To start, realize stating that a dedicated 1gbps link is $2000 may be true but it does not justify the price, which is what I'm speaking to.
What observations I had were: 1) the price of dedicated links went up in time disproportionate to their capacity; 2) network operators began moving to a per-bit charging model, forcing users to bid against each other, and; 3) at least here, non-DC dedicated links can be had much more cheaply than a link in a DC.
There is scarcity but for practical purposes the scarcity is so far off that I'm not sure it makes sense to consider it. The true scarcity is related to the availability of land and the size of the data links, neither of which are yet very limiting. Regulation related to right of way (or more commonly) city involvement in high level planning could be blamed for creating market failures if these costs really are that high.
Otherwise, in regards to #1, we would expect a downward trend of dedicated network link price while capacity goes up, following storage density and transistor density (or more appropriately computational throughput). But we don't see this, even though throughput does increase, and integration costs decrease.
For #2, going to a bidding model could imply scarcity, but as there are unjustified monopolies often there is no increase in capacity so it seems prudent to assume this is done solely to increase price.
And with #3, well, certain usages are priced higher than others, also implying some kind of bidding. This might reflect space scarcity but seeing as the price tends to not go down even if the customer provides their own network equipment and pays for its space it just seems like a reluctance of the DC operator to perform upgrades.
The fairest model I can imagine has users paying the amortized cost of the network equipment and its installation and maintenance.
What observations I had were: 1) the price of dedicated links went up in time disproportionate to their capacity; 2) network operators began moving to a per-bit charging model, forcing users to bid against each other, and; 3) at least here, non-DC dedicated links can be had much more cheaply than a link in a DC.
There is scarcity but for practical purposes the scarcity is so far off that I'm not sure it makes sense to consider it. The true scarcity is related to the availability of land and the size of the data links, neither of which are yet very limiting. Regulation related to right of way (or more commonly) city involvement in high level planning could be blamed for creating market failures if these costs really are that high.
Otherwise, in regards to #1, we would expect a downward trend of dedicated network link price while capacity goes up, following storage density and transistor density (or more appropriately computational throughput). But we don't see this, even though throughput does increase, and integration costs decrease.
For #2, going to a bidding model could imply scarcity, but as there are unjustified monopolies often there is no increase in capacity so it seems prudent to assume this is done solely to increase price.
And with #3, well, certain usages are priced higher than others, also implying some kind of bidding. This might reflect space scarcity but seeing as the price tends to not go down even if the customer provides their own network equipment and pays for its space it just seems like a reluctance of the DC operator to perform upgrades.
The fairest model I can imagine has users paying the amortized cost of the network equipment and its installation and maintenance.