The shtetl "Pionernet" through the eyes of a technician. Part two

    So, the technician’s notes about the realities of the small town “Pionernet” continue.
    The first part turned out to be interesting, although with a bias in the "organizational problems."
    I will try to fill the deficit and recall the technical aspects of the work. In particular: the organization of the overall network structure, the modernization of segments without decommissioning, the laying and termination of optical cables.
    Welcome to cat:


    Network structure.


    Perhaps it is worth noting that the city is relatively modern, the number of storeys in the building is mainly 5-9-16 floors. Geographically consists of two distant "macro-districts", divided into "micro-districts" of 20 - 40 houses.
    When I first started working, the optical backbone network was made in the form of two rings spanning the “macrodistricts”, with one point of presence on the “microdistrict”, from which chains of stupid switches on a copper twisted pair dispersed from house to house. The connection between the "macro-districts" was organized by the leased fiber of one of the "opsos".
    If you try to recall the scheme, then somehow it resembled the scheme of the Moscow metro, from which a part was removed within the circle line, and looked something like this:
    image
    At that time, we wanted to get a stable, mainstream star, from which only then to organize a smart access level.

    First stage. How it was.

    The optical network itself was terrifying. Multimode fiber in places already crumbled from old age, terminated by splice plates. Long probros could easily be made from several pieces of cable of different fibers, manufacturer and even used. Couplings at welding sites were rare - basically the joints were packaged in a cassette and stuck in a “beer bottle” suspended upside down.
    The end in the drawers also did without crosses. The cable, fiber, splice plates and ponytails were tied with twisted pair pieces to the plywood, and went into the switch. At points that were considered “exemplary”, a cassette was used instead of plywood.
    The situation was quite normal when a single-mode eight-fiber two-module cable departing from a point, on the way, in a bottle, turned into a single-module, then, in another bottle, it diverged into two multi-mode 4 fiber cables - one went further into the ring, and another one of some organization ...
    Naturally, this structure has not suffered from quality of life and stability since birth.

    There were no optical connection and crossover schemes. We collected fragmentary information about the cable structure, which were stored in the heads of installation and engineering, and began to think.
    At first, a week without getting out of the rooftops, on a satellite image of the city in the “SAS planet”, we figured out the routes for the approximate cable route, climbed into all the points and bottles, as we could track the crossover. Then they forced the authorities to order a used reflectometer and welding machine.
    The thoughts ended with the purchase of a bay of 8 fiber single-mode and single-module cables, couplings, 19 inch boxes for equipment, crosses and a heap of optical passivation.
    A month later, they received the equipment: an Anritsu reflectometer and an Ericsson FSU 905 welding machine, which to this day have served us faithfully. A month later, the materials arrived.
    From that day, “Decisive No” was voiced for bottles and splice plates.

    Laying a new structure.

    Together, we decided that we would hang 8 fiber cable in separate sections along the route of the old rings, and, as we lay them, unfold the rings into stars, with collection at the aggregation points. In total, 15 rays of the star were counted in one center and 11 in another, the rings were opened at the opposite diameter from the center of aggregation, and each point was turned on via its WDM-SFP in its fiber.
    They did as follows: In the center, two ports with ordinary duplex SFPs were allocated - in them, with a new fiber, they included an old ring that was decreasing as it worked, and 16 ports with WDM modules. The cable in the center of aggregation was welded into a cross, and the first fibers were plugged into six ports of the switch, the last pair was busy supporting the old structure.
    Then, next to the existing one at the ring point, we mounted a new box and an electrician for transferring equipment, weighed a cable section, scalded, transferred the equipment, switched the point to one fiber in a new cable, the remaining fibers were crossed through, except for the last pair that went into the old cable and closed the decreasing ring.
    Both structures worked at the same time, only one grew and the other contracted:
    image
    When there were three points in the ring, and the last two fibers were used on each side, without disturbing them, laid, terminated and crossed the last sections of the new cable, then engineers and installers We entered the night shift, and in 5 hours of a busy schedule, we moved and switched all three points.

    The second “macrodistrict” was redone in the same way. There we did without night work, because the star got 11 rays, and 16 fibers were enough to narrow the ring to two points, and then open the ring and switch them to 1 fiber. We even left a reserve of 4 fibers, well, he doesn’t pull his pocket. True, in the first "macro-district" we did not have any spare fibers, because 15 of them were used in the star, and one more on each side we used to connect aggregation centers, but this problem is now being solved.

    What happened.

    So we got two logical stars on the very same rings that were in the beginning. But now each point was completely independent of the performance of the rest, equipped with uninterrupted power, optical cross, plus we got rid of “genies in bottles”. In addition, they made their own link between the aggregation points, leaving the rented fiber as a hot reserve.
    image
    The downtime of the point and the micro-district depending on it on average 2-4 hours is quite an acceptable result, given the volume and complexity of the work.
    As they were laid, the old sections, and the boxes where the equipment was located, were dismantled. The removed pieces of optics, organoleptically and by trace, were checked for professional suitability, and sent to the warehouse. Boxes and the best part of the cables were later used to replace copper house communications.

    Second phase. Access level.


    After putting the stars into operation, we practically did not have problems on the highways. To this day, inconvenience is caused only by long power outages, which uninterruptibles cannot cope with, but this is apparently chronic.
    But with the growth of the network and the number of subscribers, other problems began to emerge - stupid twisted pair switches could not cope with the load of one and a half to two hundred customers in the segment. There were clients who organized their own “driveway” networks, with rings and Broadcast storms from all sorts of Borg and ice chats.
    There is a need to solve customer access problems.
    To do this, it was necessary to "multiply" at home. And at the same time do FTTB. For one or five access houses, they decided to put one switch in the middle of the house, in the attic. In multi-entry houses - one switch for five entrances.
    It was decided to connect the switches into rings, which should be closed to the rays of the main star.

    For tests and testing the technology, we chose one of the most densely populated, and therefore problematic areas, consisting of three huge round houses with 25-30 entrances, and several 3-4 family entrance dormitories. According to rough estimates, it turned out 23 points, in fact it turned out 25.
    In a couple of months we bought a cable, the same as for the highways - I really liked it, the switches were dlink, according to the decision of the authorities, 3526 models for houses and 3028G for a star ray, and we started installation.
    It was physically impossible to close the cable into the ring in any way, so they laid it in both directions to the extreme points, welded it up, but turned on the switches in an overlap.
    Link, leaving the root switch to the first horn, did not fall on the next one in order - but through one from it and so on to the end of the “horn”. There, he turned around and walked back through the missing switches of the first half, past the root, to the other end of the horn. There he turned around again and on the passes in the second horn he returned to the root.
    image
    The resulting ring, according to RSTP, is opened by one of the switches in the middle, which, in the event of a signal loss on one side, is deployed to the other, but if it turns off itself, it simply physically breaks it.
    The circuit is simple and stable. When de-energized on one switch, we lose only it. We risk losing the entire ring only if you disconnect two adjacent points at the same time. But this in my memory was once, when half of the neighborhood turned off.
    The transfer of customers was carried out gradually, after the commissioning of the entire ring. As they moved, the copper structure was dismantled, all customers were reduced to smart equipment. In part, it was involved, so we could not immediately turn it off, we did the same as with the optics in the old ring. When customers were taken away from the previous switch, the next one was turned on with a temporary cable from the optical one closest to it.

    All the work took two months, the correction of the jambs of the installation team allowed during the transfer is about another month. After that, the microdistrict from the incident generator turned into one of the quietest harbors of the network. For the second year, 5-7 emergency tickets a month have been received from there, no more.
    As a result, according to the results of the installation and a couple of months of operation, the circuit was approved for further use.
    In the same way, we remade another 9 microdistricts in the first “macrodistrict”, and 7 in the second. The average ring length is from 12 to 30 switches. Considering that the rings are mostly passive and only pass BPDU packets, the convergence time is about 5 seconds.

    In one microdistrict, the scheme had to be slightly changed. The development did not allow one ring to cover the entire microdistrict. Therefore, a small ring of 12 switches was made covering half of the microdistrict, and a separate cable was made to seven remote houses from the rest. They purchased a media converter chassis, and a Dlink 3100-24 switch. Each of the switches on these houses turned on through a bunch of SFP - converter in 3100-24. The result was a kind of regional "mini star".
    But the peculiarity in this section is not in the circuit, but in the method of unpacking the cable. Since it was laid in one piece with through factories in the places of the future installation of equipment, it was decided not to cut it at the points, but to get one fiber each. As a result, two centimeter incisions were made directly on the cable at a distance of 40-50 cm and the module was opened. In one incision, the desired fiber was searched for and snacked, after which the incision was sealed. The cut fiber was carefully removed into the second cut and terminated. The place of the cut and the supply of fiber were removed into the cassette, and the trailer went into the switch.
    The work is complex and almost jewelry. I spent a week and a half to break up seven points, but it was worth it. When in the future, when repairing the roofs, our cable was bent almost 180 degrees, and judging by the trace of the free fiber, in this place, even after the extension, there was a decent signal drop - none of the points stopped working. If there were 5-7 connectors on the way, it would hardly have been so rosy.

    Future plans.


    There are always enough fronts, plans for the summer to transfer four more micro-districts to FTTB (Building), start changing the ring in one of the “macro-districts” to 24 fiber cable, plus the authorities plan to connect a couple of cottage villages in the vicinity of the city to the network. In addition, recently entrepreneurs have become more active citizens who in their dreams see unlimited gigabit Internet directly to the office, which does not reduce our hassle.
    According to general results, over the past year we had one emergency trip to repair optics, which was broken by a tree that fell in a thunderstorm, while three years ago we went to emergency heels once a month - the bottle was flooded after rain, the signal disappeared, and other troubles were welcomed us almost weekly.
    The average number of emergency tickets for the week, in the last two years, if it does not fall, does not grow exactly, this is with a three-fold growth in the subscriber base. In addition, the initial diagnosis by ticket now often comes down to viewing the log and the status of the switch port. Which is much simpler and faster than leaving the place of the emergency or assembly team.
    Intra-district rings and trunk lines, designed for 1Gbps pass, are now loaded at peak hours by 30-45%, so we have room to grow, and if possible, the network can work for a couple of years, although it is already preparing an adequate change ...

    More plans, notes on copper house communications, and the interests encountered in setting up the equipment.

    Also popular now: