Download as ODP, PPTX
Distributed Computing
- 1. Distributed Computing SudarsunSanthiappan sudarsun@{burning-glass.com, gmail.com} Burning Glass Technologies Kilpauk, Chennai 600010
- 2. Technology is Changing...Computational Power gets Doubled every 18 months
- 3. Networking Bandwidth andSpeed getting Doubled every 9 months
- 4. How to tapthe benefits of this Technology ?
- 5. Should we growas an Individual ?
- 6. Should we growas a Team ?
- 7. The Coverage TodayParallel Processing
- 8. Multiprocessor or Multi-CoreComputing
- 9.
- 10.
- 11. Distributed Computing {TAO,OpenMP}
- 12. Grid Computing {GlobusToolkit}
- 13.
- 14. Parallel Computing Itis 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.
- 15. Multi-Core, Multiprocessor SMP,Massively Parallel Processing (MPP) Computers
- 16. Is it easyto write a parallel program ?
- 17. Cluster Computing Acomputer cluster is a group of linked computers, working together closely so that in many respects they form a single computer
- 18. Operate in sharedmemory mode (mostly)
- 19. Tightly coupled withhigh-speed networking, mostly with optical fiber channels.
- 20. HA, Load Balancing,Compute Clusters
- 21. Can we LoadBalance using DNS ?
- 22. 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
- 23. 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.
- 24. pcwebopedia.com : Unlikeconventional 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.
- 25. IBM: Gridcomputing 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.
- 26. Sun: GridComputing is a computing infrastructure that provides dependable, consistent, pervasive and inexpensive access to computational capabilities.
- 27. 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.
- 28. Infrastructure As AService (IaaS)
- 29. Platform As AService (PaaS)
- 30. Software as aService (SaaS)
- 31. Provide common businessapplications online accessible from a web browser.
- 32. Amazon Elastic Computing,Google Apps
- 33. Hardware: IBM p690Regatta 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
- 34. 270 Pentium4 XeonDPCPUs 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
- 35. 56 Itanium2 1.0GHz 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!
- 36. Vector Processing Itis based on array processors where the instruction set includes operations that can perform mathematical operations on data elements simultaneously
- 37. Example: Finding Scalardot product between two vectors
- 38. Is vector processinga parallel computing model?
- 39. What are thelimitations of Vector processing ?
- 40. Extensively in Videoprocessing & Games...
- 41. Pipelined Processing Thefundamental idea is to split the processing of a computer instruction into a series of independent steps, with storage at the end of each step.
- 42. This allows thecomputer'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.
- 43. A non-pipeline architectureis inefficient because some CPU components (modules) are idle while another module is active during the instruction cycle
- 44. Processors with pipeliningare organized inside into stages which can semi-independently work on separate jobs
- 45. Parallel Vs PipelinedProcessing 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
- 46. Data Dependence Parallelprocessing requires NO data dependence between processors Pipelined processing will involve inter-processor communication P1 P2 P3 P4 P1 P2 P3 P4 time time
- 47. Typical Computing ElementsHardware Operating System Applications Programming paradigms P P P P P P Microkernel Multi-Processor Computing System Threads Interface Process Processor Thread P
- 48. Why Parallel Processing? Computation requirements are ever increasing; for instance -- visualization, distributed databases, simulations, scientific prediction (ex: climate, earthquake), etc.
- 49. Sequential architectures reachingphysical limitation (speed of light, thermodynamics)
- 50. Limit on numberof transistor per square inch
- 51. Limit on inter-componentlink capacitance
- 52. Symmetric Multiprocessing SMPInvolves a multiprocessor computer architecture where two or more identical processors can connect to a single shared main memory
- 53. Kernel can executeon any processor
- 54. Typically each processordoes self-scheduling form the pool of available process or threads
- 55. Scalability problems inUniform Memory Access
- 56. NUMA to improvespeed, but limitations on data migration
- 57. Intel, AMD processorsare SMP units
- 58.
- 59.
- 60.
- 61. SISD : AConventional Computer Speed is limited by the rate at which computer can transfer information internally. Ex:PC, Macintosh, Workstations Processor Data Input Data Output Instructions
- 62. The MISD ArchitectureMore 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
- 63. 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
- 64. 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
- 65. Shared Memory MIMDmachine Communication: Source Processor writes data to GM & destination retrieves it.
- 66. 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
- 67. Distributed Memory MIMDCommunication : IPC on High Speed Network.
- 68. Network can beconfigured to ... Tree, Mesh, Cube, etc.
- 69. Unlike Shared MIMDReadily expandable
- 70. Highly reliable (anyCPU 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
- 71. 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 )
- 72. Micro Kernel basedOperating Systems for High Performance Computing
- 73. Three approaches tobuilding OS.... Monolithic OS
- 74.
- 75. Microkernel based OS Client server OS Suitable for MPP systems Simplicity, flexibility and high performance are crucial for OS. Operating System Models
- 76. Monolithic Operating SystemBetter application Performance
- 77. Difficult to extendEx: MS-DOS Application Programs Application Programs System Services Hardware User Mode Kernel Mode
- 78. Layered OS Easierto enhance
- 79. Each layer ofcode access lower level interface
- 80. Low-application performance ApplicationPrograms System Services User Mode Kernel Mode Memory & I/O Device Mgmt Hardware Process Schedule Application Programs Ex : UNIX
- 81. Microkernel/Client Server OS(for MPP Systems) Tiny OS kernel providing basic primitive (process, memory, IPC)
- 82. Traditional services becomes subsystems
- 83. Monolithic Application Perf.Competence
- 84. 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
- 85. What are MicroKernels ? Small operating system core
- 86. Contains only essentialcore operating systems functions
- 87. Many services traditionallyincluded in the operating system are now external subsystems Device drivers
- 88.
- 89.
- 90.
- 91.
- 92.
- 93. HPC Cluster ArchitectureFrontend 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
- 94. Most Critical Problemswith Clusters The largest problem in clusters is software skew When software configuration on some nodes is different than on others
- 95. Small differences (minorversion 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
- 96.
- 97. Top 3 Problemswith 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
- 98. Should be providedby package developers but … Package is required to be installed on a running system Means multi-step operation: install + update
- 99. Intermediate state canbe insecure
- 100. Clusters Classification..1 Basedon Focus (in Market) High Performance (HP) Clusters Grand Challenging Applications High Availability (HA) Clusters Mission Critical applications
- 101. HA Cluster: ServerCluster with "Heartbeat" Connection
- 102. Clusters Classification..2 Basedon Workstation/PC Ownership Dedicated Clusters
- 103. Non-dedicated clusters Adaptiveparallel computing
- 104. Also called Communalmultiprocessing
- 105. Clusters Classification..3 Basedon Node Architecture .. Clusters of PCs (CoPs)
- 106. Clusters of Workstations(COWs)
- 107. Clusters of SMPs(CLUMPs)
- 108. Building Scalable Systems: Cluster of SMPs (Clumps) Performance of SMP Systems Vs. Four-Processor Servers in a Cluster
- 109. Clusters Classification..4 Basedon Node OS Type .. Linux Clusters (Beowulf)
- 110. Solaris Clusters (Berkeley NOW)
- 111.
- 112.
- 113.
- 114. Digital VMS Clusters,HP clusters
- 115. Clusters Classification..5 Basedon Processor Arch, Node Type Homogeneous Clusters All nodes will have similar configuration Heterogeneous Clusters Nodes based on different processors and running different Operating Systems
- 116. Cluster Implementation Whatis Middleware ?
- 117. What is SingleSystem Image ?
- 118. Benefits of SingleSystem Image
- 119. What is ClusterMiddle-ware ? An interface between user applications and cluster hardware and OS platform.
- 120. Middle-ware packages supporteach other at the management, programming, and implementation levels.
- 121.
- 122. Availability Layer: Itenables the cluster services of Checkpointing, Automatic Failover, recovery from failure,
- 123. fault-tolerant operating amongall cluster nodes.
- 124. Middleware Design GoalsComplete 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..
- 125. What is SingleSystem 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.
- 126. A cluster withouta SSI is not a cluster
- 127. Benefits of SingleSystem Image Usage of system resources transparently
- 128. Improved reliability andhigher availability
- 129.
- 130. Reduction in therisk of operator errors
- 131. User need notbe aware of the underlying system architecture to use these machines effectively
- 132. Distributed Computing Noshared memory
- 133. Communication among processesSend a message
- 134. Receive a messageAsynchronous
- 135.
- 136.
- 137. Messages Messages aresequences of bytes moving between processes
- 138. The sender andreceiver must agree on the type structure of values in the message
- 139. “ Marshalling”: datalayout so that there is no ambiguity such as “four chars” v. “one integer”.
- 140. Message Passing ProcessA sends a data buffer as a message to process B.
- 141. Process B waitsfor a message from A, and when it arrives copies it into its own local memory.
- 142. No memory sharedbetween A and B.
- 143. Message Passing Obviously, Messages cannot be received before they are sent.
- 144. A receiver waitsuntil there is a message. Asynchronous Sender never blocks, even if infinitely many messages are waiting to be received
- 145. Semi-asynchronous is apractical version of above with large but finite amount of buffering
- 146. Message Passing: Pointto 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
- 147. As if x:= m
- 148. Broadcast One senderQ, multiple receivers P
- 149. Not all receiversmay receive at the same time
- 150. Q: broadcast(m) Send message M to processes P: recv(x, Q) Receive message from process Q, and place it in variable x
- 151. Synchronous Message PassingSender blocks until receiver is ready to receive.
- 152. Cannot send messagesto self.
- 153.
- 154. Asynchronous Message PassingSender never blocks.
- 155. Receiver receives whenready. Can send messages to self. Infinite buffering.
- 156. Message Passing Speednot so good Sender copies message into system buffers.
- 157. Message travels thenetwork.
- 158. Receiver copies messagefrom system buffers into local memory.
- 159. Special virtual memorytechniques help. Programming Quality less error-prone cf. shared memory
- 160. Distributed Programs Spatiallydistributed programs A part here, a part there, …
- 161.
- 162. Synergy Temporally distributedprograms Compute half today, half tomorrow
- 163. Combine the resultsat the end Migratory programs Have computation, will travel
- 164. Technological Bases ofDistributed+Parallel Programs Spatially distributed programs Message passing Temporally distributed programs Shared memory Migratory programs Serialization of data and programs
- 165. Technological Bases forMigratory programs Same CPU architecture X86, PowerPC, MIPS, SPARC, …, JVM Same OS + environment
- 166. Be able to“checkpoint” suspend, and
- 167.
- 168. without loss ofprogress
- 169. Message Passing LibrariesProgrammer is responsible for initial data distribution, synchronization, and sending and receiving information
- 170.
- 171.
- 172. Bulk Synchronous Parallelmodel (BSP)
- 173. BSP: Bulk SynchronousParallel model Divides computation into supersteps
- 174. In each superstepa processor can work on local data and send messages.
- 175. At the endof the superstep, a barrier synchronization takes place and all processors receive the messages which were sent in the previous superstep
- 176. BSP: Bulk SynchronousParallel model http://www.bsp-worldwide.org/
- 177. Book: RobH. Bisseling, “Parallel Scientific Computation: A Structured Approach using BSP and MPI,” Oxford University Press, 2004, 324 pages, ISBN 0-19-852939-2.
- 178. BSP Library Smallnumber of subroutines to implement process creation,
- 179.
- 180. bulk synchronization. Linkedto C, Fortran, … programs
- 181. Portable Batch System(PBS) Prepare a .cmd file naming the program and its arguments
- 182.
- 183. the needed resources Submit .cmd to the PBS Job Server: qsub command
- 184. Routing and Scheduling:The Job Server examines .cmd details to route the job to an execution queue.
- 185. allocates one ormore cluster nodes to the job
- 186. communicates with theExecution Servers (mom's) on the cluster to determine the current state of the nodes.
- 187. When all ofthe needed are allocated, passes the .cmd on to the Execution Server on the first node allocated (the "mother superior"). Execution Server will login on the first node as the submitting user and run the .cmd file in the user's home directory.
- 188. Run an installationdefined prologue script.
- 189. Gathers the job'soutput to the standard output and standard error
- 190. It will executeinstallation defined epilogue script.
- 191. Delivers stdout andstdout to the user.
- 192. TORQUE, an opensource PBS Tera-scale Open-source Resource and QUEue manager (TORQUE) enhances OpenPBS
- 193. Fault Tolerance Additional failure conditions checked/handled
- 194. Node health checkscript support Scheduling Interface
- 195. Scalability Significantlyimproved server to MOM communication model
- 196. Ability to handlelarger clusters (over 15 TF/2,500 processors)
- 197. Ability to handlelarger jobs (over 2000 processors)
- 198. Ability to supportlarger server messages Logging
- 199.
- 200. PVM, and MPIMessage passing primitives
- 201. Can be embeddedin many existing programming languages
- 202.
- 203.
- 204. Parallel Virtual Machine( PVM ) PVM enables a heterogeneous collection of networked computers to be used as a single large parallel computer.
- 205.
- 206. Large scientific/engineering usercommunity
- 207.
- 208. Message Passing Interface(MPI) http ://www-unix.mcs.anl.gov/mpi/
- 209.
- 210. MPI CH: www.mcs.anl.gov/mpi/mpich / by Argonne National Laboratory and Missisippy State University
- 211.
- 212.
- 213. OpenMP for sharedmemory Distributed shared memory API
- 214. User-gives hints asdirectives to the compiler
- 215.
- 216. SPMD Single program,multiple data
- 217.
- 218. Same program runson multiple nodes
- 219. May or maynot be lock-step
- 220. Nodes may beof different speeds
- 221.
- 222. Condor Cooperating workstations:come and go.
- 223.
- 224. Remote IO Resourcematching
- 225.
- 226. Migration of JobsPolicies Immediate-Eviction
- 227. Pause-and-Migrate Technical IssuesCheck-pointing: Preserving the state of the process so it can be resumed.
- 228. Migrating from onearchitecture to another
- 229. OpenMosix Distro QuantianLinux Boot from DVD-ROM
- 230. Compressed file systemon DVD
- 231. Several GB ofcluster software
- 232. http:// dirk.eddelbuettel.com/quantian.html LiveCD/DVD or Single Floppy Bootables http://bofh.be/clusterknoppix/
- 233.
- 234.
- 235.
- 236.
- 237.
- 238.
- 239.
- 240. http://gomf.sourceforge.net/ Canbe installed on HDD
- 241. What is openMOSIX?An open source enhancement to the Linux kernel
- 242.
- 243. System image model:Virtual machine with lots of memory and CPU
- 244.
- 245. Improves the overall(cluster-wide) performance.
- 246. Multi-user, time-sharing environmentfor the execution of both sequential and parallel applications
- 247. Applications unmodified (noneed to link with special library)
- 248. What is openMOSIX?Execution environment: farm of diskless x86 based nodes
- 249.
- 250. SMP (symmetric multiprocessor)
- 251. connected by standardLAN (e.g., Fast Ethernet) Adaptive resource management to dynamic load characteristics CPU, RAM, I/O, etc. Linear scalability
- 252. Users’ View ofthe Cluster Users can start from any node in the cluster, or sysadmin sets-up a few nodes as login nodes
- 253. Round-robin DNS: “hpc.clusters” with many IPs assigned to same name
- 254. Each process hasa Home-Node Migrated processes always appear to run at the home node, e.g., “ps” show all your processes, even if they run elsewhere
- 255. MOSIX architecture networktransparency
- 256.
- 257.
- 258.
- 259.
- 260. probabilistic information disseminationalgorithms
- 261. decentralized control andautonomy
- 262. A two tiertechnology Information gathering and dissemination Support scalable configurations by probabilistic dissemination algorithms
- 263. Same overhead for16 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
- 264. Transparent to applications,no change to user interface
- 265. Tier 1: Informationgathering and dissemination In each unit of time (e.g., 1 second) each node gathers information about: CPU(s) speed, load and utilization
- 266.
- 267. Free proc-table/file-table slotsInfo sent to a randomly selected node
- 268. Scalable - morenodes better scattering
- 269. Tier 2: Processmigration Load balancing: reduce variance between pairs of nodes to improve the overall performance
- 270. Memory ushering: migrateprocesses from a node that nearly exhausted its free memory, to prevent paging
- 271. Parallel FileI/O: bring the process to the file-server, direct file I/O from migrated processes
- 272. Network transparency Theuser and applications are provided a virtual machine that looks like a single machine.
- 273. Example: Disk accessfrom diskless nodes on fileserver is completely transparent to programs
- 274. Preemptive process migrationAny user’s process, trasparently and at any time, can/may migrate to any other node.
- 275. The migrating processis divided into: system context ( deputy ) that may not be migrated from home workstation (UHN);
- 276. user context (remote ) that can be migrated on a diskless node;
- 277. Splitting the Linuxprocess System context (environment) - site dependent- “home” confined
- 278. Connected by anexclusive link for both synchronous (system calls) and asynchronous (signals, MOSIX events)
- 279. Process context (code,stack, data) - site independent - may migrate Deputy Remote Kernel Kernel Userland Userland openMOSIX Link Local master node diskless node
- 280. Dynamic load balancingInitiates process migrations in order to balance the load of farm
- 281. responds to variationsin the load of the nodes, runtime characteristics of the processes, number of nodes and their speeds
- 282. makes continuous attemptsto reduce the load differences among nodes
- 283. the policy issymmetrical and decentralized all of the nodes execute the same algorithm
- 284. the reduction ofthe load differences is performed indipendently by any pair of nodes
- 285. The ACE ORBWhat Is CORBA?
- 286. CORBA Basics Clients,Servers, and Servants
- 287.
- 288. IDL and theRole of IDL Compilers
- 289.
- 290. Tying it alltogether Overview of ACE/TAO
- 291. CORBA Services NamingService
- 292.
- 293. Event Service Multi-ThreadedIssues Using CORBA
- 294. What Is CORBA?C ommon O bject R equest B roker A rchitecture Common Architecture
- 295. Object Request Broker– ORB Specification from the OMG http://www.omg.org/technology/documents/corba_spec_catalog.htm
- 296. Must be implementedbefore usable
- 297. 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)
- 298. Allows for ObjectInteroperability, regardless of: Operating Systems
- 299.
- 300. Takes care ofMarshalling and Unmarshalling of Data A method to perform Distributed Computing
- 301. What Is CORBA?Program A Running on a Windows PC
- 302. Written in JavaProgram B Running on a Linux Machine
- 303. Written in C++CORBA
- 304. CORBA Basics: Clients,Servers, and Servants CORBA Clients An Application (program)
- 305. Request services fromServant 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
- 306. CORBA Basics: Clients,Servers, and Servants CORBA Servers An Application (program)
- 307. Performs setup neededto get Servants configured properly ORB’s, POA’s Instantiates and starts Servants object(s)
- 308. Once configuration doneand Servant(s) running, Clients can begin to send messages
- 309. Implemented by SoftwareDeveloper
- 310. CORBA Basics: Clients,Servers, and Servants Servants Objects
- 311.
- 312. Respond to Clientrequests
- 313. Exists within thesame program as the Server that created and started it
- 314. Implemented by SoftwareDeveloper
- 315. ORB’s and POA’sORB: Object Request Broker The “ORB” in “CORBA” At the heart of CORBA Enables communication
- 316. Implemented by ORBVendor 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
- 317. Glue between Clientand Server applications
- 318. ORB’s and POA’sPOA: Portable Object Adapter A central CORBA goal: Programs using different ORB’s (provided by different ORB Vendors) can still communicate
- 319. The POA wasadopted as the solution
- 320. Can be viewedas 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
- 321. Activate the appropriateServant
- 322. Deliver the messageto the Servant
- 323. CORBA Basics: IDLIDL: The Interface Definition Language Keyword: Definition No “executable” code (cannot implement anything)
- 324. Very similar toC++ Header Files
- 325. Language independent fromTarget 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
- 326. Specifies messages anddata that can be sent by Clients and received by Servants Written by Software Developer
- 327. CORBA Basics: IDLUsed to define interfaces (i.e. Servants) Classes and methods that provide services IDL Provides… Primitive Data Types (int, float, boolean, char, string)
- 328. Ability to composeprimitives into more complex data structures
- 329.
- 330.
- 331. CORBA Basics: IDLIDL Compilers Converts IDL files to target language files
- 332. Done via LanguageMappings 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”
- 333. Provided by ORBVendor
- 334. CORBA Basics: IDLIDL File IDL Compiler Client Stub Files Server Skeleton Files Generates Generates Generated Files are in Target Language: C++
- 335.
- 336. etc. Generated Filesare in Target Language: C++
- 337.
- 338. etc. Client Programsused 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
- 339. CORBA Basics: IDLCan also generate empty Servant class files IDL Compiler converts to C++ (in this case)
- 340. CORBA Basics: IOR’sIOR: Interoperable Object Reference Can be thought of as a “Distributed Pointer”
- 341. Unique to eachServant
- 342. Used by ORB’sand POA’s to locate Servants For Clients, used to find Servants across networks
- 343. For Servers, usedto find proper Servant running within the application Opaque to Client and Server applications Only meaningful to ORB’s and POA’s
- 344. Contains information aboutIP Address, Port Numbers, networking protocols used, etc. The difficult part is obtaining them This is the purpose/reasoning behind developing and using CORBA Services
- 345. CORBA Basics: IOR’sCan be viewed in “stringified” format, but… Still not very meaningful
- 346. CORBA Basics: IOR’sStandardized, to some degree: … … Standardized by the OMG: Used by Client side ORB’s to locate Server side (destination) ORB’s
- 347. Contains information neededto 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
- 348. CORBA Basics: Tyingit All Together
- 349. CORBA Basics: Tyingit 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
- 350. POA’s activate servantsand deliver messages
- 351. Overview of ACE/TAOACE: Adaptive Communications Environment Object-Oriented Framework/API
- 352. Implements many concurrentprogramming design patterns
- 353. Can be usedto build more complex communications-based packages For example, an ORB
- 354. Overview of ACE/TAOTAO: The ACE ORB Built on top of ACE
- 355.
- 356. Includes many (ifnot all) CORBA features specified by the OMG Not just an ORB
- 357. Provides POA’s, CORBAServices, etc. Object-Oriented Framework/API
- 358. CORBA Services: TheNaming Service The CORBA Naming Service is similar to the White Pages (phone book)
- 359. 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
- 360. CORBA Services: TheTrading Service The CORBA Naming Service is similar to the Yellow Pages (phone book)
- 361. Servants place adescription 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
- 362. Multi-Threaded Issues UsingCORBA Server performance can be improved by using multiple threads GUI Thread
- 363.
- 364. Processing Thread Canalso use multiple ORBs and POAs to improve performance Requires a multi-threaded solution
- 365. What is GridComputing? Computational Grids Homogeneous (e.g., Clusters)
- 366. Heterogeneous (e.g., withone-of-a-kind instruments) Cousins of Grid Computing
- 367. Methods of GridComputing
- 368. Computational Grids Anetwork of geographically distributed resources including computers, peripherals, switches, instruments, and data.
- 369. Each user shouldhave a single login account to access all resources.
- 370. Resources may beowned by diverse organizations.
- 371. Computational Grids Gridsare typically managed by gridware.
- 372. Gridware can beviewed 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…)
- 373. Cousins of GridComputing Parallel Computing
- 374.
- 375.
- 376. Many others: ClusterComputing, Network Computing, Client/Server Computing, Internet Computing, etc...
- 377. Distributed Computing Peopleoften ask: Is Grid Computing a fancy new name for the concept of distributed computing?
- 378. In general, theanswer is “no.” Distributed Computing is most often concerned with distributing the load of a program across two or more processes.
- 379. PEER2PEER Computing Sharingof computer resources and services by direct exchange between systems.
- 380. Computers can actas clients or servers depending on what role is most efficient for the network.
- 381. Methods of GridComputing Distributed Supercomputing
- 382.
- 383.
- 384.
- 385.
- 386.
- 387. Distributed Supercomputing Combiningmultiple high-capacity resources on a computational grid into a single, virtual distributed supercomputer.
- 388. Tackle problems thatcannot be solved on a single system.
- 389. High-Throughput Computing Usesthe grid to schedule large numbers of loosely coupled or independent tasks, with the goal of putting unused processor cycles to work.
- 390. On-Demand Computing Usesgrid capabilities to meet short-term requirements for resources that are not locally accessible.
- 391. Models real-time computingdemands.
- 392. Data-Intensive Computing Thefocus is on synthesizing new information from data that is maintained in geographically distributed repositories, digital libraries, and databases.
- 393. Particularly useful fordistributed data mining.
- 394. Collaborative Computing Concernedprimarily with enabling and enhancing human-to-human interactions.
- 395. Applications are oftenstructured in terms of a virtual shared space.
- 396. Logistical Networking Globalscheduling and optimization of data movement.
- 397. Contrasts with traditionalnetworking, which does not explicitly model storage resources in the network.
- 398. Called "logistical" becauseof the analogy it bears with the systems of warehouses, depots, and distribution channels.
- 399. Globus A collaborationof 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.
- 400. Started in 1996and is gaining popularity year after year.
- 401. Globus A projectto develop the underlying technologies needed for the construction of computational grids.
- 402. Focuses on executionenvironments for integrating widely-distributed computational platforms, data resources, displays, special instruments and so forth.
- 403. The Globus ToolkitThe Globus Resource Allocation Manager (GRAM) Creates, monitors, and manages services.
- 404. Maps requests tolocal schedulers and computers. The Grid Security Infrastructure (GSI) Provides authentication services.
- 405. The Globus ToolkitThe 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.
- 406. 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
- 407. Clouds can bebuilt from Grids
- 408. Grids can bebuilt from Clouds
- 409. Virtualization and CloudComputing The Virtues of Virtualization Portable environments, enforcement and isolation, fast to deploy, suspend/resume, migration… Cloud computing SaaS: software as a service
- 410. Service: provide mewith a workspace
- 411. Virtualization makes iteasy to provide a workspace/VM Cloud computing resource leasing, utility computing, elastic computing
- 412. Amazon’s Elastic ComputeCloud (EC2) Is this real? Or is this just a proof-of-concept? Successfully used commercially on a large scale
- 413. More experience forscientific applications Virtual Workspaces: http//workspace.globus.org
- 414. Two major typesof cloud Compute and Data Cloud EC2, Google Map Reduce, Science clouds
- 415. Provision platform forrunning science codes
- 416. Open source infrastructure:workspace, eucalyptus, hub0
- 417. Virtualization: providing environmentsas VMs Hosting Cloud GoogleApp Engine
- 418. Highly-available, fault tolerance,robustness, etc for Web capabilities
- 419. Community example: IUhosting environment (quarry) Virtual Workspaces: http//workspace.globus.org
- 420. Technical Questions onClouds How is data compute affinity tackled in clouds? Co-locate data and compute clouds?
- 421. Lots of opticalfiber 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
- 422. Is there anintranet cloud e.g. “cloud in a box” software to manage personal (cores on my future 128 core laptop) department or enterprise cloud?
- 423. Thanks Much.. 99%of the slides are taken from the Internet from various Authors. Thanks to all of them!
- 424.
- 425. Director – R& D
- 426.
- 427.