AWS' new T2 instance type gives IT pros a way to buy Elastic Compute Cloud resources in increments of less than an hour.
The T2 line of instances has a twofold pricing model: by the hour for a percentage of a CPU core and cumulative CPU credits that accrue during slow periods for bursting up to the max performance of a full core when needed.
The new T2 line includes the t2.micro instance priced at $0.013 per hour with access to 10% of a CPU core for baseline performance, which accrues six CPU credits per hour of use; a t2.small priced at $0.026 with access to 20% of a CPU core for baseline performance, which accrues 12 CPU credits per hour of use; and a t2.medium priced at $0.052 per hour with access to 40% of a CPU core for baseline performance and accrues 24 CPU credits per hour of use.
[If] you've got the skills and the analytics and the time to manage it closely, you can save a lot of money.
SVP, Cloud Technology Partners
This pricing system is akin to the braking system in a hybrid vehicle, according to John Treadway, senior vice president with Cloud Technology Partners, a cloud consulting firm based in Boston. When a hybrid car brakes, it stores energy for the engine to use later; as a server runs at low utilization, it stores credits for later bursts of performance.
"[Amazon's] really slicing the capacity of a CPU into smaller chunks," he said. "For a lot of applications which might have fine-grained variability, that could actually be very economical, because you're paying a lot less for the capacity you're actually using, whereas before you could provision capacity, but may not be using all of it."
The previous answer to this elasticity demand was to spin servers up and down, but that can sometimes take too much time, Treadway said.
It will also be tricky for users to figure out when to use these new T2 instances and when to use reserved instances for workloads that have relatively consistent performance demands, Treadway said.
"How do I compare my pricing across clouds when I have this very interesting, innovative, but totally unique and different pricing approach for small instances?" he added. "But if you're ready to do it, and you've got the skills and the analytics and the time to manage it closely, you can save a lot of money."
The T2 line will replace the previous t1.micro instances as well as the m1.small and m1.medium, at a lower price. The older instances will still be supported for existing users, according to Amazon's Previous Generation Instance pricing page. The t1.micro cost almost twice as much as the t2.micro at $0.020 per hour; the m1.small was priced at $0.044 per hour compared with the t2.small at $0.026; and an m1.medium which cost $0.087 per hour can now be replaced with a t2.medium for $0.052 per hour. However, users must also consider the cost of Elastic Block Store volumes, which are the only storage volumes supported with the T2 line of instances.
Users can also get a pair of t2.micro instances, one running Linux and the other running Windows, completely free on the AWS free usage tier, but in general, analysts marvel how AWS is "squeezing a bit more cash out of legacy EC2," as Carl Brooks, an analyst with 451 Research based in Boston, put it.
Amazon is famous for its slim margins, but Brooks sees AWS as a cash cow for the company.
"The organization is just amazing at wringing out profit -- this is like putting a caboose on the gravy train," Brooks said.