Earlier this year, we explored cloud computing
and application-specific systems design, as Cadence Fellow Chris Rowen urged
us to "follow
the data" to understand evolving electronics-design dynamics.
Now comes Ron Wilson, Altera's editor-in-chief, who writes
about heterogenous computing design challenges in the cloud from the 2014
International Symposium on FPGAs. He dives deep into hardware acceleration
strategies, system partitioning, and how programming in this domain will evolve
in the coming years.
Arguably one of Wilson's most illuminating threads focuses on
design for reliability in the cloud. It's vital because cloud companies won't
repair failed servers in the racks. They do this because the systems are
fault-tolerant (other servers pick up the failed server's duties) and because
the lifetime of the data is just three years.
As we move deeper into the era of application-specific
system design, enabling these systems becomes so much more complex. It's about
power, performance, area, of course, but it's also about security and
reliability. They're all must-haves; they're not optional. And, therefore, the pressure on systems designers from implementation up through verification continues to mount.
Click here to read Wilson's
2014: Follow the Data to Optimize System Design--Chris Rowen
Perspectives on System-Design Evolution
-- We Need to Move
"Past EDA": Tensilica Founder Rowen
-- CDNLive Silicon