GENIUS Network

From RealityGrid

Jump to: navigation, search

[edit] GENIUS Lightpath network

Various Lightpath links will be provisioned for HemeLB cross-site runs along with providing a means for high-bandwidth low-latency visualisation and steering. The figure below outlines the GENIUS Lightpath links (in red) we are establishing.


Image:GENIUSLightpathNetwork.jpg


There is currently a 300Mb/s connection from UCL to the TeraGrid, which will be upgraded to 1Gb/s. Instead of two seperate links to TeraGrid and LONI from UCL, we are planning to have a single Lightpath from UCL to Chicago via Starlight. From there we can connect to the TeraGrid, along with a connection to the EnLIGHTened testbed, and therefore the LONI resources.

The link between the two NGS2 nodes is intended for cross-site HemeLB runs, with the link back to UCL/NH used for steering and real-time visualisation.


Lightpath Status

The following table describs the status of the various Lightpaths (under construction).

Lightpath Link Endpoints Status
UCL - NGS2/Manchester UCL NGS2/Manchester Application submitted to JANET
NGS2/Manchester - NGS2/Oxford NGS2/Manchester NGS2/Oxford Application submitted to JANET
UCL - HPCx UCL HPCx Application submitted to JANET (300Mb/s connection already in place)
UCL - TeraGrid UCL TeraGrid To be submitted shortly (300Mb/s connection already in place)
UCL - LONI UCL LONI To be submitted shortly

Optical Network Subnets

The following table describes the subnets used by the various networks.

Testbed Subnet Comments
NGS2  ? Does this include UCL?
LONI 172.30.0.0/16 (Also use, for Japan - 172.18.0.0/16)
TeraGrid  ? Unknown

[edit] GENIUS Workflow

A high-level overview of GENIUS work-flow would involve the following steps.

1. Acquisition of cerebral MRI data from a patient at the NHNN.

2. Anonymisation of MRI data and placement on a data server.

3. Transfer of MRI data from NHNN to local (or remote) HPC facilities.

4. Extraction of cerebral-vascular structure from MRI data and construction of LB boundary conditions.

5. HemeLB bloodflow simulation is run on a HPC resource, where the rendering of the high-resolution dataset is rendered as this proceeds. A steering mechanism allows the clinician, at a remote location, to interact with the simulation (viewpoint parameters, model parameters, boundary conditions, etc), giving real-time feedback.

Steps 3 and 5 require the availability of a high-speed network between NHNN and the UCL HPC cluster, for example.

Data Acquisition

MRI datasets are currently acquired at a raw resolution of 512x512, with 100 slices. Each pixel is stored as a 16 bit value, resulting in a dataset size of 50MB. New data available towards the end of 2007 using new MRI hardware (?) will consist of a 1024^3 raw voxel volume, resulting in dataset sizes of 2GB. A number of development are taking place at NHNN which...


- Two new angiography suites (capable of 3D rotational angiography, in which a volumetric image is created by obtaining many 2D projections of the brain during rapid injection of intra-arterial contrast into the carotid arteries).

- A new 64 slice CT-scanner capable of rapid acquisition of volumetric data during intravenous injection of contrast into a peripheral vein.

- Two new MRI scanners capable of generation of MRA (magnetic resonance angiography) images, also able to be processed volumetrically.

... to be completed.


Remote Visualisation and Steering

To be added.

Personal tools
projects