Terra Soft Carries the Torch for Cell and Power Platforms
Since launching Yellow Dog Linux for the PLAYSTATION3 (PS3) last year, Terra Soft has been busy expanding its presence in the Cell processor ecosystem. To get an update about what the company has been doing, we contacted Terra Soft’s CEO Kai Staats via email in Nakuru, Kenya. Staats is there working with the Pistis Orphanage & Academy to complete some important projects for the children’s home and school.
HPCwire: The last time we talked in October 2006, Terra Soft had just announced the first Cell-based clusters “E.coli” and “Amoeba.” Can you catch us up on what’s happened at the company since then? What’s the status of the “E.coli” and “Amoeba” clusters?
Staats: The two clusters originally proposed by Terra Soft, a total of 480 nodes, were delayed due to a change of hardware, Sony preferring that we use production PS3 systems in place of the slated 2U rackmount beta units as they came back from game developers, world-wide.
This spring, the cluster was initiated by a starting block 16, in operation now, with an anticipated total of 128 in motion. This cluster is comprised of an Apple G5 head node with Gigabit Ethernet interconnect to 16 PS3s. All units are running YDL [Yellow Dog Linux] v5.x with TORQUE, Moab, and MPI in place for both parallel and distributed jobs.
This cluster, now and as it grows, is available for free to all Consortium Technical members to use as a test-bed for their own cluster code.
In other areas, Terra Soft has landed two substantial board support package contracts for new Cell based servers, details to be announced later this fall via official PR. In this respect, Terra Soft has gained a pole position as a leader in the provision of a Cell operating system.
HPCwire: What’s going on with the HPC Consortium you launched back in January?
Staats: It is doing well. We spent nearly six months rebuilding the physical and logical infrastructure in preparation for growth, now unifying G-Forge (the open-source foundation for Source Forge) with LDAP and a member wiki for a single, per user account across the build-box, QS20 blades, and PS3 cluster. New systems may be rapidly deployed, seamlessly dropping into the existing system.
The goal of the Consortium has grown too, with a vision not just for the successful launch of the Cell processor via the original Hack-a-thon, but to carry Consortium members into years of forward thinking, open and private research on leading, high performance computing technologies.
Interest at Sony, IBM, and Mercury has helped shape the Consortium in its early stages, pressing for not-for-profit and eventually, 501c3 status in order to receive “donated” hardware into the 3,000 square-foot supercomputing room which has a capacity for 2,000 1U rackmount systems.
The board of directors is taking shape, attorneys are engaged, and a meeting with individuals from IBM and Sony mid September will help shape the founding charter.
Personally, this is very exciting for me. To know that we are building something beyond a sales engine or a marketing campaign, but truly a foundation for real-world research — this is why I got into this business nearly a decade ago and it is now taking form.
HPCwire: How’s the reception been for Yellow Dog Linux for the PLAYSTATION3? What is the typical YDL-PS3 user doing with the system?
Staats: Excellent reception. Sales are steady. The user base growing. And of greatest interest to me, the creativity for how PS3s are being used is continuing. Axion Racing is moving to install a PS3, running YDL, in their souped up 4×4 Jeep, for the 3rd DARPA Grand Challenge (where completely autonomous vehicles drive by computer vision and on-board processing). The Air Force has purchased dozens of PS3s from us, using them independently and building test clusters.
The Navy, DOE lab, and universities (in the U.S. and abroad) are also finding value in a lightweight, inexpensive Cell cluster.
But the vast majority of our users are running YDL on their PS3 to do exactly what was intended, be a home computer.
HPCwire: Are customers using Yellow Dog Linux on the PLAYSTATION3 as a stepping stone to applications that will end up on Mercury or IBM Cell boards?
Staats: Yes, indeed. At Los Alamos National Lab, there are both Sony PS3s and IBM/Mercury Cell blades, the natural evolution from the former to the latter as the complexity or size of the problem grows.
HPCwire: What do you see as the company’s primary focus over the next few years?
Staats: As Power.org, IBM, Freescale, and PA Semi work together to build a new level of public awareness and momentum around the Power architecture, Terra Soft’s focus will be on the development and maintenance of Board Support Packages (BSP).
Our strategy is threefold: we are paid by the chip vendors to prepare a BSP for their reference board, by the OEM who uses the vendor’s chip to build a unique system, and then by the OEM’s customer as an End User License. This proven model is increasingly working well for us, enabling growth and solidity beyond what we have previously experienced.
In addition, we will continue to focus on the development of market specific apps such as Y-HPC, Y-Bio, with a potential for movement into the film industry with Y-Film.
HPCwire: On a more personal note, can you tell us a little bit about the orphanage projects you’re involved with in Kenya and how you got interested in them?
Staats: I had been wanting to return to Africa since the late spring of 2001 when I spent two weeks in Oshigambo, northern Namibia, where I rebuilt a computer lab for a school. But the gradual growth of Terra Soft has not left room for such ventures often. However, with the recent, rapid growth of cell phone infrastructure and internet in Africa, I recognized a new opportunity to volunteer at the orphanage by day and keep up with Terra Soft by night and early morning. This experiment, while not leaving much time for sleep, has worked well. Through SPAN, Student Projects Africa Network (www.studentprojectafricanetwork.org), founded by Rebecca Mitchell, I learned of an opportunity to work with an orphanage and academy called Pistis.
In quick summary, I brought over $3,000 USD in donations, using the funds to complete electrical wiring of the middle floor of the school, to run two water lines to supply the new bath house and bring drinking water to the kitchen; and to design and build a new storage system with intent to keep the mice at bay.
There are several other, small projects that we have undertaken. The children who live at the school are incredibly fun to work with and helpful with their time, energy, and muscles.
Next week, my final week, I will concentrate on helping the Pistis management build and learn to use a cash flow management spreadsheet and bookkeeping software (on a new computer I purchased for them) in order that they may more accurately project their financial need and manage what funds they receive.
It is hard to describe in these few words the experience I have here in Nakuru, Kenya. I ask that you and your readers visit my blog, where I share a great deal more at blogs.ydl.net/kai/.