Since the email Amazon sent out doesn't seem to be reflected anywhere on the web:

--------

We are pleased to announce the immediate availability of Amazon EC2 T2 instances. T2 instances are a new low-cost, General Purpose instance type that are designed to provide a baseline level of CPU performance with the ability to burst above the baseline. With On-Demand Instance prices starting at $0.013 per hour ($9.50 per month), T2 instances are the lowest-cost Amazon EC2 instance option and are ideal for web servers, developer environments, and small databases. T2 instances are for workloads that don't use the full CPU often or consistently, but occasionally need to burst to higher CPU performance. Many applications such as web servers, developer environments, and small databases don't need consistently high levels of CPU, but benefit significantly from having full access to very fast CPUs when they need them. T2 instances are engineered specifically for these use cases.

T2 instances are available in three sizes: t2.micro, t2.small, and t2.medium and work well in combination with Amazon EBS General Purpose (SSD) volumes for instance block storage.

T2 instances are backed by the latest Intel Xeon processors with clock speeds up to 3.3GHz during burst periods. If you are currently using T1 or M1 instances, we encourage you to try T2 instances for better performance for many applications at a lower cost.

T2 instances are available in the US East (N. Virginia), US West (Oregon), EU (Ireland), Asia Pacific (Singapore), Asia Pacific (Tokyo), Australia (Sydney), and Brazil (Sao Paulo) regions and are coming soon to all other AWS Regions. T2 instances can be purchased as On-Demand and Reserved Instances. To learn more about Amazon EC2 T2 instances, see the Amazon EC2 details page.

--------

Also, at this moment these new instances don't seem to be available yet, at least not in EU.

So, with T2 instances and General Purpose volumes, Amazon is now officially in the overselling business. Kudos to them for finding a way to do it fairly. Looks like extra tough time ahead for lowendbox sellers.

As zorked noticed, these are HVM only, which is why I didn't see the option for my paravirtual AMI's.

I've never used anything but paravirtual, so I'm kinda stumped as to why this is.

These are actually quite interesting instance types. I can see the use for these.

How do Amazon handle the capacity planning internally? With a fixed resource allotment it seems pretty easy to figure out. Just divide the total capacity by the number of guests on an instance. With this setup presumably either, there are less guests on a host (to allow for more headroom), or if everyone bursts at the same time you don't actually get any additional capacity.

I suppose Amazon have a huge amount of data regarding how machines are actually used. They must have a fairly good idea of how much of the available power is used on average / other usage patterns that emerge. Would be very interesting data to look at.

It also begs another question. If you have seamless access to this additional capacity once you've earned the credits, could they introduce a system where you could have pay as you go access? I guess it's not that easy because at least the credit system puts a limit on how much each guest can burst.

Official announcement:

New Low Cost EC2 Instances with Burstable Performance

http://aws.amazon.com/blogs/aws/low-cost-burstable-ec2-insta...

Some more information here on the new t2 instance types:

http://aws.amazon.com/ec2/instance-types/

"T2 instances are Burstable Performance Instances that provide a baseline level of CPU performance with the ability to burst above the baseline. The baseline performance and ability to burst are governed by CPU Credits. Each T2 instance receives CPU Credits continuously at a set rate depending on the instance size. T2 instances accrue CPU Credits when they are idle, and use CPU credits when they are active. T2 instances are a good choice for workloads that don’t use the full CPU often or consistently, but occasionally need to burst (e.g. web servers, developer environments and small databases). "

Managing EC2 instances is becoming like filing taxes. Soon you will need EC2_Instance_Advisors for your company needs.

Seems like they already thought about that:

https://aws.amazon.com/premiumsupport/trustedadvisor/

Trusted adviser on amazon payroll, to help reduce cost of paying to amazon. #Conflict_Of_Interest

Two interesting facts: these instances are HVM only and... VPC only, weirdly enough.

I thought everything was VPC now? With a default on your account that everything is assigned to unless you create / choose another one.

Indeed. I had to delete all RDS and EC2 instances before I could remove the default VPC.

Most of the new instances are HVM only. PV vs HVM has become quite complicated but thankfully we have Brenden Gregg to straighten it out for us[1].

[1]: http://www.brendangregg.com/blog/2014-05-07/what-color-is-yo...

Non-VPC is being phased out.

where does it say VPC only?

It's not immediately apparent what the changes are

I think it's disappearance of t2.micro and addition of t1.micro, t2.small and t2.medium?

Not much information published about this yet, I wonder if Amazon sent the email out a little early and the blog post etc aren't ready yet?

At first glance I was pretty disappointed that I just reserved all my t1's last week. After seeing that the new instances are EBS only, I don't regret that decision.

It will be interesting to see how these new boxes handle after the CPU burst is over. T1 micro's were notoriously unreliable because of the CPU limitations.

These ones have specified baseline performance, and even that shouldn't be too much slower than the maximum performance on the instances they are replacing. So that plus the accumulation of credits over the entire day rather than a much smaller period for the old micro should make things much better.

AFAIK, last generation t1.micro instances were also EBS-only.

True, but the m1.small had instance storage as an option. Since there is no m3.small and EBS is out of the question, it looks like I'll be moving my m1.smalls to m3.mediums.

Why is EBS out of the question? Do you have an horror story to share? :)

I agree there are some specific uses cases ideally suited for ephemeral storage. But in most cases EBS gives you quicker provisioning times, snapshotting and of course persistance. In the past, performance and reliability sucked but AFAIK this is not an issue anymore, especially with the new General Purpose (SSD backed) EBS volumes.

"If you are planning to do this (I certainly am), don't forget that the T2 instances do not include any local (instance) storage and that you'll need to use one or more EBS volumes instead."

I find that bit a bit of a shame as we deploy using local instance storage only and avoid EBS.

Why do you avoid EBS? Cost?

The approach I am taking to EC2 instances is that I can kill it any time. So fire up instance, fetch latest code zip from S3 (not from Github), and deploy that code. Runs smoothly.

Keeps it very simple and effective, I can even ramp up the number of instances then reduce them to retire old instances.

EBS adds another point of failure (and probably the weakest link in the Amazon stack) and overcomplicates deployment.

EBS has historically had a few issues, and the performance is fairly poor. I've personally never had any issues with them in the 4 years that I've used AWS.

Interesting, it looks like the t2 instances aren't available as spot instances.

Compare that page with the 'previous generation' page: http://aws.amazon.com/ec2/previous-generation/

to see that there are new instance types, and they're cheaper.

There are very clearly new instance types on that list: t2.micro, t2.small, t2.medium.

Yeah but not everyone gets what these changes are, though they are listed in the comments section above.