8 ms·

Yeah because we all know how much of a problem those fake EC2 instances have become.

tyingq 9y ago
Calling hyperthreads "cores" is a little fake. Also, not being really upfront about the egress pricing vs market price for bandwidth. Especially for the bandwidth charges on egress to AWS direct connect. And not nitpicking. I've had to explain these things to non technical leaders many times when they make on-premisis vs cloud comparisons.
- Analog24 9y ago
  And you would consider those two examples enough to doubt their ability to succeed in a completely unrelated field? Unless your answer is "yes", it's nitpicking.
  - tyingq 9y ago
    It's not nitpicking for the specific scope of your comment: "how much of a problem those fake EC2 instances have become" You can broaden the scope of anything and say that some specific criticism is now nitpicking. Edit: A little off topic, but Amazon has failed in unrelated spaces before. Fire phone was a good example. Amazon is giant for sure, but they aren't infallible.
    - Analog24 9y ago
      My comment was a satirical response to the parent comment, giving it pretty clear and specific scope. And the fact that they have failed in some verticals and succeeded in others only strengthens my original point: success/failure in one space has very little bearing on another.
- _msw_ 9y ago
  Hi, Amazon engineer on the EC2 service here. Please let me know if there's some material that calls a hyperthread a core. On our main product detail page for EC2 instances here: https://aws.amazon.com/ec2/instance-types/ https://aws.amazon.com/ec2/instance-types/ look for the text "Each vCPU is a hyperthread of an Intel Xeon core except for T2." See also the "virtual core" information page here: https://aws.amazon.com/ec2/virtualcores/ https://aws.amazon.com/ec2/virtualcores/ and you will see that for most instances the core count is 1/2 of the total number of vCPUs. Direct connect pricing for egress traffic ("Data Transfer Out") is here: https://aws.amazon.com/directconnect/pricing/ https://aws.amazon.com/directconnect/pricing/. Please let me know if there is some other traffic classification that is not metered and billed as you'd expect. As I'm sure you're aware, comparing on-premises vs. cloud is a multidimensional problem. Customer resource to help with TCO calculations and comparisons can be found here: https://aws.amazon.com/tco-calculator/ https://aws.amazon.com/tco-calculator/. Even for a component of overall cost like network transit, not all products are created equally so a simple price comparison may not capture the whole picture.
  - tyingq 9y ago
    I'll amend it to calling a hyperthread a vcpu. People are constantly confused about this. It's not just me. There's very little common terminology across leaders in this space. Think VMware, Digital Ocean, Bluemix, Tier 1 cloud providers, in house data centers, etc. Some talk real cores, others talk hyperthreads. Business and non techie IT people do their own math without knowing the 2X difference. If you put (hyperthreads) next to "vcpus", when appropriate, it would be more clear. "Direct connect pricing for egress traffic ("Data Transfer Out") is here" Attend an AWS sales meeting with a F500 company. It's never mentioned, even when talking about obvious problem areas...like moving a data warehouse to AWS, while leaving the ETL and analytic tools on premisis. People like me are left holding the bad news bag. AWS sales people talk up the "all in" business case where all your stuff is in the cloud. They are suspiciously quiet about the transition period where the 10 to 100x market bandwidth pricing of AWS kills you. That's especially frustrating for AWS direct connect where your global network adds no value. In my mind, AWS direct connect should be competitive with market rates for bandwidth. I understand the premium for general internet connectivity because your global presence and investments add value. For direct connect, it's just plain old gouging. You're penalizing me for not going all-in on day one, which would be risky and stupid.

Calling hyperthreads "cores" is a little fake. Also, not being really upfront about the egress pricing vs market price for bandwidth. Especially for the bandwidth charges on egress to AWS direct connect. And not nitpicking. I've had to explain these things to non technical leaders many times when they make on-premisis vs cloud comparisons.

Analog24 9y ago
And you would consider those two examples enough to doubt their ability to succeed in a completely unrelated field? Unless your answer is "yes", it's nitpicking.
- tyingq 9y ago
  It's not nitpicking for the specific scope of your comment: "how much of a problem those fake EC2 instances have become" You can broaden the scope of anything and say that some specific criticism is now nitpicking. Edit: A little off topic, but Amazon has failed in unrelated spaces before. Fire phone was a good example. Amazon is giant for sure, but they aren't infallible.
  - Analog24 9y ago
    My comment was a satirical response to the parent comment, giving it pretty clear and specific scope. And the fact that they have failed in some verticals and succeeded in others only strengthens my original point: success/failure in one space has very little bearing on another.
_msw_ 9y ago
Hi, Amazon engineer on the EC2 service here. Please let me know if there's some material that calls a hyperthread a core. On our main product detail page for EC2 instances here: https://aws.amazon.com/ec2/instance-types/ https://aws.amazon.com/ec2/instance-types/ look for the text "Each vCPU is a hyperthread of an Intel Xeon core except for T2." See also the "virtual core" information page here: https://aws.amazon.com/ec2/virtualcores/ https://aws.amazon.com/ec2/virtualcores/ and you will see that for most instances the core count is 1/2 of the total number of vCPUs. Direct connect pricing for egress traffic ("Data Transfer Out") is here: https://aws.amazon.com/directconnect/pricing/ https://aws.amazon.com/directconnect/pricing/. Please let me know if there is some other traffic classification that is not metered and billed as you'd expect. As I'm sure you're aware, comparing on-premises vs. cloud is a multidimensional problem. Customer resource to help with TCO calculations and comparisons can be found here: https://aws.amazon.com/tco-calculator/ https://aws.amazon.com/tco-calculator/. Even for a component of overall cost like network transit, not all products are created equally so a simple price comparison may not capture the whole picture.
- tyingq 9y ago
  I'll amend it to calling a hyperthread a vcpu. People are constantly confused about this. It's not just me. There's very little common terminology across leaders in this space. Think VMware, Digital Ocean, Bluemix, Tier 1 cloud providers, in house data centers, etc. Some talk real cores, others talk hyperthreads. Business and non techie IT people do their own math without knowing the 2X difference. If you put (hyperthreads) next to "vcpus", when appropriate, it would be more clear. "Direct connect pricing for egress traffic ("Data Transfer Out") is here" Attend an AWS sales meeting with a F500 company. It's never mentioned, even when talking about obvious problem areas...like moving a data warehouse to AWS, while leaving the ETL and analytic tools on premisis. People like me are left holding the bad news bag. AWS sales people talk up the "all in" business case where all your stuff is in the cloud. They are suspiciously quiet about the transition period where the 10 to 100x market bandwidth pricing of AWS kills you. That's especially frustrating for AWS direct connect where your global network adds no value. In my mind, AWS direct connect should be competitive with market rates for bandwidth. I understand the premium for general internet connectivity because your global presence and investments add value. For direct connect, it's just plain old gouging. You're penalizing me for not going all-in on day one, which would be risky and stupid.