\'t like typical network usage. (more like highly
dynamic OpenFlow)

 

[Dan Purgert]

On 2023-07-26, Don Y wrote:

On 7/26/2023 3:10 AM, Dan Purgert wrote:
On 2023-07-25, Don Y wrote:
How accessible are OM3/4/5 installers in \"typical\" communities?
Yeah, I don\'t expect to find someone local in East Bumphuque, IA
who can do it. But, perhaps a nearby metro area that\'s not too
inconveniently located?

Said another way, how much does using OM3/4/5 over something like
copper inconvenience installations?

It\'s more of a pain (sometimes considerably). What\'s your end-goal
here?

I need some fat pipes to span buildings and campuses
(the latter requiring more isolation than copper typically
affords).

If it\'s not far (i.e. > a few hundred meters in a run) -- check out the
pre-terminated stuff from fiberstore (fs.com). I\'ve used it to good
effect between my garage and house (and for some small businesses).
While bare fiber is easier to pull, the LC connectors aren\'t huge
either.

Alternatively, look into microwave solutions from Mikrotik or Ubiquiti
-> they\'re used by WISP operators, so have some pretty decent bandwidth.
Chances are, WISP guys might also know some fiber installers.


--
|_|O|_|
|_|_|O| Github: https://github.com/dpurgert
|O|O|O| PGP: DDAB 23FB 19FA 7D85 1CC1 E067 6D65 70E5 4CE7 2860

 

[Dan Purgert]

On 2023-07-26, Don Y wrote:

On 7/26/2023 9:37 AM, John Walliker wrote:
On Wednesday, 26 July 2023 at 15:24:46 UTC+1, Don Y wrote:
On 7/26/2023 5:02 AM, John Walliker wrote:
Why use OM-anything? Most installations are single-mode nowadays
and there is no significant length restriction unlike multi-mode.
Multi-mode is often cheaper as well. There are connectors that don\'t
need a fusion splicer but involve pushing the fibre end into a gel
interface medium.
At higher data rates, the single-mode transceivers are pretty pricey.

That was once true, but now there is hardly any difference at speeds of
up to 10Gbit/s and ranges of up to a few km. (Except of course that
multi-mode fibre is only good for a few hundred metres at 10Gbit/s
due to dispersion.)

I\'m looking at relatively short distances (~100-300m) but much higher
data rates (40-100GBase).

As I recall, 100G over multimode tops out around 200 meters, then you
need to get 100GBASE-DR (~500m over SM). Could be wrong though, it\'s
been a while since I brushed up on the distance limits. Worst case,
you\'re ready for if/when you need more bandwidth (equiv. of spending the
extra cash for cat6a 10-15 years ago...)

It is possible to buy pre-terminated lengths of single or multi-mode
fibre with LC (or other) connectors. LC are nice because they plug
directly into SFP transceivers.These cables often come with a polymer
braid tube fitted to one end so that the cable can be pulled through
ducts without damaging the connectors. Plenty of suppliers will make
these up to your specified length for a modest extra cost.

But I\'d either have to do a site-survey to determine exact lengths
or standardize on some (number of) specific lengths (?)

\"Exact\" being \"ehhh, this\'ll need about 100 meters\".

I don\'t have \"equipment closets\" in which to hide cable service loops
(imagine if a data center used all the same length cable to connect
each server to it\'s switch, regardless of location in rack)

Where\'s your kit, if not in a closet?


--
|_|O|_|
|_|_|O| Github: https://github.com/dpurgert
|O|O|O| PGP: DDAB 23FB 19FA 7D85 1CC1 E067 6D65 70E5 4CE7 2860

 

[Don Y]

On 7/28/2023 5:48 AM, Dan Purgert wrote:

On 2023-07-26, Don Y wrote:
On 7/26/2023 9:37 AM, John Walliker wrote:
On Wednesday, 26 July 2023 at 15:24:46 UTC+1, Don Y wrote:
On 7/26/2023 5:02 AM, John Walliker wrote:
Why use OM-anything? Most installations are single-mode nowadays
and there is no significant length restriction unlike multi-mode.
Multi-mode is often cheaper as well. There are connectors that don\'t
need a fusion splicer but involve pushing the fibre end into a gel
interface medium.
At higher data rates, the single-mode transceivers are pretty pricey.

That was once true, but now there is hardly any difference at speeds of
up to 10Gbit/s and ranges of up to a few km. (Except of course that
multi-mode fibre is only good for a few hundred metres at 10Gbit/s
due to dispersion.)

I\'m looking at relatively short distances (~100-300m) but much higher
data rates (40-100GBase).

As I recall, 100G over multimode tops out around 200 meters, then you

200m in a building is pretty big building! It\'s also \"far enough\"
to connect nearby \"out buildings\".

need to get 100GBASE-DR (~500m over SM). Could be wrong though, it\'s
been a while since I brushed up on the distance limits. Worst case,
you\'re ready for if/when you need more bandwidth (equiv. of spending the
extra cash for cat6a 10-15 years ago...)

I\'ve already factored a 10X bandwidth buffer into the design.

It is possible to buy pre-terminated lengths of single or multi-mode
fibre with LC (or other) connectors. LC are nice because they plug
directly into SFP transceivers.These cables often come with a polymer
braid tube fitted to one end so that the cable can be pulled through
ducts without damaging the connectors. Plenty of suppliers will make
these up to your specified length for a modest extra cost.

But I\'d either have to do a site-survey to determine exact lengths
or standardize on some (number of) specific lengths (?)

\"Exact\" being \"ehhh, this\'ll need about 100 meters\".

That means having to figure out where to stash the \"few meters\"
that you overestimated. How often do you see a col of (e.g.) #12AWG
wire stashed somewhere because it was \"cut too long\"?

I don\'t have \"equipment closets\" in which to hide cable service loops
(imagine if a data center used all the same length cable to connect
each server to it\'s switch, regardless of location in rack)

Where\'s your kit, if not in a closet?

Factory floor (ceiling), etc. Or, in facilities that didn\'t
originally *plan* on needing a space for such infrastructure.

(E.g., doing the open-floorplan house -- without an attic *or*
a basement! -- has been a real eye opener!)

 

[Don Y]

On 7/28/2023 5:33 AM, Dan Purgert wrote:

On 2023-07-26, Don Y wrote:
On 7/26/2023 3:10 AM, Dan Purgert wrote:
On 2023-07-25, Don Y wrote:
How accessible are OM3/4/5 installers in \"typical\" communities?
Yeah, I don\'t expect to find someone local in East Bumphuque, IA
who can do it. But, perhaps a nearby metro area that\'s not too
inconveniently located?

Said another way, how much does using OM3/4/5 over something like
copper inconvenience installations?

It\'s more of a pain (sometimes considerably). What\'s your end-goal
here?

I need some fat pipes to span buildings and campuses
(the latter requiring more isolation than copper typically
affords).

If it\'s not far (i.e. > a few hundred meters in a run) -- check out the
pre-terminated stuff from fiberstore (fs.com). I\'ve used it to good
effect between my garage and house (and for some small businesses).
While bare fiber is easier to pull, the LC connectors aren\'t huge
either.

I\'ve been hoping to leave the details of \"installation\" up to
\"the locals\". E.g., you wouldn\'t worry about running *copper*
for a piece of kit (power, comms) as *someone* nearby -- even in
East Bumphugue IA -- would likely have that skillset.

The concern is how \"exotic\" is running/terminating fiber in
those off-the-beaten-path places?

Alternatively, look into microwave solutions from Mikrotik or Ubiquiti
-> they\'re used by WISP operators, so have some pretty decent bandwidth.
Chances are, WISP guys might also know some fiber installers.

I\'ve played with a few offerings for between-building applications
where there are distinct advantages of NOT having to run a physical
medium. But, *inside* buildings they don\'t seem as applicable.