The key concern is the ability to provide highly available services.

We do a lot of work for schools, and one school that ranks the lowest in available funds was the last school that I thought would ever get the needed technology upgrades. Thanks to their benefactor--also an owner of a large IT company--that changed, and now the school enjoys the benefits of convergence.

Over the summer, the donor's CTO and COO met with us. The CTO reviewed the proposed network upgrades and made a recommendation. "The point-to-point fiber from closet to closet remains a source for single points of failure." The reduced fiber cable lengths were by design for cost reduction, and we agreed it could become problematic. Since we installed multi-pair fiber between all the closets and not a spoke-to-hub, we offered to patch pairs through to accomplish the same thing.

The initial network is shown below with 10 Gbps fiber between closets. We are only using 1 Gbps transceivers and allowed for growth and expansion with 10 Gbps and multiple pairs.

Before fiber was laid during the summer, we had Category 5 copper drops from the CPU Lab and 4th closet and CPU Lab and Library closet. We back-pulled a Cat5 drop to the Offices closet from another office that connected to the Library. The Cat5 drops served as the primary uplink between switches. Once we set the gear in place we connected fiber and removed the copper uplinks.

With the new cable plant installed, we then proceeded to aggregate switch ports to create more bandwidth to the virtualized server (4 Gbps), between switches in each switch stack at each closet using uplink kits (2.5 Gbps), and the links between switches.

After this round we tested both voice and LAN traffic, and would remove either a Cat5 patch connection or the fiber patch connection to a closet. But this only provided a redundant link while pooling bandwidth. If either the 4th grade closet or the Library closet lost power, then more users would be impacted, since the fiber and copper were not home run to the Computer Lab, and this is what the CTO objected to as a single point of failure concern.

To counter this, of course we installed multi-pair fiber and we terminated all fiber pairs in an ICC modular patch assembly. This allows coupling between fiber cables by simply using a fiber patch cable. An alternative is to use a fiber sleeve that does the same thing. The modular approach allows more flexibility to patch fiber to another switch in the stack of each closet if a switch fails. The fiber transceivers are also modular and easily moved from switch to switch.

Next, we created patched direct connections to each closet from the Computer Lab using available spare pairs in each fiber cable. Again, we aggregated the fiber and copper to create more bandwidth and to allow for redundancy if a fiber link fails or a closet loses power. None of this was an extra cost and it meets needs today and as we expand connectivity to other buildings on campus.

The key concern is the ability to provide highly available services, since we are removing three separate voice and Internet service groups from the campus and providing one network resource. We installed the least amount of fiber by going closet-to-closet then reallocated fiber pairs by patching them to fiber pairs of other cables to create dedicated links. We used a modular approach to facilitate future growth or contraction.

Aggregation between closets, switches in a stack and to servers is a cool feature that will harden your network design. Instead of having a 1 Gbps fiber backbone, we have 2 Gbps with some redundancy and plenty of remaining options to expand.