Google and IBM help students learn parallel programming
Google and IBM have announced a new initiative to help students and researchers learn how to create the massively parallel programs that will be needed to address large-scale problems of the future.
Context & Ripple Effects
This lands one day after the New York Times reported the two companies' joint 'cloud computing' research effort (Google and I.B.M.'s cloud computing research), and it exposes the part of that push aimed at people rather than machines: a shared curriculum for massively parallel programming. The pairing matters because both firms run some of the largest commodity-cluster installations in the industry, and neither can hire enough programmers who think in thousands of cores rather than one.
First-order effects
- Students and researchers at participating universities get hands-on access to large-scale cluster infrastructure they could never fund themselves, making parallel-programming coursework practical rather than theoretical.
- Google and IBM effectively stand up a recruiting pipeline: graduates trained on their style of distributed systems arrive pre-qualified for each company's datacenter engineering needs.
Second-order effects
- Rival platform operators face pressure to match the program with their own university partnerships, since whoever shapes the curriculum shapes where the next generation of cluster engineers wants to work.
- Hardware and tooling vendors gain a growing academic market as multicore processors force parallelism into the standard computer science syllabus rather than leaving it a specialist niche.
Third-order effects
- If the pattern holds, university computing curricula become an extension of hyperscale vendors' infrastructure strategy — vendor-funded labs and course material deciding which skills the industry's future hires have.
- Parallel and distributed programming shifts from elective to baseline expectation across the field, with the biggest datacenter operators setting the de facto standards through what they teach.
The trend: Hyperscale operators are moving from building large clusters to cultivating the parallel-programming workforce those clusters require, using university programs as the supply chain.