Brain Imaging Center

page 1 of 6

continue to Demilitarized Zone (DMZ) Services

Warning: This is a work in progress. As such, this document is incomplete and (perhaps), in part, incorrect. Please send corrections to millman@socrates.berkeley.edu.

Network and Computer Services: Background

The Helen Wills Neuroscience Institute at the University of California at Berkeley brings various methodological and disciplinary approaches to bear on the problems of neuroscience. One aspect of this integration focuses on building three Technology Centers that will provide the tools used to understand how the brain works: the Brain Imaging Center, the Neurogenomics Center, and the Molecular Imaging Center.

BIC

The Brain Imaging Center (BIC) at UC Berkeley is a new facility dedicated to functional Magnetic Resonance Imaging (fMRI) research. fMRI is a non-invasive "neuroimaging" technique used to discern the blood-flow correlates of neural activity.

There are three main functional components on the BIC network: 1) the MRI scanner, 2) the servers, and 3) the workstations. A simplified view of how researchers are intended to use these units is: first, data is produced at the scanner; second, data is stored on a central file server; and finally, the data is processed and analyzed at the researcher's personal workstation.

The centralized computing system for the BIC currently supports about 70 users, and is expected to experience considerable growth over the next several years. This system will manage a large number of MRI image data sets (each about 300 MB each). Users can access their personal data and run software from any workstation over a dedicated and secure high-speed Local Area Network (LAN).
LAN

Due to the high volume of data to be processed and the spatially-distributed nature of our computational resources, we established high speed connections between computers in different buildings. Our LAN consists of two Cisco Catalyst 6500 switches, which are connected with optic fiber and communicate at Gigabit Ethernet speed (1000 Mbits/sec, full-duplex). Workstations are connected to the network at Fast Ethernet speed (100 Mbits/sec, full-duplex). Servers can be connected to the network with Gigabit or Fast Ethernet.

Our need to have data securely transported over this network dictated that we isolate our LAN from the Internet. We did this by insulating the BIC computers from the rest of the world with a firewall. The introduction of a firewall to our computer system usefully divides our system into a number of interesting subcomponents, which I discuss in turn below.
Linux

Finally, one of the self-imposed, overarching "constraints" framing the decisions we made in setting up the laboratory is to use a basically Linux driven computer network. A natural question for the reader, then is: "Why are we using Linux?" There are many different possible answers. Linux is fast, stable, secure, and robust. Additionally, the most useful software for the analysis of fMRI data runs on a variant of UNIX, and much of that runs on Linux.

But the most interesting, if not ultimately the most compelling, reason is that it is open source. In particular, Linux is licensed under the Gnu Public License (GPL). A recent editorial in the British Medical Journal calls for the use of free software like Linux in the National Health Service. It can be found at: Medical software's free future. Another important and more general discussion about open source software is a collection of essays by leaders and innovators in the Linux community. Open Sources. Additionally, is The Cathedral & the Bazaar. This is also online at: Eric's Random Writings.

I begin by describing what services we provide to the external world in our Demilitarized Zone (DMZ), which like its name implies is largely unprotected. Next, I explain what our firewall can and can't protect us from as well as briefly describing how it does this. I then discuss what services we provide internally. In the penultimate section, I outline some of the configuration issues involved with conecting clients to our network as well as briefly listing some of the more important software used for the analysis of fMRI data. I conclude with a brief discussion about future plans for our computer systems.

Acknowledgement. Some of this equipment was purchased by a generous donation by the Webb Family.