distributed systems, mobile computing and security
TRANSCRIPT
Antti Ylä-Jääski Feb 12th 2016
Distributed Systems, Mobile Computing and Security
Secure Systems in a NutshellWe investigate how to build systems that are simultaneously easy-to-use and inexpensive to deploy while still guaranteeing sufficient protection.
Examples of research questions:
• Can contextual data on user devices help improve security usability?
• How can cloud services ensure user privacy?
• How can we design secure software-defined networking?
Contact: N. Asokan and Tuomas Aura
Usability Deployability/Cost
Security
Research Programs and funding:Contextual Security (AoF), Cloud Security Services (AoF), CyberTrust (Tekes), Mobile System Security (Intel and Huawei)
More info:Wiki: https://wiki.aalto.fi/display/sesy/Secure%20Systems Blog: http://blog.se-sy.org/
Mobile Computing and Distributed Systems in a Nutshell
We evaluate and optimize the performance of mobile and distributed systems.We build new applications and services for mobile devices and big data scenarios.
Sample research questions:•How to save energy on
handsets and data centers with SW optimisations?
•How to optimize user experience for mobile cloud services?
•How to apply mobile crowdsensing to solve real life problems (navigation)?
•How to efficiently collect and utilize data from a massive number of devices connected to the Internet?
•How to build large scale distributed systems for big data in IoT and health?
Our current focus areas:• Mobile cloud gaming• Multimedia streaming• Indoor navigation• Crowdsensing• Internet of Things• Scientific, cloud, and
mobile edge computing
Contact: Antti Ylä-Jääski
Cloud (e.g. Amazon EC2)
Mobile Edge Computing
Mobile Cloud GamingIn Mobile Cloud Gaming the game is rendered on the cloud data center and streamed to a mobile phone• Latency is the main QoE issue in Cloud GamingVirtual machines introduce overhead into the system• Linux containers are more light-weight with
native performanceResearch questions:• How to design a distributed mobile cloud
gaming system (server placement strategy, virtualization)?
• How to model and predict end-to-end latency with mobile access network?
• What is the effect of latency on gaming experience?
4.12.2015
QoE Optimization of Mobile Video Streaming
4.12.2015
• QoE modeling and optimization• Analyze and (re)design on-demand and live mobile
video streaming systems• Use adaptive protocols and scalable video coding
• Power modeling and optimization of video delivery• Optimal use of radio resources through smart
download scheduling• No penalty in terms of video quality
HTTP server
Internet
Mobile crowd sourcing for indoor navigation
4.12.2015
• iMoon is an indoor navigation system using sensor-enriched 3D models that are created & maintained using crowd sourced photos and sensor data
• iMoon provides image-based localization and visual navigation
• iMoon user can be located with better than 2 m position accuracy and 6 degrees facing direction accuracy
Internet of Things
• More than 30 billions of smart objects will be part of the Internet by 2020– What are the consequences?
• Efficient data collection and management are key issues– User-friendly and scalable methods
to configure smart objects– Energy-efficient data collection– Modeling of large-scale networks
of smart objects
4.12.2015
Mobile Edge Computing
• Mobile Edge Computing (MEC) is a new industry initiative targeted to implement novel services next to the end user in the mobile network
• In practice, an ordinary server component is integrated into the base station providing cloud based computational and storage capacity
• Nokia’s solution is called RACS, which has been installed at our test lab• We develop and evaluate performance of potential applications using
this platform like IoT data filtering, content acceleration and video orchestration
4.12.2015
Green Big Data
Electricity has become one of the main costs of computingIn cooperation with CERN we analyze and improve the energy consumption of scientific computing and massive data analysis• Analyze profiling and log data • Model and predict power
consumption • Develop energy-efficient algorithms
and solutions for distributed computing
4.12.2015
Big Data Platforms for IoT and Health
4.12.2015
• Massive data volumes coming from e.g., IoT, Genomics, Health, and Social Networks require Big Data platforms such as Spark and Hadoop
• Our Hadoop-BAM is becoming the de facto standard to process NGS in parallel with Spark & Hadoop. Library users: Halvade (Gent), SparkSeq (ETH), SeqPig (Aalto), SEAL (CNRS4), Adam (Berkeley) and upcoming parallized version of GATK (Broad Institute)
• Health big data piloting with HUS
IoT backendarchitecture
Speedupon 64computerswithHadoop-BAM
Automated Parallel Testing and Verification
• Traditional ways of testing and simulation do not scale to validation of large distributed systems
• Model checking and automated testing are used to find bugs in concurrent systems
• Our speciality: Automated symbolic and parallelized methods for distributed systems
• Application areas: Safety critical systems (nuclear automation with VTT), multithreaded programs, hardware verification
• Organizing hardware model checking competition 2011-2015 with Prof. Armin Biere
• Visiting Professor in 2016: Prof. Roland Meyer from Univ. Kaiserslautern – “Formal-Methods-based Analysis of Geo-Replicated Big Data Applications”
4.12.2015
4.12.2015
Information-Centric Networking (ICN)
ICNNAP
IPNAP
ICNBorder
GW
IP-onlySender
UE
IP (BGP)
IP
ICNF
IP
IP
FN
TM
L2
ICNPR
ICNRT
ICNTP
ICNNAP
ICNF
IP
IP-onlyReceiver
UE
IP-onlySender & Receiver
UE
L2
ICNSR
S1
S1
IP
TM : topology managerRVZ: rendezvous pointFN : forwarding node
S2
SDNSwitch
FNSDNSwitch
RVZ
SDN Controller
• In ICN we address information - not hosts• The main applications of the Internet already
are information-centric by nature• By making the underlying network information-
centric, we can better support modern applications (e.g. IPTV) by the extensive use of multicast and caching, making CDNs obsolete
• We are coordinating our third consecutive ICN EU-project, the Horizon 2020 POINT, which is bringing ICN from laboratories to the real world
• POINT aims to show that current IP applications can run better over an information-centric core network