Comment by mdasen | Hacker News Reader

Comment by mdasen | Hacker News Reader

As a basis of comparison:

For $40/mo, you can get a 1.7GB server from EC2 with one year commitment or $31/mo with a 3 year commitment.

With the Rackspace Cloud (formerly Mosso), you can get a 1GB server for $44/mo or a 2GB server for $88/mo.

Likewise, a 7.5GB EC2 instance is now $161/mo with one year commitment and $125/mo with 3 year commitment.

Rackspace charges $350/mo at their 8GB level.

15GB EC2 - $322/mo 1-year; $250/mo 3-year. 15.5GB Rackspace - $700/mo.

So, that's a substantial savings over the provider most like EC2 - bandwidth a la carte, SLA, hourly pricing - if you're able to commit.

The sad part of this is that it only applies to newly purchased EC2 reserved instances. If you bought a reserved instance yesterday, you're in at the old pricing. And that's kind of crappy of Amazon. That's my hesitance in locking in for a 3-year term. As technology progresses, you're locked in. Granted, the rates are wonderful, but it would be nice if Amazon promised that we wouldn't get passed by on changes like this.

> If you bought a reserved instance yesterday, you're in at the old pricing. And that's kind of crappy of Amazon

You're not thinking it through. Suppose the cost of the machine declines over time as a series over 5 months: 9,8,7,6,5. They could charge you those numbers each month, or they could charge you $7 each months on upfront commitment. Either way they recoup the cost and you are paying the same.

However, with monthly plan you have try to predict hardware price declines in your business plan and you bear the consequences of getting it wrong. With commitment plan, you can lock in the expected savings immediately and put it into your business plan. If Amazon gets it wrong they eat the cost, if they get it right they win.

This is a classical risk swap, and it allows you to focus on your business, not tracking the changing environment.

The one thing that all of these GB comparisons never take into account is actual performance (which is admittedly harder to measure).

I've experimented with about 4 different dedicated virtual server type hosts (none of which was EC2), and found differences in page rendering time from 30ms to 180ms, due to things like how much load their servers were dealing with, speed of hard drives, and CPUs used.

Some tests I've seen elsewhere suggest that Mosso has better Write IO than Linode and EC2, while Linode has better read performance than Mosso and EC2.

I guess what I'm saying is that if you want to use a virtual server and performance (rather than just the ability to handle a large number of users) is important, you have to consider more than just the RAM and bandwidth.

Granted, the rates are wonderful, but it would be nice if Amazon promised that we wouldn't get passed by on changes like this.

And I want a pony! ;-)

Ever considered how they recoup that discount? Partly by not passing through their savings to you for the duration of that commitment.

Regardless the "pay later" aspect is attractive for many customers, especially when you're in a startup situation where early cash is the most expensive.

Reply on news.ycombinator.com