CSE 591 Green Computing Course

Slides:



Advertisements
Similar presentations
Tintu David Joy. Agenda Motivation Better Verification Through Symmetry-basic idea Structural Symmetry and Multiprocessor Systems Mur ϕ verification system.
Advertisements

Timed Automata.
Verification of Hybrid Systems An Assessment of Current Techniques Holly Bowen.
ECE 720T5 Fall 2012 Cyber-Physical Systems Rodolfo Pellizzoni.
Background information Formal verification methods based on theorem proving techniques and model­checking –to prove the absence of errors (in the formal.
MotoHawk Training Model-Based Design of Embedded Systems.
1 Introduction to Computability Theory Lecture12: Decidable Languages Prof. Amos Israeli.
EECE Hybrid and Embedded Systems: Computation T. John Koo, Ph.D. Institute for Software Integrated Systems Department of Electrical Engineering and.
Firewall Policy Queries Author: Alex X. Liu, Mohamed G. Gouda Publisher: IEEE Transaction on Parallel and Distributed Systems 2009 Presenter: Chen-Yu Chang.
Department of Electrical and Computer Engineering Texas A&M University College Station, TX Abstract 4-Level Elevator Controller Lessons Learned.
EECE Hybrid and Embedded Systems: Computation
November 21, 2005 Center for Hybrid and Embedded Software Systems Engine Hybrid Model A mean value model of the engine.
CSC 402, Fall Requirements Analysis for Special Properties Systems Engineering (def?) –why? increasing complexity –ICBM’s (then TMI, Therac, Challenger...)
November 18, 2004 Embedded System Design Flow Arkadeb Ghosal Alessandro Pinto Daniele Gasperini Alberto Sangiovanni-Vincentelli
Basic Concepts The Unified Modeling Language (UML) SYSC System Analysis and Design.
BAND-AiDe: A Tool for Cyber-Physical Oriented Analysis and Design of Body Area Networks and Devices Authors: Ayan Banerjee, Sailesh Kandula, Tridib Mukherjee.
Advances in Language Design
02/06/05 “Investigating a Finite–State Machine Notation for Discrete–Event Systems” Nikolay Stoimenov.
ECE 720T5 Winter 2014 Cyber-Physical Systems Rodolfo Pellizzoni.
IAY 0600 Digitaalsüsteemide disain Event-Driven Simulation Alexander Sudnitson Tallinn University of Technology.
Sensor-Based Fast Thermal Evaluation Model For Energy Efficient High-Performance Datacenters Q. Tang, T. Mukherjee, Sandeep K. S. Gupta Department of Computer.
Low-Power Wireless Sensor Networks
The Architecture of Secure Systems Jim Alves-Foss Laboratory for Applied Logic Department of Computer Science University of Idaho By, Nagaashwini Katta.
Mathematical Modeling and Formal Specification Languages CIS 376 Bruce R. Maxim UM-Dearborn.
Department of Mechanical Engineering The University of Strathclyde, Glasgow Hybrid Systems: Modelling, Analysis and Control Yan Pang Department of Mechanical.
1 Sequential Machine Theory Prof. K. J. Hintz Department of Electrical and Computer Engineering Lecture 1 Adaptation to this.
Thermal-aware Issues in Computers IMPACT Lab. Part A Overview of Thermal-related Technologies.
1 Context-dependent Product Line Practice for Constructing Reliable Embedded Systems Naoyasu UbayashiKyushu University, Japan Shin NakajimaNational Institute.
Fall 2004EE 3563 Digital Systems Design EE 3563 VHSIC Hardware Description Language  Required Reading: –These Slides –VHDL Tutorial  Very High Speed.
Electrical and Computer Engineering University of Cyprus LAB 1: VHDL.
1 Chapter 5: Harmonic Analysis in Frequency and Time Domains Contributors: A. Medina, N. R. Watson, P. Ribeiro, and C. Hatziadoniu Organized by Task Force.
HEAT TRANSFER FINITE ELEMENT FORMULATION
CSCI1600: Embedded and Real Time Software Lecture 28: Verification I Steven Reiss, Fall 2015.
Software Quality and Safety Pascal Mbayiha.  software engineering  large, complex systems  functionality, changing requirements  development difficult.
Formal Verification. Background Information Formal verification methods based on theorem proving techniques and model­checking –To prove the absence of.
IAY 0600 Digital Systems Design Event-Driven Simulation VHDL Discussion Alexander Sudnitson Tallinn University of Technology.
Introduction to Hardware Verification ECE 598 SV Prof. Shobha Vasudevan.
Superstabilizing Protocols for Dynamic Distributed Systems Authors: Shlomi Dolev, Ted Herman Presented by: Vikas Motwani CSE 291: Wireless Sensor Networks.
CS5270 Lecture 41 Timed Automata I CS 5270 Lecture 4.
Green and Sustainable Cyber-Physical Security Solutions for Body Area Networks 1. Introduction 2. Security in Body Area Networks: Need and Approach Krishna.
Lecture and laboratory No. 10 Modeling product as system Óbuda University John von Neumann Faculty of Informatics Institute of Applied Mathematics Master.
Sub-fields of computer science. Sub-fields of computer science.
INTRODUCTION TO WIRELESS SENSOR NETWORKS
OPERATING SYSTEMS CS 3502 Fall 2017
IAY 0600 Digitaalsüsteemide disain
Finite State Machines Dr K R Bond 2009
How SCADA Systems Work?.
Graph Coverage for Specifications CS 4501 / 6501 Software Testing
Instructor: Rajeev Alur
Green Software Engineering Prof
Intel Atom Architecture – Next Generation Computing
IAY 0600 Digital Systems Design
Frequently asked questions about software engineering
Real-time Software Design
Capability reporting Francesco Schillirò INAF- OACT
Computer Simulation of Networks
The Extensible Tool-chain for Evaluation of Architectural Models
Logical architecture refinement
CSCI1600: Embedded and Real Time Software
Internet of Things A Process Calculus Approach
Introduction To software engineering
What is Concurrent Programming?
John D. McGregor Session 5 Error Modeling
Simulating convective impingement heating in HASPIF
Market-based Dynamic Task Allocation in Mobile Surveillance Systems
A Trusted Safety Verifier for Process Controller Code
Presented By: Darlene Banta
Lecture One: Automata Theory Amjad Ali
Chapter 3 Modeling in the Time Domain
Rich Model Toolkit – An Infrastructure for Reliable Computer Systems
Presentation transcript:

CSE 591 Green Computing Course Modeling Based Engineering for Safe and Sustainable Body Area Network and Data Centers CSE 591 Green Computing Course

Models Model is an abstract representation of a selected part of the system Models of phenomenon – fluid flow models Models of data – regression models Model can represent an entire theory with theorems and laws. Newton’s model for gravitation Bohr’s atomic model We concentrate on the first type of models.

Types of Models Architectural model - the primary aim is to illustrate a specific set of tradeoffs inherent in the structure and design of a system or ecosystem.  Behavioral model – models the interaction of the different components of a system Control flow – Algorithmic view of the operation of a system Data flow – Input / output view of different components along with the data flow paths. State machines – Event based execution of a system

Background on Model based Verification/Analysis Model based analysis normally used to verify critical systems such as avionics. no need for actual scenario generation putting lives/property at risk. Formal models for abstraction of the system behavior. Expected system properties depend on the requirements. Formal models analyzed through model checking to verify the system properties. System Behavior System Requirements Formal Models Expected Properties Model Checking Property Verification Requirement Verification

Body Area Networks (BANs) SpO2 EKG EEG BP Base Station Motion Sensor Wearable Sensor Nodes Base Station Heating effects (Unintended interactions) Thermal Map of Human Body Issues: Thermal safety – keeping human body temperature within safe limits Sustainability – un-interrupted operation with energy scavenging Communication Range Aggregate Effects Communication Range (Intended Interactions) Body Sensor Network (BSN)

Model Based Communication in BAN Use generative models for data A light version in the sensor A full version in the base station Low communication overhead Low storage requirements Ensure required accuracy for clinically relevant data

BAND-Aide – BAN modeling

Data Centers Data Center Computing Units – CRAC Racks Raised Floor Hot Aisle Cold Aisle Computing Units – Server racks arranged in rows CRAC unit supplies cold air from underneath the floor Cold Ailse near server inlets Hot aisle at the outlets Issues: Thermal Safety – schedule tasks into servers so that their inlet temperatures do not exceed manufacture specified redline temperature Sustainability – Energy efficiency, Heat activated cooling Cool Air coming from CRAC (Intended Interaction) Hot Air coming out of chassis (Unintended Interaction) Heat Recirculation (Aggregate Effect) Interactions – Intended – CRAC cold air cooling off racks Unintended – re-circulated heat causing hot spots Ayan Banerjee, Tridib Mukherjee, Georgios Varsamopoulos, and Sandeep K. S. Gupta Integrating Cooling Awareness with Thermal Aware Workload Placement for HPC Data Centers , Elsevier Comnets Special Issue in Sustainable Computing (SUSCOM) 2011 (Accepted for publication).

Data Center Modeling

Cyber-Physical Systems A cyber-physical system is a system which has a computing units embedded in a physical environment The computing unit is constantly interacting with its environment in two ways – Intentionally – for execution of system operations Unintentionally – through side effects of its operation Interactions may have aggregate effects during networked operation of the CPS Cyber-Physical System (CPS) Computing node Space in physical environment interacted by single node Aggregate impact in space because of interactions from multiple nodes Cyber-physical interactions

System Requirements Safety – Safety of any system is defined as ensuring the impact of the interactions is within desirable limits. E.g. - keeping the temperature of the servers within redline Sustainability - Sustainability is defined as the ability of the CPS to operate by scavenging energy from the environment. In a BSN the sensor nodes operate by scavenging energy from human body

CPS Modeling Perspective Network of Local CPSs Effect of interactions are limited spatially Intended Interactions – ROIn Unintended Interaction – ROIm Network of computing units imply a network of Local CPSs Each Local CPS can affect the ROIm or ROIn of other Local CPSs leads to complex aggregate effects of interactions A. Banerjee, S. Kandula, T. Mukherjee, and S.K.S. Gupta BAND-AiDe: A Tool for Cyber-Physical Oriented Analysis and Design of Body Area Networks and Devices , ACM Transactions in Embedded Computing Systems, Special Issue on Wireless Health 2010, Accepted for publication

Example Scenario BSN Thermal Safety Computing Unit – Atom based Sensor node running health monitoring workload Physical Unit – Human body Interaction – Heat dissipation due to computation causes temperature rise at different parts of the human body. The thermal effect of a sensor is governed by Penne’s bioheat equation Sensors close to each other have aggregate effect on the skin temperature – the heat accumulated gets summed up Heat accumulated Heat transfer by conduction Heat by radiation Heat transfer by convection Heat by power dissipation Heat by metabolism

Mapping to CPS modeling perspective Human Body Thermal Effects Sensors GCPS LCPS1 LCPS2 Aggregate effects Computing Unit Physical Unit Governed by Penne’s Equation ROIm

BAND-AiDeModel Analysis

AADL Implementation Industry standard Advanced Architecture Description Language Pros - Used in the embedded industry and can model complex systems such as aircrafts Specific constructs for modeling the embedded computing devices Hierarchical model specification – matches with the CPS view Cons – No support for modeling the physical system Cannot represent dynamic variations of physical properties in terms of differential equations in AADL

BAN Model in AADL system BAN . . . end BAN; process implementation application subcomponents algorithm: thread algorithm.imp1; end application; system implementation BAN.ins1 subcomponents Sensor1: system CompUnit.Sensor1; EnergySource: system EnergySource.impl; Body system PhysicalUnit.skin; . . . connections connection between subcomponents end BAN.ins1; thread implementation algorithm.imp1 modes . . . properties end algorithm.imp1; system implementation EnergySource.impl . . . end EnergySource.impl; system CompUnit features port specification for connections properties Computing Properties Physical Properties end CompUnit; system PhysicalUnit features port specification for information transfer properties Physical properties end PhysicalUnit; system implementation CompUnit.Sensori subcomponents P1: process application; C1: system subcomponents; connections inter-connections between the subcomponents end CompUnit.Sensori; system implementation PhysicalUnit.Skin Specify physical dynamics with the help of annexes end PhysicalUnit.Skin;

Modeling in AADL – Computing Units system Computing subcomponents P1: process SignalProcApp.impl; C1: system Radio.impl; end Computing; Computing Units – Embedded System Constructs system – sensors nodes in BAN subcomponents – sensor components (e.g. radio, processor, display device etc.) threads – application specific processes (e.g. FFT computation for signal processing applications property sets computing properties (e.g. operating frequency of processor) physical properties (e.g. power dissipation of subcomponents or threads) system implementation Radio.impl properties ComputingProperty::current => 18 mA; end Radio.impl process implementation SignalProcApp.impl subcomponents FFT: thread FFT_algorithm.imp1; end SignalProcApp.impl; thread implementation FFT_algorithm.imp1 modes RadioOn: initial mode ; RadioOff: mode ; properties ComputingProperty::current => 19.56 mA in modes (RadioOn); ComputingProperty::current => 1.0 mA in modes (RadioOff); end FFT_algorithm.imp1;

Networks of computing units data implementation Comp2CompData.impl subcomponents SignalStrength: data behavior::float; ParentID: data behavior::integer; end Comp2CompData; system - used for defining the network subcomponent – used for modeling the individual computing units (sensor nodes) port group – used for modeling connections between computing units port group Comp2CompPG features Packet: inout data port Comp2CompData.impl; end Comp2CompPG; system CompUnit features C2C: port group Comp2CompPG; end CompUnit; system implementation CompUnit.Sensori . . . end CompUnit.Sensori; Use of arrays required, not supported in AADL 1.0 system implementation BAN.ins1 subcomponents Sensor1: system CompUnit.Sensor1; Sensor2: system CompUnit.Sensor2; connections port group Sensor1.C2C -> Sensor2.C2CR; . . . end BAN.ins1; Replicate code for each sensor – scalable ??

Model to analyze Sustainability system implementation computing.sensor1 properties ComputingProperty::Voltage=> 2.3V end computing.sensor1; process implementation SignalProcApp subcomponents FFT: thread FFT_algorithm.imp1; end SignalProcApp; thread FFT_algorithm ComputeProperty::Compute_Execution_Time => 2138 ms .. 2140 ms; ComputeProperty ::Frequency => 30 Hz; end FFT_algorithm; thread implementation FFT_algorithm.imp1 modes RadioOn: initial mode ; RadioOff: mode ; ComputeProperty ::current => 19.56 mA in modes (RadioOn); ComputeProperty ::current => 1.0 mA in modes (RadioOff); end FFTComputation_algorithm.imp1; system BodyHeatSource ComputeProperty ::AveragePower=> 0.26W; end BodyHeatSource; Power consumption of the sensor nodes were modeled Scavenging sources were modeled for available power Duty cycling was performed on the sensor nodes to sustain their operation using the available power The sensor radio was turned off at appropriate times

Model to analyze side effects system implementation CompUnit.impl end CompUnit.impl; port group CyberPhysical features Info: inout data port Comp2PhysData.impl; end Comp2CompPG; property set Coefficient is SpecificHeat: constant aadlinteger =>3600; Fixed_blood_Temp :constant aadlinteger => 37; . . . end Coefficient; system PhysicalUnit P2C: port group CyberPhysical; end PhysicalUnit system implementation PhysicalUnit.impl subcomponents Del1Tt: data behavior::integer; Del2Tx: data behavior::integer; annex behavior_specification {** states s0 : initial complete state; transitions s0 -[ ]-> s0 { Del1Tt := (value(Coefficient ::SpecificHeat) * Del2Tx + value(Coefficient ::blood_perfusion_constant) * (Coefficient.T - value(Coefficient ::Fixed_blood_Temp) + PowerDissipation);}; **}; end PhysicalUnit.impl; Model the physical processes Specify differential equations Extended Behavior Annex Dedicated variables for parsing the differential operators Developed a parser to recognize the operators Developed a plug-in to convert the parsed form into solvable form Used FDTD solver to solve the equations Real Value initialization not supported in behavior annex Behavior annex properties must be constant requiring separate property set definition for each annex CPS specification using the behavior annex to represent differential equations system implementation BAN is subcomponents Sensor: system CompUnit.impl; Body: system PhysicalUnit.impl connections port group Sensor.C2P  Body.P2C; end BAN; system CompUnit features C2P: port group CyberPhysical; properties Physical Property - PowerDissipation end CompUnit; data implementation Comp2PhysData.impl PowerDissipation: data behavior::float; end Comp2CompData Multiple data subcomponents in port groups cannot be accessed in the behavior annex

Formal Modeling State space representation of the problem Declare appropriate states as UNSAFE Perform reachability analysis on the model Theoretical Guarantee on Safety and Sustainabiltiy Reduces Uncertainty of Simulation Issues: Current modeling techniques support dynamic variation in only one dimension Spatio-Temporal variation of interaction effects (ROIn and ROIm) require modeling and analysis in multiple dimensions (one time and three space). Scalability of the analysis technique on multiple dimensions Algorithm error increases with large number of variables Present day tools do not handle large number of variables.

System We study systems which can be represented using a finite number of states (finite state systems). Definition A set of states Set of initial states A set of inputs A transition relation A set of outputs An output map

Finite State Automata If H maps each state in X to an yes no answer The subset of inputs U for which the automata outputs yes is called the language Examples: DFA, PDA, Turing Machine

Dynamical System A dynamical system is a pair set of continuous variables is a set of differential equations Often the real space is divided into equivalence classes Q mapping of real space to equivalent classes Concept of operating modes

Example CRAC control system The outlet temperature is the variable belonging to the set V It follows the heat flow equation which is a member of the function set f Equivalence classes can be defined on the real space to denote different operating regions of the CRAC The COP varies in different operating regions

Hybrid Dynamical System S is a finite state system In is a set of invariants for each state Invariants are conditions on the continuous variables Gu is the set of guard conditions for each edge Re is a reset function If a state x is reached then what values will the continuous variables assume ? {In,f} is a dynamical system.

Example

Timed Automata Hybrid dynamical system In consists of only operators Gu can also consist of Re can either retain the value of the variable or set it to 0 f can either be 0 or 1.

Example

Hence a variation of hybrid automata which models spatio-temporal Formal Model for CPS Requirements: R1: The states in the formal model should represent both continuous and discrete domain operation R2: The state variables can have continuous dynamics with respect to both time and space, represented by complex partial differential equations R3: State transitions can take place through events occurring in both time and space continuum R4: Composition of individual formal models to derive models of the system should reflect the aggregate behavior Hence a variation of hybrid automata which models spatio-temporal

Spatio-Temporal Hybrid Automata Discrete Time Computational States Discrete Physical States Discrete States S1 S2 Continuous Variables Continuous variables related to physical phenomenon Initial State To simulate the operation of the system in time and space S1 S2 State Transitions Guard Conditions Spatio-Temporal Threshold Equations

Formal Modeling for Safety – single sensor Single sensor node and its associated thermal effect Notion of state is in space and time – I1 is the state representing space in ROIm N1 is the state representing space not in ROIm UNSAFE state Eq1 and Eq2 are the partial differential equations representing temperature rise in human body State transitions occur due to events generated in space and time – As we move through space if T1 < Tth a transition occurs from state I1 to N1 In time also if T1 < Tth a transition occurs from state I1 to N1 In any time at any particular state if T1 > Tsafe we go to unsafe state

Single sensor thermal profile Thermal profile over time and space for a single sensor

Composition of models Given individual models how to determine the model of the system State Space Cartesian Product Set of Continuous Variables Union S1 S2 S11 S21 Union including new functions to specify aggregate effects Set of Functions S12 S22 Transitions S1 S2 Union Retain old ones. If two models change state simultaneously then combine guard conditions using and operation Guard Conditions

Thermal Safety Example – model composition Multiple sensor nodes and their aggregate thermal effect Unsafe N1I2 I1N2 N1 N2 Eq3 = f(Eq1 , Eq2 ) Eq1,Eq2 I1 I2 T1 < Tth T1 > Tth T2 < Tth T2 > Tth T1 > Tsafe T2 > Tsafe Agg > Tsafe ∩ T1 > Tth ∩ T1 < Tth States are Cartesian products Eq3 represents aggregate effect (summation of heat) Transition from I1 ,I2 to state N1 ,N2 occurs due to a combination of events

Thermal map with aggregate effects

STHA Analysis Requirements Issues System dynamics in both space and time has to be analyzed Solving multi dimensional partial differential equations are required Intersection of ROIm or ROIn has to be computed Aggregate effects in the intersecting regions have to be computed Issues Tools performing reachability analysis can handle dynamics in only one dimension Multidimensional analysis requires discretization in all but one dimension This discretization introduces error in the analysis Drastically increases the number of dynamic variables Current tools cannot handle large number of variables

STHA Analysis Procedure x=3Δx Discretized Time t = 0 t = Δt t = 2Δt t = 3Δt t = 4Δt x=0 x=Δx x=2Δx x=(n-1)Δx x=nΔx Space discretization along x axis Hybrid System Reachability/Safety Analysis in continuous space (along y axis) Control Space S1 U S2 S3 CPS STHA modified to represent dynamics in y axis CPS STHA Reachability Analysis in successive time and space steps Usafe state Reachable Halt Computation U – denotes unsafe/unsustainable state S State not yet reached States that are reached

Conclusion and Future Work Conclusions: Spatio-Temporal Hybrid Automata for modeling CPS Model composition rules to take into account the aggregate effect of cyber-physical interactions Analysis algorithm for evaluating safety and sustainability of CPS Application of the modeling and analysis technique to three diverse case studies Implementation of the modeling and analysis technique using industry standard AADL Future Work: Apply STHA for medical device control systems An accurate reachability analysis for STHA Develop a STHA modeling and analysis tool

References Frehse, G. 2005. Phaver: Algorithmic verification of hybrid systems past hytech. In HSCC. 258-273. Bartocci et al, E. 2008a. Spatial Networks of Hybrid I/O Automata for Modeling Excitable Tissue. Electronic Notes in Theoretical Computer Science (ENTCS) 194, 3, 51-67. Chow, T. 1978. Testing software design modeled by finite-state machines. Software Engineering, IEEE Transactions on SE-4, 3 (May), 178-187. Henzinger, T. 1996. The theory of hybrid automata. Logic in Computer Science, Symposium on 0, 278. Moser et al, L. E. 1990. Formal verification of safety-critical systems. Softw. Pract. Exper. 20, 9, 799-811. www.aadl.info

Thank You

Definition STHA The model M, called the interaction model of individual computing unit in a CPS is a tuple M = {Q,X, F, Init,E,G} where: is a set of n + 1 discrete states. is a set of m continuous variables associated with the model. These variables are functions of time and space. denotes a set of spatio temporal partial differential equations for each state in Q which governs the variation of elements in X is a set of initial states. is a set of discrete transition relations between different discrete states in the model is a set of conditions on the continuous variables associated with each edge in the model

Composite Model Definition Composition of models and will result in the model following a model composition relation R. The relation R consists of the following clauses Clause 1: The set of discrete states Qc in the composite model Mc, is the Cartesian product of the two sets Q1 and Q2. However there is only one blocking state Clause 2: The set of continuous variables Xc is the union of the two sets X1 and X2. Clause 3: The set of functions specifies a method to combine the functions in the individual models to determine the cumulative effects of cyber-physical interactions.

Composite Model Definition Clause 4: set of initial states Clause 5: set of edges in Mc Clause 6: Gc specifies the conditions for state transition. Gc is a union of four sets

CPS Annex extension Specification of partial differential equation not supported in AADL CPSAnnex was developed to extend AADL with the facility to specify multi-dimensional partial and total differential equations New Constructs Del, Pdel for representing total and partial derivatives Grammar