"... but again, you refer to "some of their Internet customers" and I'm not sure why."
Rear view 20/20, it appears to be a superficial layer of verbiage that would have been, granted, better left avoided. I was differentiating their tunnel-bound Internet customers from the rest of their Internet customers. It's not a major point, in any event.
"The fact that MFNX isn't responsible for creating the redundancy surely doesn't lessen its value to the customer, does it?"
Good question. It doesn't lessen their agreed-to value to the customer if they are able to support what was agreed to in the SLA. It might "appear" to lessen their value if the carrier is supporting r and d on ninety-five percent of a user's intranet, say, but can't support it on the remaining five percent. So, when they can't support r and d along some sections, they simply don't guarantee it until they can get the r and d implemented. Does the last five percent of the network represent the same level of value to the customer as the first ninety-five? You tell me.
There are several other ways of viewing that question, each yielding a slightly different answer, and I'll attempt to address a limited set of design and provisioning permutations as we go along, below.
"I suspect that in arrangements such as this, buying redundancy from another carrier is a preferred option, to improve the separation."
End user IT orgainzations are ultimately responsible for the survivability of their own networks. In many types of designs it actually behooves the end user to use different carriers as opposed to placing all of their eggs in one basket with the same carrier. This is true, not only for financial and SLA leverage, but in order to avoid catastrophic failures <meltdowns> from taking them down hard in an ubiquitous manner, as well.
So, to some extent the added value to the customer that is introduced by the same carrier providing redundancy and diversity can be a subjective call, since it may or may not come into the picture, but would instead depend on some of the variables listed below.
If it's a "managed service," then, yes, value is added when the carrier adds redundancy and diversity. But if the customer is rolling their own mesh (as opposed to subscribing to a managed service, say), then it can become less of an issue, especially if the customer elects to economize by leasing unprotected links from the carrier(s).
This is not to suggest that a redundant/diverse set of routes equipped with automatic protection switching, or some other method of self-healing, doesn't add value if those measures are invoked during cable breaks. But these are usually premium type services that would have to be contracted for and spelled out in SLAs.
What I'm saying, however, is, and to repeat it one more time: Ultimately it's not the carrier's responsibility to ensure the integrity of the end user's network design. Especially if the carrier makes no claims about being able to support features such as diversity and redundancy in the first place, as you have noted.
It's only when the carrier claims to support a managed service that it is their responsibility to assure that such diversity and redundancy exists. If they make claims of diversity and redundancy with guaranteed uptime metrics, or disavow such claims, on a section by section basis as the situation merits, then they are in the clear.
We can look at it from another perspective. If the bulk of the services that a customer subscribes to for their intranet is from a particular carrier and it happens to be a managed GbE or SONET ring service, then that design usually demands that all of the links and POP hardware on those rings are provisioned and supported by the same carrier (or their integrator, or agent). In this instance the obvious answer to the "value" question is: Yes, it does add value, both to the end user, who now only has a single point of contact to contend with, and to the carrier, because it serves to keep the customer within its fold.
Note that this last sentence assumes that the entire managed ring service is self-healing and meets the end user's other performance objectives stated in the SLA, too.
Does this take the end user off the hook, entirely? No, it doesn't. Even the best designs are subject to software-inflicted glitches during upgrades and network management system module add-ons, and other causes of meltdowns, such as plant operating personnel errors as we have seen repeatedly over the years in single-carrier supported architectures.
So, even here, where the carrier has provided redundancy and diversity, the end user might be best served if they distribute their traffic load strategically amongst two or more service providers. It's more expensive to do business this way; it takes more on the part of the customer's staff; but ultimately it can prove to be a more robust and enduring type of design if the end user organization is staffed and equipped to handle it.
As with most other business decisions, it's a matter of analyzing the tradeoffs. |
