~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Click to download the slides:
[PDF] OpenCirrus: An Open Cloud Computing Testbed
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Today, I will introduce an open cloud computing testbed: OpenCirrus. I found this testbed just the day before yesterday when I was wondering waht to talk today. I don’t know about this platform
entirely, but it’s so similar with PlanetLab. I wonder if we can use it for some potential experiments on cloud computing. So I show it to you here.
OpenCirrus was first established in 2008, and provides a single testbed of heterogeneous distributed data centers, for systems, applications and services research. This note is the presentation for OpenCirrus Summit by managing director at HP last year.
This is the outline of this talk, mainly including some general information about OpenCirrus, and also some recent research examples.
The OpenCirrus is established and sponsored by HP, Intel and Yahoo! with additional support form NSF. It aims to spur innovation in systems and applications research. Some interesting datasets may also be collected for sharing. It mainly provides global services
about signing on, monitoring, and store. OpenCirrus currently has 15 sites all over the world and more than 100 nodes for each site. This is the distribution map of the sites, mainly including some universities, such as CMU and companies like HP, Intel and
Yahoo!. China Mobile and China Telecom also join to this testbed system.
The OpenCirrus has several objectives. The most important one is to foster systems research around cloud computing. And also expose research community to enterprise level requirements. OpenCirrus is a kind of federation of heterogeneous datacenters, to support
your research and data sets collection.
This page shows the governance model for OpenCirrus. Each site runs its own research and technical teams, just similar to PlanetLab.
This is the snapshot of the website of OpenCirrus. You can request access to the OpenCirrus here. Different with PlanetLab, you can apply for access to OpenCirrus even if you are not a member of this system.
Here is the Cirrus cluster topology at HP labs. We can find that there are 256 nodes containing 1024 cores, 3328GB (giga) for RAM and 632TB (tera) for storage, all connected by 1Gb/s or 10Gb/s high bandwidth link.
Intel research cluster also has 155 nodes in its data center.
UIUC cluster also has many compute nodes and storage nodes with high performance link.
This is the detailed information for some sites, about the # of cores, servers, memory size, storage size, network bandwidth and research focus for each site.
Each access should be based on the ssh gateway for security. But web services are also supported on port 80 or 443.
The OpenCirrus software stack contains may layers, say, IT infrastructure layer, cloud infrastructure services, virtual resource sets and application services. The PRS service, Tashi and Hadoop is ongoing projects of HP.
But how can we get access to OpenCirrus sites? The Project primal investigator should apply to each site separately. We can have their emails on their website. Users are able to login to each site with the same username and password.
The OpenCirrus is seeking research in these areas, including datacenter federation, datacenter management, web services, data-intensive applications and systems, Hadoop map-reduce applications.
We can see its success form the growing number of sites, developers and users.
There are many testbeds for research, such as IBM/Google cluster, TeraGrid, PlanetLab, EmuLab. This is a comparison between them. For example, EmuLab is for network emulation and PlanetLab is for system and networking research. In addition, the distribution
scale and participants are different.
Here are lists of research projects with OpenCirrus of HP and Intel, such as Wikipedia mining & tagging at application level, Tashi at infrasture service level and OpenNet at infrastructure level. OpenCirrus is a programmable open layer-2 network. It has been
installed on OpenCirrus cluster at HP.
GENI is Global Environment for Network Innovations. OpenCirrus has been accepted to be a GENI project.
Next, here lists some research examples related to OpenCirrus. Sparql query over hadoop for very large RDF datasets, N-tier application benchmark and evaluation over Open Cirrus.
Next is the cloud sustainability dashboard. We can see that the cost in this table, when the price is low, the average entry is green, otherwise, it’s red. The objective is to express and display run-time sustainability of cloud and cloud services.
Here we can get some successful experience on OpenCirrus. For example, he said our NSDI paper would not have happed without Open Cirrus.
Finally is the summary, the OpenCirrus is growing and has enabled a lot of research. Also OpenCirrus summits have been hold for 6 times.
Thank you!