Tuesday, April 29, 2014

Another endowed data service launches

A friend pointed me to a Wired piece on Longaccess, a new endowed data service for archiving personal files. Below the fold I look at their numbers.

Longaccess charges, for their highest 100GB tier, €399 for 30 years, or $550. The underlying storage comes from Amazon; Wired says they use S3, but I doubt they can make the economics of doing so work. Glacier is the appropriate product for their usage pattern. So, assuming they use Glacier and Amazon sticks with the 1c/GB/mo price, storage will cost them $360 over 30 years (2/3 of the capital). They have not bet everything on Kryder's Law and high interest rates, but there are still considerable risks in this offering.

The first is sticker shock. At Fry's today a 4TB Hitachi Coolspin drive is $170, or 4.25c/GB as against Longaccess' price of $5.50/GB or more. A factor of 130 times the media cost is more than 4 times worse than Princeton's 30 times, which was enough to cause serious sticker shock.

The second is lock-in. Suppose Longaccess decides that Glacier is no longer a cost-effective storage option. They need to get their data out of Glacier and in to a competitor. To make the math easy, lets assume they have 10,000 100GB customers, or 1PB of data. They have collected $5.5M in capital and storage is costing them $120K/yr. They have two choices:
  • They can try to get it out for free in GB chunks using Glacier's 5%/mo "free" access. Given the necessary small margin for error it will take about 2 years and cost $50 in request charges and about $3900 in bandwidth charges, or about $0.40 per customer.
  • They can get it out in a month in GB chunks, which will cost $50 in request charges, about $60K in bandwidth charges, and at least $10K in retrieval fees. That is about $7 per customer, or nearly 5 months worth of storage income.
So a competitor is going to have to be a whole lot cheaper to motivate Longaccess to switch quickly. Switching slowly is much cheaper but carries a lot of business risk. Longaccess is not completely locked-in to Amazon, but in practice they would be even if Google succeeds in making cloud storage a two-vendor market.


  1. We have to make a lot of assumptions here, but let's say that storage, transfer and retrieval costs stay the same and Longaccess had to do this transition today. Let's assume Google or Rackspace
    have a competing service that costs less.

    5 months is less than 1.5% of the total storage cost (12 months x 30 years). Which means that a competing service should be at least 1.5% cheaper to make this worth.

    If the reason of the transition is financial, it would be for more than a 1.5% price difference anyway.

  2. Assuming constant costs through time is wrong - endowed storage is only a good business it you believe that the Kryder rate (percentage $/GB/yr drop) is significant. For the effects of on the economics of storage of a significant Kryder rate, see many previous posts such as this one.

    The effect of a move is about a 40% increase in storage cost for that year. For the effect of spikes such as this on the endowment needed see this post. Large unexpected costs in the near years have a large effect, because of both the Kryder rate and Discounted Cash Flow.