General information about Amazon’s publicly available AWS services:
developer’s guide with a comprehensive overview of ec2 components
includes Elastic IP
mapping of virtual compute unit to physical compute resources, SLAs
(contains a nice “architecture” graph, connecting EMR with S3)
Dynamo (key-value store)
“always on” key-value store that emphasizes availability and conflict resolution at the cost of consistency.
Describes the slightly more relaxed notion of consistency that AWS uses to enable high throughput and low latency for huge numbers of user requests/updates
An Evaluation of Amazon’s Grid Computing Services: EC2, S3 and SQS
Academic look at performance, utilization, pricing/use model of AWS. Provides an overview of some use cases, the API, and limitations (from the programmatic standpoint, legal perspective). Evaluation is based on throughput and availability.
Impact of virtualization on performance of EC2
Measures impact of virtualization on the performance of EC2 - concludes that in general, there are some significant issues with virtualizing EC2. also briefly considers the possibility of scientific computing in the cloud and conclues networking performance may not be sufficient.
Scientific Workflow Applications on Amazon EC2
Performance of clouds can be comparable to typical high-performance compute machines, and although EC2 doesn't have access to resources like high-performance local IO, once data is in the cloud it can be quite advantageous.