“The latest innovation with far-reaching implications for system performance is something called “software in silicon.” The concept is exactly as it sounds—putting software right into the chip for better, even faster performance.”
John Soat on Forbes.com.
As anyone who has worked with an OSS that has a large database can attest, OSS performance can be frustrating as queries on large data sets can be glacial. As such, the article above is quite interesting, particularly the quote, “it’s important to understand that the innovations in the new SPARC processor are a continuation of Oracle’s enterprise design strategy—known as Engineered Systems—which fits together servers, software, and storage into a single, highly optimized system that runs Oracle’s enterprise applications at peak performance.”
I’m not enough of an expert in this area so I’d love to hear your thoughts on whether claims such as the following might be beneficial to the operation of OSS, or at least the databases upon which they rely:
- The M7’s most significant innovations exemplify software in silicon, a technique that hardwires processes that normally would occur in a software application directly onto the processor platform. Because those functions are both part of the processor and yet separate from the processor’s core, it speeds up performance of an application without affecting overall functionality of the processor.
- One of the most compelling new features in the SPARC M7 is a query acceleration capability that ups database performance. Known as in-memory query acceleration engines, these units take over certain data-search functions from a database query that then get processed at a very high rate of speed. This offloading of hardcore number crunching directly onto hardware in the processor is what boosts database query processing so significantly.
I’d also love to hear your thoughts on whether the OSS software itself might end up in silicon?
- Could we see a time when OSS developers are working alongside processor and storage engineers to produce Engineered OSS Systems that are optimised on a per vendor or even per product basis?
- Are there any specific aspects of OSS that make optimisation for embedding into silicon attractive?
- Or does the evolving / changing nature of OSS software make it unfeasible?