Mac OS X

From GridInfo

Apple systems have been noticeably absent from the UK eScience scene, in spite of some significant advantages of the Mac OS X operating system. This article is a report on the extent to which Apple systems can be used within a production-level grid environment.

Apple’s operating system: Mac OS X

Mac OS X is build upon BSD unix, with an interface that more-or-less hides the unix layer from the user1. However, access to the unix layer is provided through a standard shell interface and an installable X11 interface. The Fink project (http://fink.sourceforge.net/) brings a range of open-source unix software to the OS X environment (including compilers and editors). Most standard languages are available for OS X, including java, python, perl, php, C, C++ and Fortran2. Additionally, OS X comes with standard web packages such as Apache.

At the interface level, OS X is capable of running most productivity software, including the standard Microsoft Office suite of programs and Adobe’s Creative Suite including Photoshop and Illustrator. There are a number of professional-level scientific applications, such as Mathematica and Igor, which will run in OS X. For the scientist, OS X has the advantage that one single interface gives access to both unix and commercial productivity tools at the same time, with no need to run a dual-boot system, to run emulators such as Cygwin or Wine, or to run alternative packages such as OpenOffice. These hybrid approaches are not without problems, inevitably so.

As noted above, Apple computers have not been prominent in the UK escience efforts. Apple are aware of grid developments3, and certainly aware of the scientific research market. For example, their Xserve cluster system is targeted at the bioinformatics community. Moreover, with the work at Virginia Tech4, the Xserve system gained a large amount of publicity with its potential to run as part of a supercomputer platform.

Experiences of using OS X within an escience environment

• Condor: The Condor team releases binaries for OS X, which work in a straightforward fashion. Incorporating the eMac computers into a larger Condor pool proved easy. The OS X installation comes with all standard Condor client tools (including Condor-G, which is important for the way that users access the eMinerals minigrid), but is restricted in the range of available operating environments.
• Globus: Installing Globus on any system is often a problem unless you have one of the systems for which there is a provided binary distribution. There is a good private distribution of GT 2.4 available for OS X5, which requires only modest changes to the instructions to make work. GT 4 appears to have good support for OS X.
• Storage Resource Broker: We have built both the client tools and the tools for constructing SRB vaults from provided sources, for both versions 2 and 3 of the SRB, and have set up an SRB3 vault. However, we have not tried to build an MCAT server under OS X.
• Access Grid: Until recently there was no support for running the Access Grid on OS X. However, since February 2005 there has been a release of the Access Grid 2 software for OS X, and our experience is that this has worked well “out of the box”. It has the requirement to work with Apple’s own iSight camera, which is only slightly more expensive than most consumer-level webcams.
• Web services: Apple’s development team were an early adopter of web services, and OS X is a straightforward platform for the development of web services applications. OS X includes essential tools such as Apache axis, tomcat, with support for SOAP and WSDL for example6.
• Computing environment: As mentioned above, to support production level computations on Apple hardware, there is a good range of compilers, libraries and other tools. The eMinerals project uses IBM’s commercial xlf Fortran compiler, although there is now a g95 compiler available (the g77 compiler has been available for some time). The gcc3 compiler, and other language tools, are available either with the standard installation or as freely-available downloads. Libraries such as MPICH and LAM/MPI, BLAS, LAPACL and ScaLAPAC, are available, to enable Apple hardware to be used for robust computational work. Part of the driver behind these developments is for applications in the bioinformatics area7, as noted above, although the same tools are in use within the eMinerals projects for large scale molecular dynamics and ab initio quantum mechanical simulations of materials.