#NodeSummit: The Internet of Things

Date: Sat Apr 26 2014
Speaker is Steve Pawlowski, of Intel.

The way Industry has used relational databases to store data has to change with the new kinds of devices. Leveraging the information that's available can give significant benefit to the whole industry. The issue is very broad ranging. A lot of unstructured data such as sensors that are out in the environment.

The unprocessed data is growing at a phenomenal rate.

Intel is working with CERN closely because of the tremendous amount of data ... their instruments give tremendous amounts of data per second that the machine is running. They're looking for specific signatures (Higgs Boson) and they'll throw away any other data.

They're also working with oil & gas industry that also generates a huge amount of data. This industry wants to keep the whole dataset and care that the data remains in integrity over many years, whereas CERN doesn't care if the data degrades.

Hence users have different models of what they want from the data.

Traditional Data relies on relational databases (SQL) and batch-style analytics, while "Big Data" relies on NoSQL or NewSQL databases, flexible schema's, distributed hierarchy, distributed analytics.

Algorithms will have to evolve. Old style algorithms will use 8x or more the resources required over new style algorithms. This makes the sort of difference called the energy efficiency of computation.

On one level this is an environmental question, because as the Internet grows and becomes even more ubiquitous, would we cook the planet with the energy and other resources required to run the machines? What's needed is the computing infrastructure required to run the Internet is to be as efficient as possible. It means algorithms and methodologies that scale out at the massive levels.

Embrace Heterogeneity - Different types of architectures, etc, on the needs of the customers rather than one size fits all

Integrate New Memory Technologies - e.g. new memory technologies that develop different characteristics and compute models.

Reduce Data Movement - Data deduplication algorithms are nascent. Network bandwidth will always be a precious commodity

Security is a huge concern especially with the way small compute devices are being baked into the woodwork. Example are security exploits through bluetooth to get into on-board computation on the car.

Prevent the kind of things they know (signature of the attacks they know about); mechanisms to detect aberrant behavior; recover quickly and gracefully

Exascale program .. energy efficiency is a big issue. Efficient data center usage, 2/3rds of power goes to servers, 1/4 goes to CRAC's and UPS. In 2006 data centers accounted for 1.5% of total U.S. energy consumption, the choice is whether the energy usage curve has to continue growing as the Internet grows.

An energy efficient architecture could get the Internet onto an energy usage curve that heads downward while still growing the number of transactions on the Internet.

An aspect of the growth of the Internet is all the mobile devices being built out, with the varying transaction load as apps on mobile devices query services on the cloud. Node.js obviously plays a role in implementing this world, and in particular the efficiency of the Node platform could be the basis for implementing power efficient server platforms for the PAAS cloud systems.