Cloud computing now offers a variety of resources for scientists. In addition to a large number of commercial clouds, several projects make resources available to scientific communities. This page details some of those projects.

Many clouds have been configured in academia. Many are available only to specific projects and communities. Those listed below provide general access provided via allocation mechanisms of various projects or on a pay-for basis.

fg-logo-lg The FutureGrid Project supports several clouds, distributed among five sites, in aggregate providing the capacity of over a thousand cores. The FutureGrid clouds are configured with Nimbus, OpenStack and Eucalyptus all of which support interfaces that are roughly compatible with AWS EC2/S3, allowing users to move between clouds relatively easily. Both KVM and Xen are supported over a variety of instance. Read More…

Access to the FutureGrid clouds is possible for members of the research community participating in active FutureGrid projects.

RedCloud Red Cloud is a cloud from Cornell University based on Eucalyptus 3. Red Cloud is accessible through the AWS EC2/S3 API. The physical infrastructure is based on 8 servers (96 cores total) with 384GB of RAM and 10 Gb Ethernet. It uses the KVM hypervisor. Five instance types are provided with 1 to 12 CPU cores, 4GB to 48GB of memory, and 20GB to 1TB of disk. Additionally, 7.5 TB of storage are accessible via 10Gb iSCSI through EBS volumes of up to 1TB in 1GB increments.

Access to RedCloud is provided on a subscription basis, starting at $500 for Cornell and $750 for other academic users for the equivalent of one CPU core for one year, plus 50 GB of storage (extra charge for additional storage). There are no network charges.

Wispy at Purdue

Wispy is a Nimbus cloud operated by the Purdue University as part of the XSEDE project. As such, it is accessible through the AWS EC2/S3 API. Its physical infrastructure consists of eight nodes, each with 16-core Intel Xeon E5-2670 processors (128 cores in total), 32GB of memory, and Gigabit connectivity. It uses the KVM hypervisor and Nimbus version 2.8.

Access to Wispy is available to researchers with an XSEDE allocation.

SDSC Cloud Storage

SDSC Cloud Storage provides an object-based storage system based on OpenStack Swift, with access through the Rackspace and AWS S3 APIs. It features continuous automatic data verification, maintenance of at least two on-site replicas, and off-site replication. Objects are accessible via the web. It advertises peak transfer rates of up to 8GB/sec, and also provides high performance NFS and CIFS storage.

Access to the SDSC Cloud is provided on a pay-for basis. Pricing is different for UCSD members, academics, or industry partners. There are no network charges.

AWS_LOGO_CMYK The Amazon Elastic Compute Cloud, part of Amazon Web Services, was launched in August 2006 and is now recognized as a leader among the commercial IaaS providers. Users can select among up to 18 instance types grouped into 10 families. The CPU power of each instance is measured in “EC2 Compute Units” equivalent to CPU capacity of a 1.0-1.2 GHz 2007 Opteron or 2007 Xeon processor. Instance types range from 1 to 88 EC2 Compute Units. Amazon EC2 is based on the Xen hypervisor. Hardware-assisted virtualization (HVM) is used for cluster compute instances, high I/O instances, second generation (m3) instances, and Windows instances.On-demand instances are priced per compute hour; the instance prices vary per type of instance. Users can also use spot instances with market-based prices, or buy subscriptions via reserved instances which can then be resold on the Amazon EC2 Reserved Instance Marketplace. Users also pay for storage (via EBS or S3) as well as for data transferred between availability zones, between EC2 regions, and sent to the Internet. Transfer of incoming data is free. For more information see the AWS website.

Windows Azure is Microsoft’s entry in the Cloud Computing space. In addition to Infrastructure as a Service, they also provide Platform as a Service, and website hosting services. Windows Azure provides five instance types for its users. These range from small VMs with a shared processor and 768 MB memory all the way to 8 core 56 GB machines. Windows Azure is built on the Windows Azure Hypervisor a customized version of Microsoft Hyper-V. It can run Linux and Windows virtual machines.

Instances are priced on an hourly basis, or discounted on six or twelve month plans. Bandwidth and support are charged separately.

Rackspace Cloud provides Infrastructure as a Service with Rackspace Cloud Servers, a file storage and download service, as well as other services like block storage, database, and load balancing. Rackspace provides seven instance types ranging from 1 virtual CPU instances with 512 MB memory to 8 virtual CPU instances with 30 GB memory. Rackspace VMs are based on Xen for Linux and XenServer for Windows virtualization. Rackspace Cloud can run Windows, FreeBSD, or a variety of Linux distributions.

Rackspace Cloud instances are priced by the hour with extra charges for bandwidth or Windows or Redhat Enterprise Linux instances.

Google Compute Engine is Google’s Infrastructure as a Service platform. Google Compute Engine provides a variety of instance type from single CPU instances to 8 core 52GB of memory instances. Google Compute Engine uses KVM to provide virtual machine instances. Users can use Google Compute Engine Linux or Centos images.

Google Compute Engine instances are priced by the minute, with additional charges for persistent disk and network transfer.

Resources