Amazon EC2 Instances and cpuinfo

Amazon EC2 provides 5 different instance types and publishes compute capacity as EC2 Compute Unit. But what do they really mean as actual cpu’s reported to the operating system?

Its quite easy to find out. /proc/cpuinfo provides  information about CPU and their speed. In the table below I have attached the output of more /proc/cpuinfo for each instance type and the actual proc report on cpuinfo so you can see how tthe OS treats them from a physical CPU standpoint.

Instance type Amazon compute units and physical cpu cpuinfo
Small Instance 1 EC2 Compute Unit (1 virtual core with 1 EC2 Compute Unit) 1 physical core @2.6 GHz small cpuinfo
Large Instance 4 EC2 Compute Units (2 virtual cores with 2 EC2 Compute Units each) 2 physical cores @ 2.0 GHz large cpuinfo
Extra Large Instance 8 EC2 Compute Units (4 virtual cores with 2 EC2 Compute Units each) 4 phyiscal cores @ 2.6 GHz extra-large cpuinfo
High-CPU Medium Instance 5 EC2 Compute Units (2 virtual cores with 2.5 EC2 Compute Units each) 2 physical cores @ 2.4GHz high-cpu-medium cpuinfo
High-CPU Extra Large Instance 20 EC2 Compute Units (8 virtual cores with 2.5 EC2 Compute Units each) 8 physical cores @ 2.4 GHz high-cpu-extra-large cpuinfo
Share and Enjoy:
  • Digg
  • Sphinn
  • del.icio.us
  • Facebook
  • Google
  • Live
  • MySpace
  • description
  • Technorati
  • TwitThis
  • Yahoo! Buzz

11 thoughts on “Amazon EC2 Instances and cpuinfo

  1. Hi
    The information is very useful. I would also like to get cpuinfo of the “High Double memory extra large” and “High quadruple memory extra large” instance which are newly instance types introdused by the AWS. Please help me to get this.
    Thanks
    Arpit

  2. This is completely wrong. What you are providing is the info of the physical core. A small instance get assigned only a virtual core of that physical core.

  3. Pingback: Comparing cloud providers’ pricing models to EC2 … “The devil is in the details”

  4. This is interesting, however I must wonder how relevant the output of /proc/cpuinfo is. For example, let’s say it tells us the actual info on the CPU we are running on. But it doesn’t tell us if we are sharing that physical CPU with anyone.

    If Amazon believes they can deliver 2 Compute Units on a single core, and you are only paying for a single Compute Unit, do you really think you have exclusive access to that CPU?

    The only real way to tell is to run some comparative benchmarks.

  5. Of course the most reliable way is to measure the performance of your application or run a benchmark that mimics closest your applications workload on the given hardware. This information is relevant e.g when tuning a multi-threaded application but no more than that.

  6. Pingback: [转载]What are Amazon EC2 Compute Units? « Feisky

  7. So, from the above, it looks like an Amazon EC2 “virtual core” maps to a physical processor (or, more accurately, has a physical processor underlying it) and an ECU maps to a slice of a physical core. The % of a physical core that is allocated to an ECU depends on the speed of the processor. An ECU is equivalent to a 1.0 – 1.2GHZ Opteron; all of the above processors are dual-core CPU’s running in the range of 2.0-2.6GHz so I’m guessing that on ECU is roughly 50% of a single core. Small, Large and Extra Large instances seem to support this breakdown but the high CPU instances which have 2.5 ECU’s confuses things…

  8. Pingback: Comparing cloud providers’ pricing models to EC2 … “The devil is in the details” | zen smith

  9. http://aws.amazon.com/ec2/instance-types/#instance-details says

    *1 CC2 and CR1 Instances are backed by 2 x Intel Xeon E5-2670 processors, eight-cores with hyperthreading
    *2 CG1 Instances are backed by 2 x Intel Xeon X5570, quad-core with hyperthread plus 2 NVIDIA Tesla M2050 GPUs
    *3 HI1 Instances can deliver more than 120,000 4 KB random read IOPS and between 10,000 and 85,000 4 KB random write IOPS (depending on active logical block addressing span) to applications. The maximum sequential throughput on is approximately 2 GB/s read and 1.1 GB/s write.
    *4 HS1 Instances can deliver 2.4 GB/s of 2 MB sequential read performance and 2.6 GB/s of sequential write performance.
    *5 Instances launched into the same cluster placement group are placed into a non-blocking 10 Gigabit ethernet network

Leave a Reply

Your email address will not be published. Required fields are marked *

*

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>