Friday, June 8, 2018

Medium to large campuses.... how do you handle OOBM and redundant fiber runs?

My campus has ~750 IDFs. Some buildings have multiple IDFs - some buildings have only one.

My question is twofold....


How do you handle OOBM? If my organization were completely enclosed in a large office building, our density would be greater, and it would be cheaper to do OOBM. However, we have quite a few buildings that have 3-4 users in it - if that. So, we don't have a very dense network in terms of users to IDFs. The cost to implement a proper OOBM system is.... prohibitive to say the least.

Here are our options that we see now:

  • OpenGear ACM7004-5 w/ SFPs - $925 per IDF, total of approx $700,000. But - we now need an additional pair of fiber to each IDF (see the next problem I have)
  • OpenGear ACM7004-2-LMV (cellular only) - $1,015 per IDF, total of approx $765,000, with an added regular occurring cost for cellular service for 750 devices (and in my line of work, CAPEX is easier to get than OPEX).
  • OpenGear ACM7004-5-LMA (SFP + cellular) - $1485 per IDF, total of approx $1.1 million. And... we now have the issue of additional fiber PLUS the cellular service.
  • Establish an isolated network for OOBM, switches in each MDF (we have six), and use media converters + fiber to transport to the IDFs. I got a rough estimate for a decent media converter, it's $500 - so, $375,000. And... we still have the problem of not enough fiber. (See below)

In addition, none of these cost estimates address what we're going to do with the MDFs (we have six, two of which are 'data centers' as well as a core network node).


Next up, fiber. We have VERY few IDFs that have fiber laid to more than one MDF. For most of them, if we wanted to connect a switch to multiple MDFs, we'd have to build a fiber path through the first MDF to get there. So, rough numbers, if 1/6 of the switches go to an MDF (we have six MDFs), and we want to hang each switch off of both MDFs in the pair, that means:

  • 125 pair/250 strand coming from IDFs, into MDF A that is destined for MDF A
  • 125 pair/250 strand coming from IDFs, into MDF A that is destined for MDF B
  • 125 pair/250 strand coming from MDF B into MDF A to connect IDFs from MDF B to MDF A
  • 125 pair/250 strand coming from MDF A into MDF B to connect IDFs from MDF A to MDF B

This leaves me with a fiber requirement for 500 strands between the MDFs in the MDF pair, and 500 strands coming from IDFs. This doesnt even address the issue of the fiber at the IDFs. Here's an example of a particularly troublesome building:

  • 18 stands coming from MDF-A (in building 2) to IDF-A (in building 1234)
  • 6 stands coming from IDF-A (bldg 1234) to IDF-B (same building - 1234)
  • 6 stands coming from IDF-A (bldg 1234) to IDF-C (same building - 1234)
  • This building also has a separate network housed in it for a tenant organization - they are using 4 strands for their switches, and we are using 2 for ours (with no redundant path). Therefore, we have used up every single fiber in that building, and have no more to run redundant links.

Obviously, the real solution is to run more fiber. This is VERY cost prohibitive. We do own the entirety of the land, and we do have a manhole/duct bank system - but the labor costs are intensive, as well as a very long project to accomplish. We'd be looking at probably a many millions to a billion dollars to completely overhaul it, and it'd likely take ten years.

So - that leaves me with other workarounds, such as bidirectional optics from the IDF to the MDF (allowing us to have two connections per pair, rather than only one). Then, a DWDM system to connect the MDFs together to reduce fiber requirements there.


Any thoughts on either?



No comments:

Post a Comment