SAGrid Diary - a week at the SAAO

During week 4 of 2012, I have been working at the South African Astronomical Observatory on the integration of their compute cluster and data systems into the grid. During this week, I've had the opportunity to meet with some old friends from the astro community and especially have some long and productive chats to the SALT data team and the simulation team from the LADUMA collaboration who happened to have a collaboration meeting up at UCT's main campus this week. A case of killing many birds with one stone, certainly, and it's made for a very productive week here in Cape Town, with a daunting and stimulating to-do list.

User Induction

First on that list is to prepare a follow-up session with researchers to provide them with the initial tools to get started on the grid. To catch the most interested parties possible, we've prepared a poll to identify availability : http://www.doodle.com/uv49x6kn9qxkddhf If you're interested, just follow the link and tell us when you're available, it will go a long way to helping us to organise the event. What's covered in the user training addresses two main issues :

• Getting started on the grid : what is it, how does it work, how do I do basic things ?
• Making the grid useful : identifying your application, porting it to the grid, running applications, managing data, etc.

See the recent EPIKH application porting school for an idea of what is covered, but bear in mind that the event is organised with the principle of making the grid a useful and productive set of services for you, the scientist.

Site Administrator Training

A further task to do is to ensure a transfer of knowledge. To this end, we (the SAGrid operations team and the fine people we work with) will be preparing a site administration course for the SAAO IT team. This is a process which will ensure that the local experts who operate the equipment can become part of the team of experts which keep the grid running, and is more a process of imparting a consistent methodology and procedure rather than any particular technical knowledge.

The operation of Africa's largest (and probably only) distributed computing infrastructure is coordinated by the Africa and Arabia Regional Operations Centre, which itself forms part of the global federation of e-Infrastructure initiatives which we collectively refer to as "the grid" (or this fine instructive video). Ensuring this operational consistency, although apparently not very interesting to "the scientist on the street", is actually fundamental to the ability to build an integrated infrastructure for collaboration that research infrastructure such as SALT or MeerKAT and the scientists that use it depend on. With a federated computing infrastructure and consistent methodology, we can make geography somewhat irrelevant, allowing scientists to collaborate smoothly with each other no matter which institute they belong to.

All of this, of course assumes that the physical (fibre-optic) network is in place to connect things together, over and above the human networks which we discussed... which brings us to an interesting point.

Data management and movement

It came to my attention while working a the observatory that while SANReN connectivity has changed the game in terms of network capacity within South Africa, collaboration is still somewhat stifled by the constraints on international connectivity. After some discussion with the IT staff here, the SALT data manager and colleagues on SANReN, we have come up with a plan to try to address that in the near future. Watch this space - but if you have a large data set which you're having trouble sharing with international colleagues, please let me know - see below for contact details.

Applications, applications, applications !

The grid, as part of an integrated infrastructure, needs to be a platform for improving productivity and enabling collaboration. Two specific examples of this are solving the issues of :

• Make my simulation go faster !
• I want to share my data with my collaborators... and do strange, complicated things to it together !

These are things which a distributed, federated computing environment like the grid does exceptionally well, considering that it's made up of many individual resource centres (sites, which contribute shared computing power and data storage in the form of HPC clusters and mass storage devices. Whether you need massively-parallel or embarrassingly-parallel capacity, simple access to data, or complex workflows of data processing, the grid can do it. To demonstrate how, I'll be working on two particular cases :

1. SALT data management
2. porting the GADGET application to the grid

These will be used to demonstrate how the services provided by the grid can enable collaboration and ease the day-to-day business of science. Co-incidentally, they will also be contributed back to CHAIN project (Coordination and Harmonisation of Advanced e-Infrastructures),which will be having it's annual review in Athens next week. This is of course nowhere near the end of the story - I've heard of many applications which are being used by the researchers here, all of which we want to deploy onto the grid. This will be done during the induction session I mentioned above, and eventually once they are working on the grid, we will register them in the European Grid Initative (EGI) Appliation Database,so that others can benefit.

See you in Feb/March !

So, it's goodbye for now to Cape Town, at least in the physical world. I'm on my way to Athens for the CHAIN Review, where snow is expected apparently, so I'm enjoying the last few rays of sunshine here in the Mother City. Please don't forget to tell us when you will be available for the training session in Feburary or March. If you want to keep in touch, feel free to follow us on Facebook and Twitter:

https://www.facebook.com/SAGrid

http://twitter.com/thesagrid

or drop me a line directly : This e-mail address is being protected from spambots. You need JavaScript enabled to view it

Bruce

Share