SONET simply means you are on a Sonet ring

you wish you can easily find yourself on a linear stub and, even if you are on a ring, you need to ensure o trench diversity o that it is actually configured to fall over o ... randy

On Thu, 2005-04-14 at 16:15 -0400, Luke Youngblood wrote:

SONET simply means you are on a Sonet ring: Two redundant connections to the central office. If someone gets a little crazy with a backhoe your line is guaranteed to stay up (ask about SLAs, and make sure they will refund part of your monthly bill if you have an outage). That's why it costs over twice as much.

It is worth mentioning that the value of SONET redundancy has a lot to do with your location to the ring. SONET's value is proportional to the available space for backhoes between your location and the point where 1 trench splits into two trenches, if this ever occurs, prior to hitting the SONET ring. For instance: If you are more likely to experience digging near your location, then SONET may not really be worth the cost as nearby trenching activity could just as easily knock you off the ring. SONET is very good if you are worried about the provider's metro network confronting a backhoe. Does anyone have any good statistics on SONET outages? -Jim P.

On Thu, 14 Apr 2005, Valdis.Kletnieks@vt.edu wrote:

On Thu, 14 Apr 2005 16:15:41 EDT, Luke Youngblood said:

...

SONET simply means you are on a Sonet ring: Two redundant connections to the central office. If someone gets a little crazy with a backhoe your line is guaranteed to stay up (ask about SLAs, and make sure they will refund part of your monthly bill if you have an outage). That's why it costs over twice as much.

And remember to ask questions - make sure they've actually got the two connections routed differently. Remember that if the backhoe hits the conduit, *all* the fiber pairs go - and if both runs were in the same conduit, you're still dead....

(Anybody here *NOT* seen cases where the 2 fibers leave the building on opposite sides, go down different streets - and rejoin 2 miles down the way because there's only one convenient bridge/tunnel/etc over the river, or similar?)

yes. but in my case we checked it and it was okay on install but was rerouted at some point. someone broke the ducting and we lost a bunch of oc48s which was bad. you'll never get better redundancy than having more than one carrier. Steve

On Thu, 14 Apr 2005, Randy Bush wrote:

...

you'll never get better redundancy than having more than one carrier.

very often, they buy segments from eachother, run in the same trench, change the dlr six months later, ... trust and test just as you would a single carrier. trust 0 test 100.

when this has been critical i've insisted on seeing duct maps to ensure they really are different. but as you say even that doesnt prevent a change of path in the future i guess it helps us to know a bit more than the average circuit buyer tho too.. Steve

On Thu, 2005-04-14 at 17:12 -0400, Patrick W Gilmore wrote:

On Apr 14, 2005, at 5:07 PM, Stephen J. Wilcox wrote:

...

when this has been critical i've insisted on seeing duct maps to ensure they really are different. but as you say even that doesnt prevent a change of path in the future

i guess it helps us to know a bit more than the average circuit buyer tho too..

I've seen duct maps which "proved" the fibers were truly diverse, but somehow the circuit died with a single backhoe strike.

The way to somewhat solve this problem is to buy a 12 pack (or two) for the tech(s) that are setting up your circuit. Get to know them and ask them about the pathing and what they know about common conduits, tunneling, railroad tracks and/or watermains subject to being dug up on the weekend, etc. Whatever you do, don't trust the sales or support people to know. They mean well, but their interests aren't in keeping their butt in a warm bed in the middle of a winter weekend night. -Jim P.

On 4/14/05, Stephen J. Wilcox <steve@telecomplete.co.uk> wrote:

you'll never get better redundancy than having more than one carrier.

One carrier can often tell if two circuits they're providing you are on the same route, and can develop processes for building circuits that are not only installed diversely, but don't get rerouted onto the same path. Doesn't mean that *every* carrier can do this , or that the ones who can always do it reliably, but it's a start. Two carriers can almost never tell if that's the case, even if neither of them is buying services from the other or from a common third carrier. Nor are they good at developing processes together. They are less likely to have common equipment failures, or common operator failures, but they're more likely to have common backhoe targets. Also, while it may be possible to get diversity by deploying two circuits at the same time, that doesn't mean that you can get diversity by deploying one circuit and later deploying another. A common problem is that routes ABCDEFGZ and AHIJKLMZ are diverse, but the first circuit you buy is on route ABCKLMZ and there's no way to build diversity with the leftovers. Another common problem is that routes ABCDEFGZ and AHIJKLMZ are both more expensive than ABCKLMZ, and AHINOPQRSTUVWXYZ is much more expensive, and you've awarded the initial purchase to the lowest bidder and expect others to be competitive. ---- Thanks; Bill Note that this isn't my regular email account - It's still experimental so far. And Google probably logs and indexes everything you send it.

Michael.Dillon@radianz.com writes:

On the contrary, you get better redundancy by sticking to one carrier and making sure that they really provide separacy though the entire span of the circuit. If you have two carriers running fibre to yoiur building down the same conduit, then you do NOT have separacy and as a result, the redundancy is not there.

The problem with this theory is that one carrier is completely free to reroute your connectivity among its resources. Two carriers can certainly move from being diverse relative to each other to sharing common infrastructure, but if it costs money it's much less likely to happen. Note that many carriers, though perhaps not the LECs, will answer questions about the underlying resources they are using if they are sufficiently motivated, but you have to reask every now and again to make sure that the answers are still satisfactory.

Of course, you can get separacy with two carriers but it is generally more work to verify that the two companies do not share fibre or conduit or tunnels.

Well, then it's a question of how much one cares about diversity. If it's important, it may be worth some effort. If it's really important, then you try harder to control the infrastructure in question directly, which can mean anything from constructing your own underground facilities to leasing someone elses - at least to get out of the building. If it's really, really important, then you pick the building based on the selection of providers and the redundancy offered by the set.

(Anybody here *NOT* seen cases where the 2 fibers leave the building on opposite sides, go down different streets - and rejoin 2 miles down the way because there's only one convenient bridge/tunnel/etc over the river, or similar?)

Even if that's not the case, and it's still perfectly separated all the way to the CO, the CO is a common point of failure. Granted, the failure modes are very unlikely to occur for a CO, but they do exist. Those two separate paths of the ring have a way of always coming together somewhere, by design. The only way to insure that doesn't happen is to have two sources of connectivity to a building, from two separate local carriers that have fiber going in two opposite directions (eg., one carrier to the east, one to the west), to two opposite area codes/LATAs that get transit from two different transit providers that have POPs in cities that are geographically the furthest apart (one to the north, one to the south, or east west, or whatever). As long as everything keeps heading in complete opposite directions, it becomes very assured that the common modes of failure diminish with distance. This tactic works, and works well with IP using BGP, but it's something that would be beyond my scope of expertise to attempt to implement with anything else. (someone mentioned earlier charging the 2 9's rate for providing 5 9's service...... it was a wake up call to myself.....I'm that guy!) On a somewhat related, but kind of a little off topic note: It always makes me chuckle inside to hear data centers tout their "dual grid connections" as a way to insure that the power "is hardly ever interrupted" Same basic principal. Sure they might be separate distribution feeders, and they might even come from separate distribution substations, and the subtransmission that feeds the distribution substations might even come from separate transmission substations... but within about a minimum of a 60-100 mile radius, it's nearly always connected together by the transmission grid. Now, if there was a data center that had a power feed connection to say, ERCOT, the Eastern Interconnection, and the Western Interconnection.... THAT would be something to brag about.

Speaking on Deep Background, the Press Secretary whispered:

(Anybody here *NOT* seen cases where the 2 fibers leave the building on opposite sides, go down different streets - and rejoin 2 miles down the way because there's only one convenient bridge/tunnel/etc over the river, or similar?)

A friend works for Uncle Sam, and he is required to audit the routes used on certain high-availability systems, to insure the diversity we pay for. (It's in the contracts with the carriers..) He describes it as a long drawn-out exercise in futility. A non-trivial employee has to spend eons on the task. It's a recursive onion peeling, or a data version of Tom Lehrer's "I Got It From Agnes"... And once done... the errors found, the diversity restored, and the report signed off; it's soon worthless...because the carriers soon shuffle things around Yet Again. And I agree re: the building entrance issue and later choke points. Anyone recall the time several years ago that most of the Valley was isolated? One route was across the ?Bay? Bridge; it was down for planned maintenance when backhoe fade struck around San Jose. How many paths is enough? -- A host is a host from coast to coast.................wb8foz@nrk.com & no one will talk to a host that's close........[v].(301) 56-LINUX Unless the host (that isn't close).........................pob 1433 is busy, hung or dead....................................20915-1433

Michael.Dillon@radianz.com writes:

In my opinion, the following rule of thumb is reasonable.

1 path is enough for a site/enterprise that shuts down its services evenings and weekends.

2 paths is enough for a site/enterprise that provides a 24 hour, 7 day per week service.

3 paths is enough for a population center with under a million inhabitants.

5 paths is enough for a population center with over a million inhabitants.

And a very few population centers such as New York, London, Tokyo, and Cheyenne Mountain should probably have more than 5 paths.

Given that anything larger than a single enterprise has no central coordinating body, how is it useful to say how many paths is "enough" for a city of any size? Service providers will build as many paths as make commercial sense, whatever that may be, and if customers have opinions and are willing to back it up with money, they should express those opinions to their providers.

On Sun, 17 Apr 2005, Jay R. Ashworth wrote: So here's the 64GB/s question: If carriers are being paid to ensure physical separation between circuits for the life of the circuit, why is it that they haven't implemented change management systems (and I don't solely mean the software) to ensure they they *can* (not even that they will) manage to ensure such separation? Simple math. The cost of the occasional SLA credit and/or circuit regrooming when the customer discovers a non-diverse path where one was specified is obviously much less than the cost of tracking, maintaining ( and surely providing ) path diversity. Surely large providers have spent a lot more time and money developing processes and software that allow them to groom circuits into the least number of physical paths possible. Or at least I would, if I were paying for the facilities. matto --matt@snark.net------------------------------------------<darwin>< The only thing necessary for the triumph of evil is for good men to do nothing. - Edmund Burke