Scalable Distributed Memory Multiprocessors Todd C. Mowry CS 495 October 24 & 29, 2002.

Slides:



Advertisements
Similar presentations
System Area Network Abhiram Shandilya 12/06/01. Overview Introduction to System Area Networks SAN Design and Examples SAN Applications.
Advertisements

ECE669 L3: Design Issues February 5, 2004 ECE 669 Parallel Computer Architecture Lecture 3 Design Issues.
1 (Review of Prerequisite Material). Processes are an abstraction of the operation of computers. So, to understand operating systems, one must have a.
1/1/ / faculty of Electrical Engineering eindhoven university of technology Architectures of Digital Information Systems Part 1: Interrupts and DMA dr.ir.
Graduate Computer Architecture, Fall 2005 Lecture 10 Distributed Memory Multiprocessors Shih-Hao Hung Computer Science & Information Engineering National.
1/1/ / faculty of Electrical Engineering eindhoven university of technology Introduction Part 3: Input/output and co-processors dr.ir. A.C. Verschueren.
CS 258 Parallel Computer Architecture Lecture 15.1 DASH: Directory Architecture for Shared memory Implementation, cost, performance Daniel Lenoski, et.
Computer Systems/Operating Systems - Class 8
Realizing Programming Models CS 258, Spring 99 David E. Culler Computer Science Division U.C. Berkeley.
Latency Tolerance: what to do when it just won’t go away CS 258, Spring 99 David E. Culler Computer Science Division U.C. Berkeley.
VIA and Its Extension To TCP/IP Network Yingping Lu Based on Paper “Queue Pair IP, …” by Philip Buonadonna.
ECE669 L20: Evaluation and Message Passing April 13, 2004 ECE 669 Parallel Computer Architecture Lecture 20 Evaluation and Message Passing.
EECC756 - Shaaban #1 lec # 13 Spring Scalable Distributed Memory Machines Goal: Parallel machines that can be scaled to hundreds or thousands.
Scalability CS 258, Spring 99 David E. Culler Computer Science Division U.C. Berkeley.
Fundamental Design Issues for Parallel Architecture Todd C. Mowry CS 495 January 22, 2002.
EECC756 - Shaaban #1 lec # 12 Spring Scalable Distributed Memory Machines Goal: Parallel machines that can be scaled to hundreds or thousands.
I/O Hardware n Incredible variety of I/O devices n Common concepts: – Port – connection point to the computer – Bus (daisy chain or shared direct access)
EECS 570: Fall rev3 1 Chapter 7 (excl. 7.9): Scalable Multiprocessors.
Active Messages: a Mechanism for Integrated Communication and Computation von Eicken et. al. Brian Kazian CS258 Spring 2008.
CS252 Graduate Computer Architecture Lecture 17 Multiprocessor Networks (con’t) March 31 th, 2010 John Kubiatowicz Electrical Engineering and Computer.
Distributed Memory Multiprocessors CS 252, Spring 2005 David E. Culler Computer Science Division U.C. Berkeley.
Chapter 13: I/O Systems I/O Hardware Application I/O Interface
General Purpose Node-to-Network Interface in Scalable Multiprocessors CS 258, Spring 99 David E. Culler Computer Science Division U.C. Berkeley.
7/2/2015 slide 1 PCOD: Scalable Parallelism (ICs) Per Stenström (c) 2008, Sally A. McKee (c) 2011 Scalable Multiprocessors What is a scalable design? (7.1)
Node-to-Network Interface in Scalable Multiprocessors CS 258, Spring 99 David E. Culler Computer Science Division U.C. Berkeley.
CS 258 Parallel Computer Architecture Lecture 8 Network Interface Design February 20, 2008 Prof John D. Kubiatowicz
Chapter 13: I/O Systems I/O Hardware Application I/O Interface
Switching, routing, and flow control in interconnection networks.
CS-334: Computer Architecture
Introduction to Interconnection Networks. Introduction to Interconnection network Digital systems(DS) are pervasive in modern society. Digital computers.
Synchronization and Communication in the T3E Multiprocessor.
Hardware Definitions –Port: Point of connection –Bus: Interface Daisy Chain (A=>B=>…=>X) Shared Direct Device Access –Controller: Device Electronics –Registers:
Switched network.
CHAPTER 3 TOP LEVEL VIEW OF COMPUTER FUNCTION AND INTERCONNECTION
Cluster Computers. Introduction Cluster computing –Standard PCs or workstations connected by a fast network –Good price/performance ratio –Exploit existing.
Top Level View of Computer Function and Interconnection.
CCNA 3 Week 4 Switching Concepts. Copyright © 2005 University of Bolton Introduction Lan design has moved away from using shared media, hubs and repeaters.
CSE 661 PAPER PRESENTATION
Computers Operating System Essentials. Operating Systems PROGRAM HARDWARE OPERATING SYSTEM.
ECE 259 / CPS 221 Advanced Computer Architecture II (Parallel Computer Architecture) Machine Organizations Copyright 2004 Daniel J. Sorin Duke University.
I/O Computer Organization II 1 Interconnecting Components Need interconnections between – CPU, memory, I/O controllers Bus: shared communication channel.
Cisco 3 - Switching Perrine. J Page 16/4/2016 Chapter 4 Switches The performance of shared-medium Ethernet is affected by several factors: data frame broadcast.
EEE440 Computer Architecture
CS252 Graduate Computer Architecture Lecture 17 Multiprocessor Networks (con’t) March 18 th, 2012 John Kubiatowicz Electrical Engineering and Computer.
Graduate Computer Architecture I Lecture 11: Distribute Memory Multiprocessors.
Chapter 13 – I/O Systems (Pgs ). Devices  Two conflicting properties A. Growing uniformity in interfaces (both h/w and s/w): e.g., USB, TWAIN.
Chapter 6 Storage and Other I/O Topics. Chapter 6 — Storage and Other I/O Topics — 2 Introduction I/O devices can be characterized by Behaviour: input,
Caltech CS184 Spring DeHon 1 CS184b: Computer Architecture (Abstractions and Optimizations) Day 14: May 7, 2003 Fast Messaging.
3/12/2013Computer Engg, IIT(BHU)1 PARALLEL COMPUTERS- 2.
CCNA3 Module 4 Brierley Module 4. CCNA3 Module 4 Brierley Topics LAN congestion and its effect on network performance Advantages of LAN segmentation in.
CS252 Graduate Computer Architecture Lecture 24 Network Interface Design Memory Consistency Models Prof John D. Kubiatowicz
Spring EE 437 Lillevik 437s06-l22 University of Portland School of Engineering Advanced Computer Architecture Lecture 22 Distributed computer Interconnection.
Chapter 3 System Buses.  Hardwired systems are inflexible  General purpose hardware can do different tasks, given correct control signals  Instead.
Cluster Computers. Introduction Cluster computing –Standard PCs or workstations connected by a fast network –Good price/performance ratio –Exploit existing.
1 Device Controller I/O units typically consist of A mechanical component: the device itself An electronic component: the device controller or adapter.
CDA-5155 Computer Architecture Principles Fall 2000 Multiprocessor Architectures.
Architectures of Digital Information Systems Part 1: Interrupts and DMA dr.ir. A.C. Verschueren Eindhoven University of Technology Section of Digital.
John Kubiatowicz Electrical Engineering and Computer Sciences
Packet Switching Outline Store-and-Forward Switches
CMSC 611: Advanced Computer Architecture
Architecture of Parallel Computers CSC / ECE 506 Summer 2006 Scalable Programming Models Lecture 11 6/19/2006 Dr Steve Hunter.
Computer Science Division
Peng Liu Lecture 14 I/O Peng Liu
Chapter 13: I/O Systems I/O Hardware Application I/O Interface
Chapter 13: I/O Systems.
Presentation transcript:

Scalable Distributed Memory Multiprocessors Todd C. Mowry CS 495 October 24 & 29, 2002

CS 495 F’02– 2 – Outline Scalability physical, bandwidth, latency and cost level of integration Realizing Programming Models network transactions protocols safety –input buffer problem: N-1 –fetch deadlock Communication Architecture Design Space how much hardware interpretation of the network transaction?

CS 495 F’02– 3 – Limited Scaling of a Bus Bus: each level of the system design is grounded in the scaling limits at the layers below and assumptions of close coupling between components CharacteristicBus Physical Length~ 1 ft Number of Connectionsfixed Maximum Bandwidthfixed Interface to Comm. mediummemory inf Global Orderarbitration ProtectionVirt -> physical Trusttotal OSsingle comm. abstractionHW

CS 495 F’02– 4 – Workstations in a LAN? No clear limit to physical scaling, little trust, no global order, consensus difficult to achieve. Independent failure and restart CharacteristicBusLAN Physical Length~ 1 ftKM Number of Connectionsfixedmany Maximum Bandwidthfixed??? Interface to Comm. mediummemory infperipheral Global Orderarbitration??? ProtectionVirt -> physicalOS Trusttotalnone OSsingleindependent comm. abstractionHWSW

CS 495 F’02– 5 – Scalable Machines What are the design trade-offs for the spectrum of machines between? specialize or commodity nodes? capability of node-to-network interface supporting programming models? What does scalability mean? avoids inherent design limits on resources bandwidth increases with P latency does not cost increases slowly with P

CS 495 F’02– 6 – Bandwidth Scalability What fundamentally limits bandwidth? single set of wires Must have many independent wires Connect modules through switches Bus vs Network Switch?

CS 495 F’02– 7 – Dancehall MP Organization Network bandwidth? Bandwidth demand? independent processes? communicating processes? Latency?

CS 495 F’02– 8 – Generic Distributed Memory Org. Network bandwidth? Bandwidth demand? independent processes? communicating processes? Latency?

CS 495 F’02– 9 – Key Property Large # of independent communication paths between nodes allow a large # of concurrent transactions using different wires Initiated independently No global arbitration Effect of a transaction only visible to the nodes involved effects propagated through additional transactions

CS 495 F’02– 10 – Latency Scaling T(n) = Overhead + Channel Time + Routing Delay Overhead? Channel Time(n) = n/B Bandwidth at bottleneck RoutingDelay(h,n)

CS 495 F’02– 11 – Typical Example max distance: log n number of switches:  n log n overhead = 1 us, BW = 64 MB/s, 200 ns per hop Pipelined T 64 (128) = 1.0 us us + 6 hops * 0.2 us/hop = 4.2 us T 1024 (128) = 1.0 us us + 10 hops * 0.2 us/hop = 5.0 us Store and Forward T sf 64 (128) = 1.0 us + 6 hops * ( ) us/hop = 14.2 us T sf 1024 (128) = 1.0 us + 10 hops * ( ) us/hop = 23 us

CS 495 F’02– 12 – Cost Scaling cost(p,m) = fixed cost + incremental cost (p,m) Bus Based SMP? Ratio of processors : memory : network : I/O ? Parallel efficiency(p) = Speedup(P) / P Costup(p) = Cost(p) / Cost(1) Cost-effective: speedup(p) > costup(p)

CS 495 F’02– 13 – Physical Scaling Chip-level integration Board-level integration System-level integration

CS 495 F’02– 14 – nCUBE/2 Machine Organization Entire machine synchronous at 40 MHz Single-chip node Basic module Hypercube network configuration DRAM interface D M A c h a n n e l s R o u t e r MMU I-Fetch & decode 64-bit integer IEEE floating point Operand $ Execution unit 1024 Nodes

CS 495 F’02– 15 – CM-5 Machine Organization Board-level integration ctrl DRAM ctrl Vector unit FPUData networks Control network $ ctrl $ SRAM NI

CS 495 F’02– 16 – System Level Integration IBM SP-2

CS 495 F’02– 17 – Outline Scalability physical, bandwidth, latency and cost level of integration Realizing Programming Models network transactions protocols safety –input buffer problem: N-1 –fetch deadlock Communication Architecture Design Space how much hardware interpretation of the network transaction?

CS 495 F’02– 18 – Programming Models Realized by Protocols CAD MultiprogrammingShared address Message passing Data parallel DatabaseScientific modeling Parallel applications Programming models Communication abstraction User/system boundary Compilation or library Operating systems support Communication hardware Physical communication medium Hardware/software boundary Network Transactions

CS 495 F’02– 19 – Network Transaction Primitive one-way transfer of information from a source output buffer to a dest. input buffer causes some action at the destination occurrence is not directly visible at source deposit data, state change, reply

CS 495 F’02– 20 – Bus Transactions vs Net Transactions Issues: protection checkV->P?? formatwiresflexible output bufferingreg, FIFO?? media arbitrationgloballocal destination naming and routing input bufferinglimitedmany source action completion detection

CS 495 F’02– 21 – Shared Address Space Abstraction Fundamentally a two-way request/response protocol writes have an acknowledgement Issues fixed or variable length (bulk) transfers remote virtual or physical address, where is action performed? deadlock avoidance and input buffer full coherent? consistent?

CS 495 F’02– 22 – The Fetch Deadlock Problem Even if a node cannot issue a request, it must sink network transactions. Incoming transaction may be a request, which will generate a response. Closed system (finite buffering)

CS 495 F’02– 23 – Consistency write-atomicity violated without caching

CS 495 F’02– 24 – Key Properties of SAS Abstraction Source and destination data addresses are specified by the source of the request a degree of logical coupling and trust No storage logically “outside the application address space(s)” –may employ temporary buffers for transport Operations are fundamentally request-response Remote operation can be performed on remote memory logically does not require intervention of the remote processor

CS 495 F’02– 25 – Message Passing Bulk transfers Complex synchronization semantics more complex protocols more complex action Synchronous Send completes after matching recv and source data sent Receive completes after data transfer complete from matching send Asynchronous Send completes after send buffer may be reused

CS 495 F’02– 26 – Synchronous Message Passing Constrained programming model. Deterministic! What happens when threads added? Destination contention very limited. User/System boundary? Processor Action?

CS 495 F’02– 27 – Asynch. Message Passing: Optimistic More powerful programming model Wildcard receive => non-deterministic Storage required within msg layer?

CS 495 F’02– 28 – Asynch. Msg Passing: Conservative Where is the buffering? Contention control? Receiver initiated protocol? Short message optimizations

CS 495 F’02– 29 – Key Features of Msg Passing Abstraction Source knows send data address, dest. knows receive data address after handshake they both know both Arbitrary storage “outside the local address spaces” may post many sends before any receives non-blocking asynchronous sends reduces the requirement to an arbitrary number of descriptors –fine print says these are limited too Fundamentally a 3-phase transaction includes a request / response can use optimisitic 1-phase in limited “safe” cases –credit scheme

CS 495 F’02– 30 – Active Messages User-level analog of network transaction transfer data packet and invoke handler to extract it from the network and integrate with on-going computation Request/Reply Event notification: interrupts, polling, events? May also perform memory-to-memory transfer Request handler Reply

CS 495 F’02– 31 – Common Challenges Input buffer overflow N-1 queue over-commitment => must slow sources Reserve space per source(credit) –when available for reuse? »Ack or Higher level Refuse input when full –backpressure in reliable network –tree saturation –deadlock free –what happens to traffic not bound for congested dest? Reserve ack back channel Drop packets Utilize higher-level semantics of programming model

CS 495 F’02– 32 – Challenges (cont) Fetch Deadlock For network to remain deadlock free, nodes must continue accepting messages, even when cannot source msgs what if incoming transaction is a request? –Each may generate a response, which cannot be sent! –What happens when internal buffering is full? Logically independent request/reply networks physical networks virtual channels with separate input/output queues Bound requests and reserve input buffer space K(P-1) requests + K responses per node service discipline to avoid fetch deadlock? NACK on input buffer full NACK delivery?

CS 495 F’02– 33 – Challenges in Realizing Programming Models in the Large One-way transfer of information No global knowledge, nor global control barriers, scans, reduce, global-OR give fuzzy global state Very large number of concurrent transactions Management of input buffer resources many sources can issue a request and over-commit destination before any see the effect Latency is large enough that you are tempted to “take risks” optimistic protocols large transfers dynamic allocation Many many more degrees of freedom in design and engineering of these system

CS 495 F’02– 34 – Summary Scalability physical, bandwidth, latency and cost level of integration Realizing Programming Models network transactions protocols safety –input buffer problem: N-1 –fetch deadlock Communication Architecture Design Space how much hardware interpretation of the network transaction?

CS 495 F’02– 35 – Network Transaction Processing Key Design Issues: How much interpretation of the message? How much dedicated processing in the Comm. Assist? PM CA PM ° ° ° Scalable Network Node Architecture Communication Assist Message Output Processing – checks – translation – formating – scheduling Input Processing – checks – translation – buffering – action

CS 495 F’02– 36 – Spectrum of Designs None: Physical bit stream blind, physical DMAnCUBE, iPSC,... User/System User-level portCM-5, *T User-level handlerJ-Machine, Monsoon,... Remote virtual address Processing, translationParagon, Meiko CS-2 Global physical address Proc + Memory controllerRP3, BBN, T3D Cache-to-cache Cache controllerDash, KSR, Flash Increasing HW Support, Specialization, Intrusiveness, Performance (???)

CS 495 F’02– 37 – Net Transactions: Physical DMA DMA controlled by regs, generates interrupts Physical => OS initiates transfers Send-side construct system “envelope” around user data in kernel area Receive must receive into system buffer, since no interpretation inCA senderauth dest addr

CS 495 F’02– 38 – nCUBE Network Interface independent DMA channel per link direction leave input buffers always open segmented messages routing interprets envelope dimension-order routing on hypercube bit-serial with 36 bit cut-through Os16 ins 260 cy13 us Or18200 cy15 us - includes interrupt

CS 495 F’02– 39 – Conventional LAN Network Interface NIC Controller DMA addr len trncv TX RX Addr Len Status Next Addr Len Status Next Addr Len Status Next Addr Len Status Next Addr Len Status Next Addr Len Status Next Data Host Memory NIC IO Bus mem bus Proc

CS 495 F’02– 40 – User Level Ports initiate transaction at user level deliver to user without OS intervention network port in user space User/system flag in envelope protection check, translation, routing, media access in src CA user/sys check in dest CA, interrupt on system

CS 495 F’02– 41 – User Level Network ports Appears to user as logical message queues plus status What happens if no user pop?

CS 495 F’02– 42 – Example: CM-5 Input and output FIFO for each network 2 data networks tag per message index NI mapping table context switching? *T integrated NI on chip iWARP also ctrl DRAM ctrl Vector unit FPUData networks Control network $ ctrl $ SRAM NI Os50 cy1.5 us Or53 cy1.6 us interrupt10us

CS 495 F’02– 43 – User Level Handlers Hardware support to vector to address specified in message message ports in registers User/system P Mem DestDataAddress P Mem 

CS 495 F’02– 44 – J-Machine Each node a small msg driven processor HW support to queue msgs and dispatch to msg handler task

CS 495 F’02– 45 – *T

CS 495 F’02– 46 – iWARP Nodes integrate communication with computation on systolic basis Msg data direct to register Stream into memory Interface unit Host

CS 495 F’02– 47 – Dedicated Message Processing Without Specialized Hardware Design General Purpose processor performs arbitrary output processing (at system level) General Purpose processor interprets incoming network transactions (at system level) User Processor Msg Processor via shared memory Msg Processor Msg Processor via system network transaction Network ° ° ° dest Mem P M P NI UserSystem Mem P M P NI UserSystem

CS 495 F’02– 48 – Levels of Network Transaction User Processor stores cmd / msg / data into shared output queue must still check for output queue full (or make elastic) Communication assists make transaction happen checking, translation, scheduling, transport, interpretation Effect observed on destination address space and/or events Protocol divided between two layers Network ° ° ° dest Mem P M P NI UserSystem Mem P M P NI

CS 495 F’02– 49 – Example: Intel Paragon Network ° ° ° Mem P M P NI i860xp 50 MHz 16 KB $ 4-way 32B Block MESI sDMA rDMA MB/s $$ MB/s Duplex I/O Nodes rte MP handler Var data EOP I/O Nodes Service Devices 2048 B

CS 495 F’02– 50 – User Level Abstraction Any user process can post a transaction for any other in protection domain communication layer moves OQ src –> IQ dest may involve indirection: VAS src –> VAS dest Proc OQ IQ VAS Proc OQ IQ VAS Proc OQ IQ VAS Proc OQ IQ VAS

CS 495 F’02– 51 – Msg Processor Events Dispatcher User Output Queues Send FIFO ~Empty Rcv FIFO ~Full Send DMA Rcv DMA DMA done Compute Processor Kernel System Event

CS 495 F’02– 52 – Basic Implementation Costs: Scalar Cache-to-cache transfer (two 32B lines, quad word ops) producer: read(miss,S), chk, write(S,WT), write(I,WT),write(S,WT) consumer: read(miss,S), chk, read(H), read(miss,S), read(H),write(S,WT) to NI FIFO: read status, chk, write,... from NI FIFO: read status, chk, dispatch, read, read,... CP User OQ MP Registers Cache Net FIFO User IQ MPCP Net µs5.4 µs 10.5 µs 7 wds ns + H*40ns

CS 495 F’02– 53 – Virtual DMA -> Virtual DMA Send MP segments into 8K pages and does VA –> PA Recv MP reassembles, does dispatch and VA –> PA per page CP User OQ MP Registers Cache Net FIFO User IQ MP CP Net wds 222 Memory sDMA hdr rDMA MP MB/s 175 MB/s 400 MB/s

CS 495 F’02– 54 – Single Page Transfer Rate Actual Buffer Size: 2048 Effective Buffer Size: 3232

CS 495 F’02– 55 – Msg Processor Assessment Concurrency Intensive Need to keep inbound flows moving while outbound flows stalled Large transfers segmented Reduces overhead but adds latency User Output Queues Send FIFO ~Empty Rcv FIFO ~Full Send DMA Rcv DMA DMA done Compute Processor Kernel System Event User Input Queues VAS Dispatcher