Thanks for your previous answer...
But the analysis in “The Business Case for Direct Connect” really pivots on the fact that Amazon has a “data transfer fee” that is substantially reduced when interconnecting with AWS via the “Direct Connect Service Model” as opposed to the “Internet Connect Service Model.”
How does it work for the general case where there is no data transfer fee charged by the cloud company? I would think this to be the more common case.
-- Enrico Pollini
You are spot on.
As a quick aside, it is interesting to me that the “Direct Connect Service Model” I described is essentially peering for enterprise and cloud companies. One difference is that the enterprises and the cloud companies don’t do the peering waltz. If they can swing it, they both simply want to connect with each other, regardless of things like traffic ratios. And unlike the peering world, in the Direct Connect world, they really do want to interconnect with their customers. Interesting twist, no?
But back to your point...
Yes, in our example last month, the sustained 50Mbps of traffic generated 16.2TB per month of traffic resulting in data transfer fees of $1346/mo when connecting over the Internet and only $486/mo when using the Direct Connect Service Model. That is indeed where the savings come from, and the same type of analysis can be applied to Google, Microsoft, and most others that compete against Amazon Web Services.
But the far more interesting case is the one where the cloud company does not provide an incentive to direct connect? When does it make sense then?
For this we need to quantify what we ignored in the previous analysis: the value of the traffic itself. Here we will use the the cost of downtime as a proxy. We must estimate the cost of Internet failure and compare that against the cost of the Direct Connect Service Model.
There are at least a few first order outage costs here: lost revenue, damage to brand, and increased costs. When a mission-critical connection between the two companies does not work for whatever reason, what is the financial downside?
Let’s assume that the firm has correctly identified their “mission-critical” destinations, ones for which an hour of downtime at busy or crisis periods would prevent the firm from completing is mission.
Let’s further assume that your upstream ISP(s) has a one hour outage, one that breaks access to to one of those mission-critical services. Think about that for a minute: What would most likely happen?
Lost Revenue. Think here about customers that can’t use the service and delay their purchases. Consider your lost access to ad bidding networks, payment processing systems, CRM systems, cloud storage systems that hold critical updated data for your sales teams, etc. If your company loses revenue because it can’t access, by definition, mission critical systems, then this the type of lost revenue to calculate. The first question is,
How much revenue is lost when connectedness breaks?
Damage to Brand. When customers perceive poor quality they consider competitor’s services. The outage will be beyond your control, but the customer doesn’t care. You need to win them back or lose them. Consider the number of customers affected and the cost of retaining them and the cost of replacing them.
How much will it cost to replace the lost customers?
Increased Cost. In almost all cases there is follow up work to be done as a result of the outage, particularly in the marketing, sales, and customer services side of the house. Marketing costs will climb as energies are spent on brand repair and messaging. The engineering energies spent on post-mortem and to identify steps to mitigate outages and outage losses moving forward.
How much will the post-outage management efforts cost?
Finally let’s plug in some numbers to illustrate.
Assume one hour of Internet outage prevents your smallish firm from accessing one or more of its “mission-critical” network services. This outage results in revenue loses of about $4000, brand damage of about $2000, and lost productivity, increased meetings to handle the outage its aftermath costs about $10,000. The total cost of this outage is therefore estimated to be about $16,000. If the direct connect service model would have prevented this failure for less than $16,000 for the year, then the direct connect strategy provably makes sense to explore.
It is interesting that the cost of transporting the bits might be in the hundreds of dollars, while the value of that very same traffic, as approximated by the cost of the outage when the traffic can’t be delivered, is in the thousands of dollars. The Direct Connect Service Model looks a little like insurance in this regard.