top of page
  • Writer's pictureDAVID MITLYNG

Weekly Takeaways-August 2, 2023

Updated: Apr 1

AI's Dirty Little Secret Artificial Intelligence (AI) is all the buzz nowadays. Many experts believe it can be a boon for the world – if it doesn’t destroy it first. It has developed to the point where it can write articles and software, compose music, create award-winning artwork, replace actors, or be a virtual companion. This rapid advancement has led to concerns that AI will wipe out jobs or even take over the world. Or maybe not. But there is one concern that is indisputable: it is adding to our carbon emission woes. Behind the scenes are "vast data centers capable of running A.I. systems," which in turn requires more power and water for cooling “as Google and every other tech company in the AI arms race speed to build new data centers.” But there is an option to increase the efficiency of data centers: better timing. Meta and NVIDIA found that a synchronization improvement of 80x made the distributed database run 3x faster - "an incredible performance boost on the same server hardware, just from keeping more accurate and more reliable time." This resulted in reduced power consumption and cooling. This project convinced Meta, NVIDIA, and other like-minded colleagues to set up the Open Compute Project Time Appliances Project (TAP). Their motto: "Time is a key element to get the highest efficiency in a distributed system." Like it or not, AI is coming. Better timing has to come along too. Last Week's Theme: The World Changes Forever

Industry News

Conferences

  • Euroconsult, September 11 – 15, Paris, France

  • APSCC, October 10 – 12, KL, Malaysia

  • ITSF, Oct 30 – Nov 2, Antwerp, Belgium

  • UK National Quantum Technologies Showcase 2023, Nov 2, London, UK

  • SLUSH, Nov 30 – Dec 1, Helsinki, Finland

The More You Know...

So how does better timing improve the efficiency of data centers? The technical answer: better timing precision reduces surge events in databases, eliminates centralized nodes and reduces the effort to work with the database. The simpler answer: it helps reduce the read and write times for many distributed users. Better synchronization allows these users to efficiently access the information across one large, or many distributed, servers. And as AI grows along with our insatiable need for data, there is a trend from huge hyperscale data centers to large numbers of point-of-presence micro data centers (POPs). And these distributed data centers need better synchronization than GPS can deliver.


bottom of page