Distributed Computing Sudarsun Santhiappan sudarsun@{burning-glass.com, gmail.com} Burning Glass Technologies Kilpauk, Chennai 600010

Technology is Changing... Computational Power gets Doubled every 18 months Networking Bandwidth and Speed getting Doubled every 9 months How to tap the benefits of this Technology ? Should we grow as an Individual ? Should we grow as a Team ?

Computational Power gets Doubled every 18 months

Networking Bandwidth and Speed getting Doubled every 9 months

How to tap the benefits of this Technology ?

Should we grow as an Individual ?

Should we grow as a Team ?

The Coverage Today Parallel Processing Multiprocessor or Multi-Core Computing Symmetric Multiprocessing Cluster Computing {PVM} Distributed Computing {TAO, OpenMP} Grid Computing {Globus Toolkit} Cloud Computing {Amazon EC2}

Parallel Processing

Multiprocessor or Multi-Core Computing

Symmetric Multiprocessing

Cluster Computing {PVM}

Distributed Computing {TAO, OpenMP}

Grid Computing {Globus Toolkit}

Cloud Computing {Amazon EC2}

Parallel Computing It is a form of computation in which many calculations are carried out simultaneously, operating on the principle that large problems can often be divided into smaller ones, which are then solved concurrently in parallel. Multi-Core, Multiprocessor SMP, Massively Parallel Processing (MPP) Computers Is it easy to write a parallel program ?

It is a form of computation in which many calculations are carried out simultaneously, operating on the principle that large problems can often be divided into smaller ones, which are then solved concurrently in parallel.

Multi-Core, Multiprocessor SMP, Massively Parallel Processing (MPP) Computers

Is it easy to write a parallel program ?

Cluster Computing A computer cluster is a group of linked computers, working together closely so that in many respects they form a single computer Operate in shared memory mode (mostly) Tightly coupled with high-speed networking, mostly with optical fiber channels. HA, Load Balancing, Compute Clusters Can we Load Balance using DNS ?

A computer cluster is a group of linked computers, working together closely so that in many respects they form a single computer

Operate in shared memory mode (mostly)

Tightly coupled with high-speed networking, mostly with optical fiber channels.

HA, Load Balancing, Compute Clusters

Can we Load Balance using DNS ?

Distributed Computing Wikipedia : It deals with hardware and software systems containing more than one processing element or storage element, concurrent processes, or multiple programs, running under a loosely or tightly controlled regime

Wikipedia : It deals with hardware and software systems containing more than one processing element or storage element, concurrent processes, or multiple programs, running under a loosely or tightly controlled regime

Grid Computing Wikipedia: A form of distributed computing whereby a super and virtual computer is composed of a cluster of networked, loosely-coupled computers, acting in concert to perform large tasks. pcwebopedia.com : Unlike conventional networks that focus on communication among devices, grid computing harnesses unused processing cycles of all computers in a network for solving problems too intensive for any stand-alone machine. IBM: Grid computing enables the virtualization of distributed computing and data resources such as processing, network bandwidth and storage capacity to create a single system image, granting users and applications seamless access to vast IT capabilities. Just as an Internet user views a unified instance of content via the Web, a grid user essentially sees a single, large virtual computer. Sun: Grid Computing is a computing infrastructure that provides dependable, consistent, pervasive and inexpensive access to computational capabilities.

Wikipedia: A form of distributed computing whereby a super and virtual computer is composed of a cluster of networked, loosely-coupled computers, acting in concert to perform large tasks.

pcwebopedia.com : Unlike conventional networks that focus on communication among devices, grid computing harnesses unused processing cycles of all computers in a network for solving problems too intensive for any stand-alone machine.

IBM: Grid computing enables the virtualization of distributed computing and data resources such as processing, network bandwidth and storage capacity to create a single system image, granting users and applications seamless access to vast IT capabilities. Just as an Internet user views a unified instance of content via the Web, a grid user essentially sees a single, large virtual computer.

Sun: Grid Computing is a computing infrastructure that provides dependable, consistent, pervasive and inexpensive access to computational capabilities.

Cloud Computing Wikipedia: It is a style of computing in which dynamically stable and often virtualised resources are provided as a service over the Internet. Infrastructure As A Service (IaaS) Platform As A Service (PaaS) Software as a Service (SaaS) Provide common business applications online accessible from a web browser. Amazon Elastic Computing, Google Apps

Wikipedia: It is a style of computing in which dynamically stable and often virtualised resources are provided as a service over the Internet.

Infrastructure As A Service (IaaS)

Platform As A Service (PaaS)

Software as a Service (SaaS)

Provide common business applications online accessible from a web browser.

Amazon Elastic Computing, Google Apps

Hardware: IBM p690 Regatta 32 POWER4 CPUs (1.1 GHz) 32 GB RAM 218 GB internal disk OS: AIX 5.1 Peak speed: 140.8 GFLOP/s * Programming model: shared memory multithreading (OpenMP) (also supports MPI) * GFLOP/s: billion floating point operations per second

270 Pentium4 XeonDP CPUs 270 GB RAM 8,700 GB disk OS: Red Hat Linux Enterprise 3 Peak speed: 1.08 TFLOP/s * Programming model: distributed multiprocessing (MPI) * TFLOP/s: trillion floating point operations per second Hardware: Pentium4 Xeon Cluster

56 Itanium2 1.0 GHz CPUs 112 GB RAM 5,774 GB disk OS: Red Hat Linux Enterprise 3 Peak speed: 224 GFLOP/s * Programming model: distributed multiprocessing (MPI) * GFLOP/s: billion floating point operations per second Hardware: Itanium2 Cluster schooner.oscer.ou.edu New arrival!

Vector Processing It is based on array processors where the instruction set includes operations that can perform mathematical operations on data elements simultaneously Example: Finding Scalar dot product between two vectors Is vector processing a parallel computing model? What are the limitations of Vector processing ? Extensively in Video processing & Games...

It is based on array processors where the instruction set includes operations that can perform mathematical operations on data elements simultaneously

Example: Finding Scalar dot product between two vectors

Is vector processing a parallel computing model?

What are the limitations of Vector processing ?

Extensively in Video processing & Games...

Pipelined Processing The fundamental idea is to split the processing of a computer instruction into a series of independent steps, with storage at the end of each step. This allows the computer's control circuitry to issue instructions at the processing rate of the slowest step, which is much faster than the time needed to perform all steps at once. A non-pipeline architecture is inefficient because some CPU components (modules) are idle while another module is active during the instruction cycle Processors with pipelining are organized inside into stages which can semi-independently work on separate jobs

The fundamental idea is to split the processing of a computer instruction into a series of independent steps, with storage at the end of each step.

This allows the computer's control circuitry to issue instructions at the processing rate of the slowest step, which is much faster than the time needed to perform all steps at once.

A non-pipeline architecture is inefficient because some CPU components (modules) are idle while another module is active during the instruction cycle

Processors with pipelining are organized inside into stages which can semi-independently work on separate jobs

Parallel Vs Pipelined Processing Parallel processing Pipelined processing a1 a2 a3 a4 b1 b2 b3 b4 c1 c2 c3 c4 d1 d2 d3 d4 a1 b1 c1 d1 a2 b2 c2 d2 a3 b3 c3 d3 a4 b4 c4 d4 P1 P2 P3 P4 P1 P2 P3 P4 time Colors: different types of operations performed a, b, c, d: different data streams processed Less inter-processor communication Complicated processor hardware time More inter-processor communication Simpler processor hardware

Parallel processing

Pipelined processing

Data Dependence Parallel processing requires NO data dependence between processors Pipelined processing will involve inter-processor communication P1 P2 P3 P4 P1 P2 P3 P4 time time

Parallel processing requires NO data dependence between processors

Pipelined processing will involve inter-processor communication

Typical Computing Elements Hardware Operating System Applications Programming paradigms P P P P P P   Microkernel Multi-Processor Computing System Threads Interface Process Processor Thread P

Why Parallel Processing ? Computation requirements are ever increasing; for instance -- visualization, distributed databases, simulations, scientific prediction (ex: climate, earthquake), etc. Sequential architectures reaching physical limitation (speed of light, thermodynamics) Limit on number of transistor per square inch Limit on inter-component link capacitance

Computation requirements are ever increasing; for instance -- visualization, distributed databases, simulations, scientific prediction (ex: climate, earthquake), etc.

Sequential architectures reaching physical limitation (speed of light, thermodynamics)

Limit on number of transistor per square inch

Limit on inter-component link capacitance

Symmetric Multiprocessing SMP Involves a multiprocessor computer architecture where two or more identical processors can connect to a single shared main memory Kernel can execute on any processor Typically each processor does self-scheduling form the pool of available process or threads Scalability problems in Uniform Memory Access NUMA to improve speed, but limitations on data migration Intel, AMD processors are SMP units What is ASMP ?

Involves a multiprocessor computer architecture where two or more identical processors can connect to a single shared main memory

Kernel can execute on any processor

Typically each processor does self-scheduling form the pool of available process or threads

Scalability problems in Uniform Memory Access

NUMA to improve speed, but limitations on data migration

Intel, AMD processors are SMP units

What is ASMP ?

SISD : A Conventional Computer Speed is limited by the rate at which computer can transfer information internally. Ex:PC, Macintosh, Workstations Processor Data Input Data Output Instructions

Speed is limited by the rate at which computer can transfer information internally.

The MISD Architecture More of an intellectual exercise than a practical configuration. Few built, but commercially not available Data Input Stream Data Output Stream Processor A Processor B Processor C Instruction Stream A Instruction Stream B Instruction Stream C

More of an intellectual exercise than a practical configuration. Few built, but commercially not available

SIMD Architecture Ex: CRAY machine vector processing, Intel MMX (multimedia support) C i <= A i * B i Instruction Stream Processor A Processor B Processor C Data Input stream A Data Input stream B Data Input stream C Data Output stream A Data Output stream B Data Output stream C

Unlike SISD, MISD, MIMD computer works asynchronously. Shared memory (tightly coupled) MIMD Distributed memory (loosely coupled) MIMD MIMD Architecture Processor A Processor B Processor C Data Input stream A Data Input stream B Data Input stream C Data Output stream A Data Output stream B Data Output stream C Instruction Stream A Instruction Stream B Instruction Stream C

Shared Memory MIMD machine Communication: Source Processor writes data to GM & destination retrieves it. Limitation : reliability & expandability A memory component or any processor failure affects the whole system. Increase of processors leads to memory contention. Ex. : Silicon graphics supercomputers.... Global Memory System Processor A Processor B Processor C MEMORY BUS MEMORY BUS MEMORY BUS

Communication: Source Processor writes data to GM & destination retrieves it.

Limitation : reliability & expandability A memory component or any processor failure affects the whole system.

Increase of processors leads to memory contention.

Distributed Memory MIMD Communication : IPC on High Speed Network. Network can be configured to ... Tree, Mesh, Cube, etc. Unlike Shared MIMD Readily expandable Highly reliable (any CPU failure does not affect the whole system) Processor A Processor B Processor C IPC channel IPC channel MEMORY BUS MEMORY BUS MEMORY BUS Memory System A Memory System B Memory System C

Communication : IPC on High Speed Network.

Network can be configured to ... Tree, Mesh, Cube, etc.

Unlike Shared MIMD

Readily expandable

Highly reliable (any CPU failure does not affect the whole system)

Laws of caution..... Speed of computers is proportional to the square of their cost. i.e. cost = Speed Speedup by a parallel computer increases as the logarithm of the number of processors. Speedup = log 2 (no. of processors) S P logP C S (speed = cost 2 )

Speed of computers is proportional to the square of their cost.

Speedup by a parallel computer increases as the logarithm of the number of processors.

Speedup = log 2 (no. of processors)

Micro Kernel based Operating Systems for High Performance Computing

Three approaches to building OS.... Monolithic OS Layered OS Microkernel based OS Client server OS Suitable for MPP systems Simplicity, flexibility and high performance are crucial for OS. Operating System Models

Monolithic OS

Layered OS

Microkernel based OS

Simplicity, flexibility and high performance are crucial for OS.

Monolithic Operating System Better application Performance Difficult to extend Ex: MS-DOS Application Programs Application Programs System Services Hardware User Mode Kernel Mode

Better application Performance

Difficult to extend

Layered OS Easier to enhance Each layer of code access lower level interface Low-application performance Application Programs System Services User Mode Kernel Mode Memory & I/O Device Mgmt Hardware Process Schedule Application Programs Ex : UNIX

Easier to enhance

Each layer of code access lower level interface

Low-application performance

Microkernel/Client Server OS (for MPP Systems) Tiny OS kernel providing basic primitive (process, memory, IPC) Traditional services becomes subsystems Monolithic Application Perf. Competence OS = Microkernel + User Subsystems Client Application Thread lib. File Server Network Server Display Server Microkernel Hardware Send Reply Ex: Mach, PARAS, Chorus, etc. User Kernel

Tiny OS kernel providing basic primitive (process, memory, IPC)

Traditional services becomes subsystems

Monolithic Application Perf. Competence

OS = Microkernel + User Subsystems

What are Micro Kernels ? Small operating system core Contains only essential core operating systems functions Many services traditionally included in the operating system are now external subsystems Device drivers File systems Virtual memory manager Windowing system Security services

Small operating system core

Contains only essential core operating systems functions

Many services traditionally included in the operating system are now external subsystems

Device drivers

File systems

Virtual memory manager

Windowing system

Security services

HPC Cluster Architecture Frontend Node Public Ethernet Private Ethernet Network Application Network (Optional) Power Distribution (Net addressable units as option) Node Node Node Node Node Node Node Node Node Node

Most Critical Problems with Clusters The largest problem in clusters is software skew When software configuration on some nodes is different than on others Small differences (minor version numbers on libraries) can cripple a parallel program The second most important problem is lack of adequate job control of the parallel process Signal propagation Cleanup

The largest problem in clusters is software skew

When software configuration on some nodes is different than on others

Small differences (minor version numbers on libraries) can cripple a parallel program

The second most important problem is lack of adequate job control of the parallel process

Signal propagation

Cleanup

Top 3 Problems with Software Packages Software installation works only in interactive mode Need a significant work by end-user Often rational default settings are not available Extremely time consuming to provide values Should be provided by package developers but … Package is required to be installed on a running system Means multi-step operation: install + update Intermediate state can be insecure

Software installation works only in interactive mode

Need a significant work by end-user

Often rational default settings are not available

Extremely time consuming to provide values

Should be provided by package developers but …

Package is required to be installed on a running system

Means multi-step operation: install + update

Intermediate state can be insecure

Clusters Classification..1 Based on Focus (in Market) High Performance (HP) Clusters Grand Challenging Applications High Availability (HA) Clusters Mission Critical applications

Based on Focus (in Market)

High Performance (HP) Clusters

Grand Challenging Applications

High Availability (HA) Clusters

Mission Critical applications

HA Cluster: Server Cluster with &quot;Heartbeat&quot; Connection

Clusters Classification..2 Based on Workstation/PC Ownership Dedicated Clusters Non-dedicated clusters Adaptive parallel computing Also called Communal multiprocessing

Based on Workstation/PC Ownership

Dedicated Clusters

Non-dedicated clusters

Adaptive parallel computing

Also called Communal multiprocessing

Clusters Classification..3 Based on Node Architecture .. Clusters of PCs (CoPs) Clusters of Workstations (COWs) Clusters of SMPs (CLUMPs)

Based on Node Architecture ..

Clusters of PCs (CoPs)

Clusters of Workstations (COWs)

Clusters of SMPs (CLUMPs)

Building Scalable Systems: Cluster of SMPs (Clumps) Performance of SMP Systems Vs. Four-Processor Servers in a Cluster

Clusters Classification..4 Based on Node OS Type .. Linux Clusters (Beowulf) Solaris Clusters (Berkeley NOW) NT Clusters (HPVM) AIX Clusters (IBM SP2) SCO/Compaq Clusters (Unixware) Digital VMS Clusters, HP clusters

Based on Node OS Type ..

Linux Clusters (Beowulf)

Solaris Clusters (Berkeley NOW)

NT Clusters (HPVM)

AIX Clusters (IBM SP2)

SCO/Compaq Clusters (Unixware)

Digital VMS Clusters, HP clusters

Clusters Classification..5 Based on Processor Arch, Node Type Homogeneous Clusters All nodes will have similar configuration Heterogeneous Clusters Nodes based on different processors and running different Operating Systems

Homogeneous Clusters

All nodes will have similar configuration

Heterogeneous Clusters

Nodes based on different processors and running different Operating Systems

Cluster Implementation What is Middleware ? What is Single System Image ? Benefits of Single System Image

What is Middleware ?

What is Single System Image ?

Benefits of Single System Image

What is Cluster Middle-ware ? An interface between user applications and cluster hardware and OS platform. Middle-ware packages support each other at the management, programming, and implementation levels. Middleware Layers: SSI Layer Availability Layer: It enables the cluster services of Checkpointing, Automatic Failover, recovery from failure, fault-tolerant operating among all cluster nodes.

An interface between user applications and cluster hardware and OS platform.

Middle-ware packages support each other at the management, programming, and implementation levels.

Middleware Layers:

SSI Layer

Availability Layer: It enables the cluster services of

Checkpointing, Automatic Failover, recovery from failure,

fault-tolerant operating among all cluster nodes.

Middleware Design Goals Complete Transparency Lets the see a single cluster system.. Single entry point, ftp, telnet, software loading... Scalable Performance Easy growth of cluster no change of API & automatic load distribution. Enhanced Availability Automatic Recovery from failures Employ checkpointing & fault tolerant technologies Handle consistency of data when replicated..

Complete Transparency

Lets the see a single cluster system..

Single entry point, ftp, telnet, software loading...

Scalable Performance

Easy growth of cluster

no change of API & automatic load distribution.

Enhanced Availability

Automatic Recovery from failures

Employ checkpointing & fault tolerant technologies

Handle consistency of data when replicated..

What is Single System Image (SSI) ? A single system image is the illusion , created by software or hardware, that a collection of computing elements appear as a single computing resource. SSI makes the cluster appear like a single machine to the user, to applications, and to the network. A cluster without a SSI is not a cluster

A single system image is the illusion , created by software or hardware, that a collection of computing elements appear as a single computing resource.

SSI makes the cluster appear like a single machine to the user, to applications, and to the network.

A cluster without a SSI is not a cluster

Benefits of Single System Image Usage of system resources transparently Improved reliability and higher availability Simplified system management Reduction in the risk of operator errors User need not be aware of the underlying system architecture to use these machines effectively

Usage of system resources transparently

Improved reliability and higher availability

Simplified system management

Reduction in the risk of operator errors

User need not be aware of the underlying system architecture to use these machines effectively

Distributed Computing No shared memory Communication among processes Send a message Receive a message Asynchronous Synchronous Synergy among processes

No shared memory

Communication among processes

Send a message

Receive a message

Asynchronous

Synchronous

Synergy among processes

Messages Messages are sequences of bytes moving between processes The sender and receiver must agree on the type structure of values in the message “ Marshalling”: data layout so that there is no ambiguity such as “four chars” v. “one integer”.

Messages are sequences of bytes moving between processes

The sender and receiver must agree on the type structure of values in the message

“ Marshalling”: data layout so that there is no ambiguity such as “four chars” v. “one integer”.

Message Passing Process A sends a data buffer as a message to process B. Process B waits for a message from A, and when it arrives copies it into its own local memory. No memory shared between A and B.

Process A sends a data buffer as a message to process B.

Process B waits for a message from A, and when it arrives copies it into its own local memory.

No memory shared between A and B.

Message Passing Obviously, Messages cannot be received before they are sent. A receiver waits until there is a message. Asynchronous Sender never blocks, even if infinitely many messages are waiting to be received Semi-asynchronous is a practical version of above with large but finite amount of buffering

Obviously,

Messages cannot be received before they are sent.

A receiver waits until there is a message.

Asynchronous

Sender never blocks, even if infinitely many messages are waiting to be received

Semi-asynchronous is a practical version of above with large but finite amount of buffering

Message Passing: Point to Point Q: send(m, P) Send message M to process P P: recv(x, Q) Receive message from process Q, and place it in variable x The message data Type of x must match that of m As if x := m

Q: send(m, P)

Send message M to process P

P: recv(x, Q)

Receive message from process Q, and place it in variable x

The message data

Type of x must match that of m

As if x := m

Broadcast One sender Q, multiple receivers P Not all receivers may receive at the same time Q: broadcast (m) Send message M to processes P: recv(x, Q) Receive message from process Q, and place it in variable x

One sender Q, multiple receivers P

Not all receivers may receive at the same time

Q: broadcast (m)

Send message M to processes

P: recv(x, Q)

Receive message from process Q, and place it in variable x

Synchronous Message Passing Sender blocks until receiver is ready to receive. Cannot send messages to self. No buffering.

Sender blocks until receiver is ready to receive.

Cannot send messages to self.

No buffering.

Asynchronous Message Passing Sender never blocks. Receiver receives when ready. Can send messages to self. Infinite buffering.

Sender never blocks.

Receiver receives when ready.

Can send messages to self.

Infinite buffering.

Message Passing Speed not so good Sender copies message into system buffers. Message travels the network. Receiver copies message from system buffers into local memory. Special virtual memory techniques help. Programming Quality less error-prone cf. shared memory

Speed not so good

Sender copies message into system buffers.

Message travels the network.

Receiver copies message from system buffers into local memory.

Special virtual memory techniques help.

Programming Quality

less error-prone cf. shared memory

Distributed Programs Spatially distributed programs A part here, a part there, … Parallel Synergy Temporally distributed programs Compute half today, half tomorrow Combine the results at the end Migratory programs Have computation, will travel

Spatially distributed programs

A part here, a part there, …

Parallel

Synergy

Temporally distributed programs

Compute half today, half tomorrow

Combine the results at the end

Migratory programs

Have computation, will travel

Technological Bases of Distributed+Parallel Programs Spatially distributed programs Message passing Temporally distributed programs Shared memory Migratory programs Serialization of data and programs

Spatially distributed programs

Message passing

Temporally distributed programs

Shared memory

Migratory programs

Serialization of data and programs

Technological Bases for Migratory programs Same CPU architecture X86, PowerPC, MIPS, SPARC, …, JVM Same OS + environment Be able to “checkpoint” suspend, and then resume computation without loss of progress

Same CPU architecture

X86, PowerPC, MIPS, SPARC, …, JVM

Same OS + environment

Be able to “checkpoint”

suspend, and

then resume computation

without loss of progress

Message Passing Libraries Programmer is responsible for initial data distribution, synchronization, and sending and receiving information Parallel Virtual Machine (PVM) Message Passing Interface (MPI) Bulk Synchronous Parallel model (BSP)

Programmer is responsible for initial data distribution, synchronization, and sending and receiving information

Parallel Virtual Machine (PVM)

Message Passing Interface (MPI)

Bulk Synchronous Parallel model (BSP)

BSP: Bulk Synchronous Parallel model Divides computation into supersteps In each superstep a processor can work on local data and send messages. At the end of the superstep, a barrier synchronization takes place and all processors receive the messages which were sent in the previous superstep

Divides computation into supersteps

In each superstep a processor can work on local data and send messages.

At the end of the superstep, a barrier synchronization takes place and all processors receive the messages which were sent in the previous superstep

BSP: Bulk Synchronous Parallel model http://www.bsp-worldwide.org/ Book: Rob H. Bisseling, “Parallel Scientific Computation: A Structured Approach using BSP and MPI,” Oxford University Press, 2004, 324 pages, ISBN 0-19-852939-2.

http://www.bsp-worldwide.org/

Book: Rob H. Bisseling, “Parallel Scientific Computation: A Structured Approach using BSP and MPI,” Oxford University Press, 2004, 324 pages, ISBN 0-19-852939-2.

BSP Library Small number of subroutines to implement process creation, remote data access, and bulk synchronization. Linked to C, Fortran, … programs

Small number of subroutines to implement

process creation,

remote data access, and

bulk synchronization.

Linked to C, Fortran, … programs

Portable Batch System (PBS) Prepare a .cmd file naming the program and its arguments properties of the job the needed resources Submit .cmd to the PBS Job Server: qsub command Routing and Scheduling: The Job Server examines .cmd details to route the job to an execution queue. allocates one or more cluster nodes to the job communicates with the Execution Servers (mom's) on the cluster to determine the current state of the nodes. When all of the needed are allocated, passes the .cmd on to the Execution Server on the first node allocated (the &quot;mother superior&quot;). Execution Server will login on the first node as the submitting user and run the .cmd file in the user's home directory. Run an installation defined prologue script. Gathers the job's output to the standard output and standard error It will execute installation defined epilogue script. Delivers stdout and stdout to the user.

Prepare a .cmd file

naming the program and its arguments

properties of the job

the needed resources

Submit .cmd to the PBS Job Server: qsub command

Routing and Scheduling: The Job Server

examines .cmd details to route the job to an execution queue.

allocates one or more cluster nodes to the job

communicates with the Execution Servers (mom's) on the cluster to determine the current state of the nodes.

When all of the needed are allocated, passes the .cmd on to the Execution Server on the first node allocated (the &quot;mother superior&quot;).

Execution Server

will login on the first node as the submitting user and run the .cmd file in the user's home directory.

Run an installation defined prologue script.

Gathers the job's output to the standard output and standard error

It will execute installation defined epilogue script.

Delivers stdout and stdout to the user.

TORQUE, an open source PBS Tera-scale Open-source Resource and QUEue manager (TORQUE) enhances OpenPBS Fault Tolerance Additional failure conditions checked/handled Node health check script support Scheduling Interface Scalability Significantly improved server to MOM communication model Ability to handle larger clusters (over 15 TF/2,500 processors) Ability to handle larger jobs (over 2000 processors) Ability to support larger server messages Logging http://www.supercluster.org/projects/torque/

Tera-scale Open-source Resource and QUEue manager (TORQUE) enhances OpenPBS

Fault Tolerance

Additional failure conditions checked/handled

Node health check script support

Scheduling Interface

Scalability

Significantly improved server to MOM communication model

Ability to handle larger clusters (over 15 TF/2,500 processors)

Ability to handle larger jobs (over 2000 processors)

Ability to support larger server messages

Logging

http://www.supercluster.org/projects/torque/

PVM, and MPI Message passing primitives Can be embedded in many existing programming languages Architecturally portable Open-sourced implementations

Message passing primitives

Can be embedded in many existing programming languages

Architecturally portable

Open-sourced implementations

Parallel Virtual Machine ( PVM ) PVM enables a heterogeneous collection of networked computers to be used as a single large parallel computer. Older than MPI Large scientific/engineering user community http://www.csm.ornl.gov/pvm/

PVM enables a heterogeneous collection of networked computers to be used as a single large parallel computer.

Older than MPI

Large scientific/engineering user community

http://www.csm.ornl.gov/pvm/

Message Passing Interface (MPI) http ://www-unix.mcs.anl.gov/mpi/ MPI-2.0 http://www.mpi-forum.org/docs/ MPI CH: www.mcs.anl.gov/mpi/mpich / by Argonne National Laboratory and Missisippy State University LAM: http://www.lam-mpi.org/ http://www.open-mpi.org/

http ://www-unix.mcs.anl.gov/mpi/

MPI-2.0 http://www.mpi-forum.org/docs/

MPI CH: www.mcs.anl.gov/mpi/mpich / by Argonne National Laboratory and Missisippy State University

LAM: http://www.lam-mpi.org/

http://www.open-mpi.org/

OpenMP for shared memory Distributed shared memory API User-gives hints as directives to the compiler http://www.openmp.org

Distributed shared memory API

User-gives hints as directives to the compiler

http://www.openmp.org

SPMD Single program, multiple data Contrast with SIMD Same program runs on multiple nodes May or may not be lock-step Nodes may be of different speeds Barrier synchronization

Single program, multiple data

Contrast with SIMD

Same program runs on multiple nodes

May or may not be lock-step

Nodes may be of different speeds

Barrier synchronization

Condor Cooperating workstations: come and go. Migratory programs Checkpointing Remote IO Resource matching http://www.cs.wisc.edu/condor/

Cooperating workstations: come and go.

Migratory programs

Checkpointing

Remote IO

Resource matching

http://www.cs.wisc.edu/condor/

Migration of Jobs Policies Immediate-Eviction Pause-and-Migrate Technical Issues Check-pointing: Preserving the state of the process so it can be resumed. Migrating from one architecture to another

Policies

Immediate-Eviction

Pause-and-Migrate

Technical Issues

Check-pointing: Preserving the state of the process so it can be resumed.

Migrating from one architecture to another

OpenMosix Distro Quantian Linux Boot from DVD-ROM Compressed file system on DVD Several GB of cluster software http:// dirk.eddelbuettel.com/quantian.html Live CD/DVD or Single Floppy Bootables http://bofh.be/clusterknoppix/ http://sentinix.org/ http://itsecurity.mq.edu.au/chaos/ http://openmosixloaf.sourceforge.net/ http://plumpos.sourceforge.net/ http://www.dynebolic.org/ http://bccd.cs.uni.edu/ http://eucaristos.sourceforge.net/ http://gomf.sourceforge.net/ Can be installed on HDD

Quantian Linux

Boot from DVD-ROM

Compressed file system on DVD

Several GB of cluster software

http:// dirk.eddelbuettel.com/quantian.html

Live CD/DVD or Single Floppy Bootables

http://bofh.be/clusterknoppix/

http://sentinix.org/

http://itsecurity.mq.edu.au/chaos/

http://openmosixloaf.sourceforge.net/

http://plumpos.sourceforge.net/

http://www.dynebolic.org/

http://bccd.cs.uni.edu/

http://eucaristos.sourceforge.net/

http://gomf.sourceforge.net/

Can be installed on HDD

What is openMOSIX? An open source enhancement to the Linux kernel Cluster with come-and-go nodes System image model: Virtual machine with lots of memory and CPU Granularity: Process Improves the overall (cluster-wide) performance. Multi-user, time-sharing environment for the execution of both sequential and parallel applications Applications unmodified (no need to link with special library)

An open source enhancement to the Linux kernel

Cluster with come-and-go nodes

System image model: Virtual machine with lots of memory and CPU

Granularity: Process

Improves the overall (cluster-wide) performance.

Multi-user, time-sharing environment for the execution of both sequential and parallel applications

Applications unmodified (no need to link with special library)

What is openMOSIX? Execution environment: farm of diskless x86 based nodes UP (uniprocessor), or SMP (symmetric multi processor) connected by standard LAN (e.g., Fast Ethernet) Adaptive resource management to dynamic load characteristics CPU, RAM, I/O, etc. Linear scalability

Execution environment:

farm of diskless x86 based nodes

UP (uniprocessor), or

SMP (symmetric multi processor)

connected by standard LAN (e.g., Fast Ethernet)

Adaptive resource management to dynamic load characteristics

CPU, RAM, I/O, etc.

Linear scalability

Users’ View of the Cluster Users can start from any node in the cluster, or sysadmin sets-up a few nodes as login nodes Round-robin DNS: “hpc.clusters” with many IPs assigned to same name Each process has a Home-Node Migrated processes always appear to run at the home node, e.g., “ps” show all your processes, even if they run elsewhere

Users can start from any node in the cluster, or sysadmin sets-up a few nodes as login nodes

Round-robin DNS: “hpc.clusters” with many IPs assigned to same name

Each process has a Home-Node

Migrated processes always appear to run at the home node, e.g., “ps” show all your processes, even if they run elsewhere

MOSIX architecture network transparency preemptive process migration dynamic load balancing memory sharing efficient kernel communication probabilistic information dissemination algorithms decentralized control and autonomy

network transparency

preemptive process migration

dynamic load balancing

memory sharing

efficient kernel communication

probabilistic information dissemination algorithms

decentralized control and autonomy

A two tier technology Information gathering and dissemination Support scalable configurations by probabilistic dissemination algorithms Same overhead for 16 nodes or 2056 nodes Pre-emptive process migration that can migrate any process, anywhere, anytime - transparently Supervised by adaptive algorithms that respond to global resource availability Transparent to applications, no change to user interface

Information gathering and dissemination

Support scalable configurations by probabilistic dissemination algorithms

Same overhead for 16 nodes or 2056 nodes

Pre-emptive process migration that can migrate any process, anywhere, anytime - transparently

Supervised by adaptive algorithms that respond to global resource availability

Transparent to applications, no change to user interface

Tier 1: Information gathering and dissemination In each unit of time (e.g., 1 second) each node gathers information about: CPU(s) speed, load and utilization Free memory Free proc-table/file-table slots Info sent to a randomly selected node Scalable - more nodes better scattering

In each unit of time (e.g., 1 second) each node gathers information about:

CPU(s) speed, load and utilization

Free memory

Free proc-table/file-table slots

Info sent to a randomly selected node

Scalable - more nodes better scattering

Tier 2: Process migration Load balancing: reduce variance between pairs of nodes to improve the overall performance Memory ushering: migrate processes from a node that nearly exhausted its free memory, to prevent paging Parallel File I/O: bring the process to the file-server, direct file I/O from migrated processes

Load balancing: reduce variance between pairs of nodes to improve the overall performance

Memory ushering: migrate processes from a node that nearly exhausted its free memory, to prevent paging

Parallel File I/O: bring the process to the file-server, direct file I/O from migrated processes

Network transparency The user and applications are provided a virtual machine that looks like a single machine. Example: Disk access from diskless nodes on fileserver is completely transparent to programs

The user and applications are provided a virtual machine that looks like a single machine.

Example: Disk access from diskless nodes on fileserver is completely transparent to programs

Preemptive process migration Any user’s process, trasparently and at any time, can/may migrate to any other node. The migrating process is divided into: system context ( deputy ) that may not be migrated from home workstation (UHN); user context ( remote ) that can be migrated on a diskless node;

Any user’s process, trasparently and at any time, can/may migrate to any other node.

The migrating process is divided into:

system context ( deputy ) that may not be migrated from home workstation (UHN);

user context ( remote ) that can be migrated on a diskless node;

Splitting the Linux process System context (environment) - site dependent- “home” confined Connected by an exclusive link for both synchronous (system calls) and asynchronous (signals, MOSIX events) Process context (code, stack, data) - site independent - may migrate Deputy Remote Kernel Kernel Userland Userland openMOSIX Link Local master node diskless node

System context (environment) - site dependent- “home” confined

Connected by an exclusive link for both synchronous (system calls) and asynchronous (signals, MOSIX events)

Process context (code, stack, data) - site independent - may migrate

Dynamic load balancing Initiates process migrations in order to balance the load of farm responds to variations in the load of the nodes, runtime characteristics of the processes, number of nodes and their speeds makes continuous attempts to reduce the load differences among nodes the policy is symmetrical and decentralized all of the nodes execute the same algorithm the reduction of the load differences is performed indipendently by any pair of nodes

Initiates process migrations in order to balance the load of farm

responds to variations in the load of the nodes, runtime characteristics of the processes, number of nodes and their speeds

makes continuous attempts to reduce the load differences among nodes

the policy is symmetrical and decentralized

all of the nodes execute the same algorithm

the reduction of the load differences is performed indipendently by any pair of nodes

The ACE ORB What Is CORBA? CORBA Basics Clients, Servers, and Servants ORBs and POAs IDL and the Role of IDL Compilers IORs Tying it all together Overview of ACE/TAO CORBA Services Naming Service Trading Service Event Service Multi-Threaded Issues Using CORBA

What Is CORBA?

CORBA Basics

Clients, Servers, and Servants

ORBs and POAs

IDL and the Role of IDL Compilers

IORs

Tying it all together

Overview of ACE/TAO

CORBA Services

Naming Service

Trading Service

Event Service

Multi-Threaded Issues Using CORBA

What Is CORBA? C ommon O bject R equest B roker A rchitecture Common Architecture Object Request Broker – ORB Specification from the OMG http://www.omg.org/technology/documents/corba_spec_catalog.htm Must be implemented before usable

C ommon O bject R equest B roker A rchitecture

Common Architecture

Object Request Broker – ORB

Specification from the OMG

http://www.omg.org/technology/documents/corba_spec_catalog.htm

Must be implemented before usable

What Is CORBA? More specifically: “ ( CORBA ) is a standard defined by the Object Management Group (OMG) that enables software components written in multiple computer languages and running on multiple computers to work together ” (1) Allows for Object Interoperability, regardless of: Operating Systems Programming Language Takes care of Marshalling and Unmarshalling of Data A method to perform Distributed Computing

More specifically:

“ ( CORBA ) is a standard defined by the Object Management Group (OMG) that enables software components written in multiple computer languages and running on multiple computers to work together ” (1)

Allows for Object Interoperability, regardless of:

Operating Systems

Programming Language

Takes care of Marshalling and Unmarshalling of Data

A method to perform Distributed Computing

What Is CORBA? Program A Running on a Windows PC Written in Java Program B Running on a Linux Machine Written in C++ CORBA

Running on a Windows PC

Written in Java

Running on a Linux Machine

Written in C++

CORBA Basics: Clients, Servers, and Servants CORBA Clients An Application (program) Request services from Servant object Invoke a method call Can exist on a different computer from Servant Can also exist on same computer, or even within the same program, as the Servant Implemented by Software Developer

CORBA Clients

An Application (program)

Request services from Servant object

Invoke a method call

Can exist on a different computer from Servant

Can also exist on same computer, or even within the same program, as the Servant

Implemented by Software Developer

CORBA Basics: Clients, Servers, and Servants CORBA Servers An Application (program) Performs setup needed to get Servants configured properly ORB’s, POA’s Instantiates and starts Servants object(s) Once configuration done and Servant(s) running, Clients can begin to send messages Implemented by Software Developer

CORBA Servers

An Application (program)

Performs setup needed to get Servants configured properly

ORB’s, POA’s

Instantiates and starts Servants object(s)

Once configuration done and Servant(s) running, Clients can begin to send messages

Implemented by Software Developer

CORBA Basics: Clients, Servers, and Servants Servants Objects Implement interfaces Respond to Client requests Exists within the same program as the Server that created and started it Implemented by Software Developer

Servants

Objects

Implement interfaces

Respond to Client requests

Exists within the same program as the Server that created and started it

Implemented by Software Developer

ORB’s and POA’s ORB: Object Request Broker The “ORB” in “CORBA” At the heart of CORBA Enables communication Implemented by ORB Vendor An organization that implements the CORBA Specification (a company, a University, etc.) Can be viewed as an API/Framework Set of classes and method Used by Clients and Servers to properly setup communication Client and Server ORB’s communicate over a network Glue between Client and Server applications

ORB: Object Request Broker

The “ORB” in “CORBA”

At the heart of CORBA

Enables communication

Implemented by ORB Vendor

An organization that implements the CORBA Specification (a company, a University, etc.)

Can be viewed as an API/Framework

Set of classes and method

Used by Clients and Servers to properly setup communication

Client and Server ORB’s communicate over a network

Glue between Client and Server applications

ORB’s and POA’s POA: Portable Object Adapter A central CORBA goal: Programs using different ORB’s (provided by different ORB Vendors) can still communicate The POA was adopted as the solution Can be viewed as an API/Framework Set of classes and method Sits between ORB’s and Servants Glue between Servants and ORBs Job is to: Receive messages from ORB’s Activate the appropriate Servant Deliver the message to the Servant

POA: Portable Object Adapter

A central CORBA goal: Programs using different ORB’s (provided by different ORB Vendors) can still communicate

The POA was adopted as the solution

Can be viewed as an API/Framework

Set of classes and method

Sits between ORB’s and Servants

Glue between Servants and ORBs

Job is to:

Receive messages from ORB’s

Activate the appropriate Servant

Deliver the message to the Servant

CORBA Basics: IDL IDL: The Interface Definition Language Keyword: Definition No “executable” code (cannot implement anything) Very similar to C++ Header Files Language independent from Target Language Allows Client and Server applications to be written in different (several) languages A “contract” between Clients and Servers Both MUST have the exact same IDL Specifies messages and data that can be sent by Clients and received by Servants Written by Software Developer

IDL: The Interface Definition Language

Keyword: Definition

No “executable” code (cannot implement anything)

Very similar to C++ Header Files

Language independent from Target Language

Allows Client and Server applications to be written in different (several) languages

A “contract” between Clients and Servers

Both MUST have the exact same IDL

Specifies messages and data that can be sent by Clients and received by Servants

Written by Software Developer

CORBA Basics: IDL Used to define interfaces (i.e. Servants) Classes and methods that provide services IDL Provides… Primitive Data Types (int, float, boolean, char, string) Ability to compose primitives into more complex data structures Enumerations, Unions, Arrays, etc. Object-Oriented Inheritance

Used to define interfaces (i.e. Servants)

Classes and methods that provide services

IDL Provides…

Primitive Data Types (int, float, boolean, char, string)

Ability to compose primitives into more complex data structures

Enumerations, Unions, Arrays, etc.

Object-Oriented Inheritance

CORBA Basics: IDL IDL Compilers Converts IDL files to target language files Done via Language Mappings Useful to understand your Language Mapping scheme Target language files contain all the implementation code that facilitates CORBA-based communication More or less “hides” the details from you Creates client “stubs” and Server “skeletons” Provided by ORB Vendor

IDL Compilers

Converts IDL files to target language files

Done via Language Mappings

Useful to understand your Language Mapping scheme

Target language files contain all the implementation code that facilitates CORBA-based communication

More or less “hides” the details from you

Creates client “stubs” and Server “skeletons”

Provided by ORB Vendor

CORBA Basics: IDL IDL File IDL Compiler Client Stub Files Server Skeleton Files Generates Generates Generated Files are in Target Language: C++ Java etc. Generated Files are in Target Language: C++ Java etc. Client Programs used the classes in the Client Stub files to send messages to the Servant objects Client Program Servant Object Servant Objects inherit from classes in the Server Skeleton files to receive messages from the Client programs Association Inheritance

C++

Java

etc.

C++

Java

etc.

CORBA Basics: IDL Can also generate empty Servant class files IDL Compiler converts to C++ (in this case)

Can also generate empty Servant class files

CORBA Basics: IOR’s IOR: Interoperable Object Reference Can be thought of as a “Distributed Pointer” Unique to each Servant Used by ORB’s and POA’s to locate Servants For Clients, used to find Servants across networks For Servers, used to find proper Servant running within the application Opaque to Client and Server applications Only meaningful to ORB’s and POA’s Contains information about IP Address, Port Numbers, networking protocols used, etc. The difficult part is obtaining them This is the purpose/reasoning behind developing and using CORBA Services

IOR: Interoperable Object Reference

Can be thought of as a “Distributed Pointer”

Unique to each Servant

Used by ORB’s and POA’s to locate Servants

For Clients, used to find Servants across networks

For Servers, used to find proper Servant running within the application

Opaque to Client and Server applications

Only meaningful to ORB’s and POA’s

Contains information about IP Address, Port Numbers, networking protocols used, etc.

The difficult part is obtaining them

This is the purpose/reasoning behind developing and using CORBA Services

CORBA Basics: IOR’s Can be viewed in “stringified” format, but… Still not very meaningful

Can be viewed in “stringified” format, but…

Still not very meaningful

CORBA Basics: IOR’s Standardized, to some degree: … … Standardized by the OMG: Used by Client side ORB’s to locate Server side (destination) ORB’s Contains information needed to make physical connection NOT Standardized by the OMG; proprietary to ORB Vendors Used by Server side ORB’s and POA’s to locate destination Servants

Standardized, to some degree:

Used by Client side ORB’s to locate Server side (destination) ORB’s

Contains information needed to make physical connection

Used by Server side ORB’s and POA’s to locate destination Servants

CORBA Basics: Tying it All Together

CORBA Basics: Tying it All Together Client Program IOR (Servant Ref) Server Program Servant Message(Data) Logical Flow Client Program Server Program Servant Actual Flow POA ORB IOR (Servant Ref) ORB Once ORB’s and POA’s set up and configured properly, transparency is possible ORB’s communicate over network POA’s activate servants and deliver messages

ORB’s communicate over network

POA’s activate servants and deliver messages

Overview of ACE/TAO ACE: Adaptive Communications Environment Object-Oriented Framework/API Implements many concurrent programming design patterns Can be used to build more complex communications-based packages For example, an ORB

ACE: Adaptive Communications Environment

Object-Oriented Framework/API

Implements many concurrent programming design patterns

Can be used to build more complex communications-based packages

For example, an ORB

Overview of ACE/TAO TAO: The ACE ORB Built on top of ACE A CORBA implementation Includes many (if not all) CORBA features specified by the OMG Not just an ORB Provides POA’s, CORBA Services, etc. Object-Oriented Framework/API

TAO: The ACE ORB

Built on top of ACE

A CORBA implementation

Includes many (if not all) CORBA features specified by the OMG

Not just an ORB

Provides POA’s, CORBA Services, etc.

Object-Oriented Framework/API

CORBA Services: The Naming Service The CORBA Naming Service is similar to the White Pages (phone book) Servants place their “names,” along with their IOR’s, into the Naming Service The Naming Service stores these as pairs Later, Clients obtain IOR’s from the Naming Service by passing the name of the Servant object to it The Naming Service returns the IOR Clients may then use to make requests

The CORBA Naming Service is similar to the White Pages (phone book)

Servants place their “names,” along with their IOR’s, into the Naming Service

The Naming Service stores these as pairs

Later, Clients obtain IOR’s from the Naming Service by passing the name of the Servant object to it

The Naming Service returns the IOR

Clients may then use to make requests

CORBA Services: The Trading Service The CORBA Naming Service is similar to the Yellow Pages (phone book) Servants place a description of the services they can provide (i.e. their “Trades”), along with their IOR’s, into the Trading Services The Trading Service stores these Clients obtain IOR’s from the Trading Service by passing the type(s) of Services they require The Trading Service returns an IOR Clients may then use to make requests

The CORBA Naming Service is similar to the Yellow Pages (phone book)

Servants place a description of the services they can provide (i.e. their “Trades”), along with their IOR’s, into the Trading Services

The Trading Service stores these

Clients obtain IOR’s from the Trading Service by passing the type(s) of Services they require

The Trading Service returns an IOR

Clients may then use to make requests

Multi-Threaded Issues Using CORBA Server performance can be improved by using multiple threads GUI Thread Listening Thread Processing Thread Can also use multiple ORBs and POAs to improve performance Requires a multi-threaded solution

Server performance can be improved by using multiple threads

GUI Thread

Listening Thread

Processing Thread

Can also use multiple ORBs and POAs to improve performance

Requires a multi-threaded solution

What is Grid Computing? Computational Grids Homogeneous (e.g., Clusters) Heterogeneous (e.g., with one-of-a-kind instruments) Cousins of Grid Computing Methods of Grid Computing

Computational Grids

Homogeneous (e.g., Clusters)

Heterogeneous (e.g., with one-of-a-kind instruments)

Cousins of Grid Computing

Methods of Grid Computing

Computational Grids A network of geographically distributed resources including computers, peripherals, switches, instruments, and data. Each user should have a single login account to access all resources. Resources may be owned by diverse organizations.

A network of geographically distributed resources including computers, peripherals, switches, instruments, and data.

Each user should have a single login account to access all resources.

Resources may be owned by diverse organizations.

Computational Grids Grids are typically managed by gridware. Gridware can be viewed as a special type of middleware that enable sharing and manage grid components based on user requirements and resource attributes (e.g., capacity, performance, availability…)

Grids are typically managed by gridware.

Gridware can be viewed as a special type of middleware that enable sharing and manage grid components based on user requirements and resource attributes (e.g., capacity, performance, availability…)

Cousins of Grid Computing Parallel Computing Distributed Computing Peer-to-Peer Computing Many others: Cluster Computing, Network Computing, Client/Server Computing, Internet Computing, etc...

Parallel Computing

Distributed Computing

Peer-to-Peer Computing

Many others: Cluster Computing, Network Computing, Client/Server Computing, Internet Computing, etc...

Distributed Computing People often ask: Is Grid Computing a fancy new name for the concept of distributed computing? In general, the answer is “no.” Distributed Computing is most often concerned with distributing the load of a program across two or more processes.

People often ask: Is Grid Computing a fancy new name for the concept of distributed computing?

In general, the answer is “no.” Distributed Computing is most often concerned with distributing the load of a program across two or more processes.

PEER2PEER Computing Sharing of computer resources and services by direct exchange between systems. Computers can act as clients or servers depending on what role is most efficient for the network.

Sharing of computer resources and services by direct exchange between systems.

Computers can act as clients or servers depending on what role is most efficient for the network.

Methods of Grid Computing Distributed Supercomputing High-Throughput Computing On-Demand Computing Data-Intensive Computing Collaborative Computing Logistical Networking

Distributed Supercomputing

High-Throughput Computing

On-Demand Computing

Data-Intensive Computing

Collaborative Computing

Logistical Networking

Distributed Supercomputing Combining multiple high-capacity resources on a computational grid into a single, virtual distributed supercomputer. Tackle problems that cannot be solved on a single system.

Combining multiple high-capacity resources on a computational grid into a single, virtual distributed supercomputer.

Tackle problems that cannot be solved on a single system.

High-Throughput Computing Uses the grid to schedule large numbers of loosely coupled or independent tasks, with the goal of putting unused processor cycles to work.

Uses the grid to schedule large numbers of loosely coupled or independent tasks, with the goal of putting unused processor cycles to work.

On-Demand Computing Uses grid capabilities to meet short-term requirements for resources that are not locally accessible. Models real-time computing demands.

Uses grid capabilities to meet short-term requirements for resources that are not locally accessible.

Models real-time computing demands.

Data-Intensive Computing The focus is on synthesizing new information from data that is maintained in geographically distributed repositories, digital libraries, and databases. Particularly useful for distributed data mining.

The focus is on synthesizing new information from data that is maintained in geographically distributed repositories, digital libraries, and databases.

Particularly useful for distributed data mining.

Collaborative Computing Concerned primarily with enabling and enhancing human-to-human interactions. Applications are often structured in terms of a virtual shared space.

Concerned primarily with enabling and enhancing human-to-human interactions.

Applications are often structured in terms of a virtual shared space.

Logistical Networking Global scheduling and optimization of data movement. Contrasts with traditional networking, which does not explicitly model storage resources in the network. Called &quot;logistical&quot; because of the analogy it bears with the systems of warehouses, depots, and distribution channels.

Global scheduling and optimization of data movement.

Contrasts with traditional networking, which does not explicitly model storage resources in the network.

Called &quot;logistical&quot; because of the analogy it bears with the systems of warehouses, depots, and distribution channels.

Globus A collaboration of Argonne National Laboratory’s Mathematics and Computer Science Division, the University of Southern California’s Information Sciences Institute, and the University of Chicago's Distributed Systems Laboratory. Started in 1996 and is gaining popularity year after year.

A collaboration of Argonne National Laboratory’s Mathematics and Computer Science Division, the University of Southern California’s Information Sciences Institute, and the University of Chicago's Distributed Systems Laboratory.

Started in 1996 and is gaining popularity year after year.

Globus A project to develop the underlying technologies needed for the construction of computational grids. Focuses on execution environments for integrating widely-distributed computational platforms, data resources, displays, special instruments and so forth.

A project to develop the underlying technologies needed for the construction of computational grids.

Focuses on execution environments for integrating widely-distributed computational platforms, data resources, displays, special instruments and so forth.

The Globus Toolkit The Globus Resource Allocation Manager (GRAM) Creates, monitors, and manages services. Maps requests to local schedulers and computers. The Grid Security Infrastructure (GSI) Provides authentication services.

The Globus Resource Allocation Manager (GRAM)

Creates, monitors, and manages services.

Maps requests to local schedulers and computers.

The Grid Security Infrastructure (GSI)

Provides authentication services.

The Globus Toolkit The Monitoring and Discovery Service (MDS) Provides information about system status, including server configurations, network status, and locations of replicated datasets, etc. Nexus and globus_io provides communication services for heterogeneous environments.

The Monitoring and Discovery Service (MDS)

Provides information about system status, including server configurations, network status, and locations of replicated datasets, etc.

Nexus and globus_io

provides communication services for heterogeneous environments.

What are Clouds? Clouds are “Virtual Clusters” (“Virtual Grids”) of possibly “Virtual Machines” They may cross administrative domains or may “just be a single cluster”; the user cannot and does not want to know Clouds support access (lease of) computer instances Instances accept data and job descriptions (code) and return results that are data and status flags Each Cloud is a “Narrow” (perhaps internally proprietary) Grid Clouds can be built from Grids Grids can be built from Clouds

Clouds are “Virtual Clusters” (“Virtual Grids”) of possibly “Virtual Machines”

They may cross administrative domains or may “just be a single cluster”; the user cannot and does not want to know

Clouds support access (lease of) computer instances

Instances accept data and job descriptions (code) and return results that are data and status flags

Each Cloud is a “Narrow” (perhaps internally proprietary) Grid

Clouds can be built from Grids

Grids can be built from Clouds

Virtualization and Cloud Computing The Virtues of Virtualization Portable environments, enforcement and isolation, fast to deploy, suspend/resume, migration… Cloud computing SaaS: software as a service Service: provide me with a workspace Virtualization makes it easy to provide a workspace/VM Cloud computing resource leasing, utility computing, elastic computing Amazon’s Elastic Compute Cloud (EC2) Is this real? Or is this just a proof-of-concept? Successfully used commercially on a large scale More experience for scientific applications Virtual Workspaces: http//workspace.globus.org

The Virtues of Virtualization

Portable environments, enforcement and isolation, fast to deploy, suspend/resume, migration…

Cloud computing

SaaS: software as a service

Service: provide me with a workspace

Virtualization makes it easy to provide a workspace/VM

Cloud computing

resource leasing, utility computing, elastic computing

Amazon’s Elastic Compute Cloud (EC2)

Is this real? Or is this just a proof-of-concept?

Successfully used commercially on a large scale

More experience for scientific applications

Two major types of cloud Compute and Data Cloud EC2, Google Map Reduce, Science clouds Provision platform for running science codes Open source infrastructure: workspace, eucalyptus, hub0 Virtualization: providing environments as VMs Hosting Cloud GoogleApp Engine Highly-available, fault tolerance, robustness, etc for Web capabilities Community example: IU hosting environment (quarry) Virtual Workspaces: http//workspace.globus.org

Compute and Data Cloud

EC2, Google Map Reduce, Science clouds

Provision platform for running science codes

Open source infrastructure: workspace, eucalyptus, hub0

Virtualization: providing environments as VMs

Hosting Cloud

GoogleApp Engine

Highly-available, fault tolerance, robustness, etc for Web capabilities

Community example: IU hosting environment (quarry)

Technical Questions on Clouds How is data compute affinity tackled in clouds? Co-locate data and compute clouds? Lots of optical fiber i.e. “just” move the data? What happens in clouds when demand for resources exceeds capacity – is there a multi-day job input queue? Are there novel cloud scheduling issues? Do we want to link clouds (or ensembles as atomic clouds); if so how and with what protocols Is there an intranet cloud e.g. “cloud in a box” software to manage personal (cores on my future 128 core laptop) department or enterprise cloud?

How is data compute affinity tackled in clouds?

Co-locate data and compute clouds?

Lots of optical fiber i.e. “just” move the data?

What happens in clouds when demand for resources exceeds capacity – is there a multi-day job input queue?

Are there novel cloud scheduling issues?

Do we want to link clouds (or ensembles as atomic clouds); if so how and with what protocols

Is there an intranet cloud e.g. “cloud in a box” software to manage personal (cores on my future 128 core laptop) department or enterprise cloud?

Thanks Much.. 99% of the slides are taken from the Internet from various Authors. Thanks to all of them! Sudarsun Santhiappan Director – R & D Burning Glass Technologies Kilpauk, Chennai 600010

99% of the slides are taken from the Internet from various Authors. Thanks to all of them!

Sudarsun Santhiappan

Director – R & D

Burning Glass Technologies

Kilpauk, Chennai 600010