|
The growing energy consumption of data centers has emerged as
a key challenge to the IT industry. In the joint Data
center energy management project with the IBM Watson Research
Center, we are contributing to the evolution of measurement and
management technology (MMT) for acquiring and processing environmental
data produced by a data center. The original MMT tool employs a
mobile cart with wired sensors for acquiring temperature and humidity
data. Air flow readings are taken with a manual sensor. To allow
real-time data capturing, we deploy wireless sensors that use IEEE
802.15.4 technology with ZigBee/mesh networking or proprietary
protocol stacks, and IBM's MQTT messaging for connecting the sensors
to monitoring and processing applications. Measurement campaigns
are carried out in IBM data centers and at customer sites. Temperature
maps derived from these measurements are used for identifying hot-spots
and cooling inefficiencies. The insights show us what measures
need to be taken to improve air flow and cooling efficiency to
achieve significant energy savings and better operational reliability
of a data center.
Middleware for sensor systems:
Publish/subscribe messaging is used widely for efficiently
connecting a large number of data sources with multiple applications.
Topic lists are maintained by a broker that accepts data published
by the sources and passes it to the subscribers. This is especially
efficient in systems where the data sources are active only infrequently
and the applications subscribe only to a subset of the available
data. We have developed a version of IBM's MQTT pub/sub messaging
protocol to allow operation over wireless links, e.g. ZigBee, Bluetooth,
or GPRS. The specification of MQTT-S has been published at http://mqtt.org,
and a lean implementation for low-power sensor nodes is also available.
In addition, we are working with IBM business units on customer
opportunities where MQTT-S is an important enabler for connecting
sensor devices with IBM application environments.
To evaluate the functionality and performance of wireless technologies
that are relevant for sensor networking, we have built a sensor
system lab that supports IEEE 802.15.4 devices with ZigBee
or TinyOS protocol stacks, and multi-hop / mesh networking. ZigBee
sensor networks have been deployed in our office building and in
the local data center. Each sensor node is equipped with sensors
that measure temperature, humidity, and light intensity. To assess
the behavior of MQTT-S, we have written several applications that
use MQTT-S for capturing sensor data and sending control information
to sensor nodes. A "live example" demonstrating the real-time
acquisition of light intensity data is shown here.
We are working on location sensing
and routing protocols for tracking objects and persons in
indoor environments where GPS is not feasible. Wireless
sensor nodes are installed at static reference positions. Mobile
nodes estimate their position using radio signal-strength
measurements, data derived from inertial navigation sensors,
and Kalman filter processing. Positions are broadcast
to the static nodes and used by a geographic routing protocol
in the wireless mesh network. We have prototyped the concept
in a real-time testbed. Measurement results show a position
accuracy of 1-2 m. Potential applications include scenarios
in the retail industry ("Store of the future"), asset
management, and transportation.
EU projects. We participated in
the two-year project e-SENSE. Together with leading partners from
industry and academia, we developed a new architecture for wireless
sensor networks and their integration into "beyond 3G" mobile
communication systems. We contributed to the e-SENSE system architecture,
the design of a lightweight protocol stack ("e-stack"), and novel
distributed middleware concepts. A special focus was placed on
developing advanced "cooperative relaying" algorithms and lightweight
distributed pub/sub protocols to obtain a scalable, power-efficient
multi-hop WSN.
In addition, we participated in the initial phase
of follow-on project SENSEI, which focused on the integration of
sensors and actuators into the "future Internet" to enable context-aware
services. Leading the key technical work package in SENSEI, we
contributed to the definition of the initial reference model, system
requirements, and functional architecture of the SENSEI system.
Past activities include projects
on ultra-wideband (UWB) radio technology and 802.11x radio LANs.
|
