NEC R&D meeting 2009

Stream Computing: Real-time Processing of Massive Data

What does it solve?

Visualization of traffic congestion
● The Currently Future（with new NEC technology）
amount of data dealt with in the ubiquitous
society will expand explosively by a factor of 200
by 2025
●A huge increase in IT devices and amount of data
processing will produce a rapid 500% increase in
power consumption by 2025 Only main roads; updated every Smaller roads included; updated
several minutes in less than a minute

More efficient data center administration

Thin client terminals
● Fine visualization and control by the real-time
collection and analysis of huge amounts of time-series
data continuously acquired from various sensors VM

Use example Server 1 Server 1

 Very fine traffic congestion data supplied by the VM

real-time processing of floating car sensors on Server 2 relocation Server 2

individual vehicles
 Eco-friendly data center administration by control Server 3 Server 3
*VM: Virtual Machine Move the VM and power
based on understanding detailed server load (thin client environment) down the server.
NEC R&D meeting 2009

Real-time Processing of Massive Data

Features of the technology

● Assembly-line analysis in the process of Conventional technology (database)

collecting data
Massive Analysis Application
 Real-time analysis without data data from
stream stagnation by multi-stage sensors, etc.
partial analysis of data
Database
● Efficient partial analysis of massive time Data overflow Decreased processing
series data speed due to access
conflict
 Faster processing through partial This technology
Partial analysis 1 Assembly-line
analysis execution of an incremental processing
algorithm that re-uses prior results Partial analysis 2
Partial analysis 3

Massive
Application
data from
sensors, etc.

Efficient processing of
massive time-series data
Raw data is discarded
NEC R&D meeting 2009

Real-time Processing of Massive Data

Future development

Market needs/technical prospects R&D roadmap

● Utilizing explosively increasing data Social infrastructure

Health management,
Discover new value by advanced analysis of Traffic analysis, Crime
Expanding market
massive data Prevention, etc.
→ Fusion of real-world data and cyber data New business Mobile advertising and
marketing that makes use of cell phones
ＩＴ systems Data center administration (energy conservation）
● Interworking with ubiquitous sensors
Collect and make use of the huge amount of
data from sensors embedded in various More added-value Distributed sensor
locations and devices or terminals that anyone data such as network (location and
marketing data behavior/movement)
can carry (PC or cell phone)
Sensor fusion More sensor data
→ Sensor fusion and sensor networking
(CPU, memory, temperature, GPS, IC cards,
etc.）
● New computer architecture Stream processing for
Break through limits on processing speed and distributed system
scale with a new architecture for handling Hierarchical stream processing
massive data
Stream processing for multi-core
→ Stream computing CPUs
2009 2010 2012 2014