1. (C) 2003 Mulitfacet ProjectUniversity of Wisconsin-Madison Simulating a $2M Commercial Server on a $2K PC Alaa Alameldeen, Milo Martin, Carl Mauer, Kevin Moore, Min Xu, Daniel Sorin, Mark D. Hill, & David A. Wood Multifacet Project (www.cs.wisc.edu/multifacet) Computer Sciences Department University of Wisconsin—Madison January 2003

2. Wisconsin Multifacet Project Methods 2 Dangers of a Sabbatical Mark on Sabbatical at Universidad Politecnica de Catalunya (UPC) Mark Normally

3. Wisconsin Multifacet Project Methods 3 Context –Commercial server design is important –Multifacet project seeks improved designs –Must evaluate alternatives Commercial Servers –Processors, memory, disks  $2M –Run large multithreaded transaction-oriented workloads –Use commercial applications on commercial OS To Simulate on $2K PC –Scale & tune workloads –Manage simulation complexity –Cope with workload variability Summary Keep L2 miss rates, etc. Separate timing & function Use randomness & statistics

4. Wisconsin Multifacet Project Methods 4 Outline Context –Commercial Servers –Multifacet Project Workload & Simulation Methods Separate Timing & Functional Simulation Cope with Workload Variability Summary

5. Wisconsin Multifacet Project Methods 5 Why Commercial Servers? Many (Academic) Architects –Desktop computing –Wireless appliances We focus on servers – (Important Market) – Performance Challenges – Robustness Challenges – Methodological Challenges

6. Wisconsin Multifacet Project Methods 6 3-Tier Internet Service PCs w/ “soft” state Servers running databases for “hard” state Servers running applications for “business” rules LAN / SAN LAN / SAN Multifacet Focus

7. Wisconsin Multifacet Project Methods 7 Multifacet: Commercial Server Design Wisconsin Multifacet Project –Directed by Mark D. Hill & David A. Wood –Sponsors: NSF, WI, Compaq, IBM, Intel, & Sun –Current Contributors: Alaa Alameldeen, Brad Beckman, Pacia Harper, Milo Martin, Carl Mauer, Kevin Moore, Daniel Sorin, & Min Xu –Past Contributors: Anastassia Ailamaki, Ender Bilir, Ross Dickson, Ying Hu, Manoj Plakal, & Anne Condon Analysis –Want 4-64 processors –Many cache-to-cache misses –Neither snooping nor directories ideal Multifacet Designs –Snooping w/ multicast [ISCA99] or unordered network [ASPLOS00] – Bandwidth-adaptive [HPCA02] & token coherence [ISCA03]

8. Wisconsin Multifacet Project Methods 8 Outline Context Workload & Simulation Methods –Select, scale, & tune workloads –Transition workload to simulator –Specify & test the proposed design –Evaluate design with simple/detailed processor models Separate Timing & Functional Simulation Cope with Workload Variability Summary

9. Wisconsin Multifacet Project Methods 9 Multifacet Simulation Overview Virtutech Simics (www.virtutech.com)www.virtutech.com Rest is Multifacet software Full System Functional Simulator (Simics) Pseudo-Random Protocol Checker Memory Timing Simulator (Ruby) Processor Timing Simulator (Opal) Commercial Server (Sun E6000) Scaled WorkloadsFull Workloads Memory Protocol Generator (SLICC) Timing SimulatorProtocol Development Workload Development

10. Wisconsin Multifacet Project Methods 10 Select Important Workloads Online Transaction Processing: DB2 w/ TPC-C-like Java Server Workload: SPECjbb Static web content serving: Apache Dynamic web content serving: Slashcode Java-based Middleware: (soon) Full Workloads

11. Wisconsin Multifacet Project Methods 11 Setup & Tune Workloads (on real hardware) Tune workload, OS parameters Measure transaction rate, speed-up, miss rates, I/O Compare to published results Commercial Server (Sun E6000) Full Workloads

12. Wisconsin Multifacet Project Methods 12 Scale & Re-tune Workloads Scale-down for PC memory limits Retaining similar behavior (e.g., L2 cache miss rate) Re-tune to achieve higher transaction rates (OLTP: raw disk, multiple disks, more users, etc.) Commercial Server (Sun E6000) Scaled Workloads

13. Wisconsin Multifacet Project Methods 13 Transition Workloads to Simulation Create disk dumps of tuned workloads In simulator: Boot OS, start, & warm application Create Simics checkpoint (snapshot) Full System Functional Simulator (Simics) Scaled Workloads

14. Wisconsin Multifacet Project Methods 14 Specify Proposed Computer Design Coherence Protocol (control tables: states X events) Cache Hierarchy (parameters & queues) Interconnect (switches & queues) Processor (later) Memory Timing Simulator (Ruby) Memory Protocol Generator (SLICC)

15. Wisconsin Multifacet Project Methods 15 Test Proposed Computer Design Randomly select write action & later read check Massive false-sharing for interaction Perverse network stresses design Transient error & deadlock detection Sound but not complete Memory Timing Simulator (Ruby) Pseudo-Random Protocol Checker

16. Wisconsin Multifacet Project Methods 16 Simulate with Simple Blocking Processor Warm-up caches or sometimes sufficient (SafetyNet) Run for fixed number of transactions –Some transaction partially done at start –Other transactions partially done at end Cope with workload variability (later) Full System Functional Simulator (Simics) Memory Timing Simulator (Ruby) Scaled Workloads

17. Wisconsin Multifacet Project Methods 17 Simulate with Detailed Processor Accurate (future) timing & (current) function Simulation complexity decoupled (discussed soon) Same transaction methodology & work variability issues Full System Functional Simulator (Simics) Memory Timing Simulator (Ruby) Processor Timing Simulator (Opal) Scaled Workloads

18. Wisconsin Multifacet Project Methods 18 Simulation Infrastructure & Workload Process Select important workloads: run, tune, scale, & re-tune Specify system & pseudo-randomly test Create warm workload checkpoint Simulate with simple or detailed processor Fixed #transactions, manage simulation complexity (next), cope with workload variability (next next) Full System Functional Simulator (Simics) Memory Timing Simulator (Ruby) Processor Timing Simulator (Opal) Commercial Server (Sun E6000) Scaled WorkloadsFull Workloads Pseudo-Random Protocol Checker Memory Protocol Generator (SLICC)

19. Wisconsin Multifacet Project Methods 19 Outline Context Simulation Infrastructure & Workload Process Separate Timing & Functional Simulation –Simulation Challenges –Managing Simulation Complexity –Timing-First Simulation –Evaluation Cope with Workload Variability Summary

20. Wisconsin Multifacet Project Methods 20 Challenges to Timing Simulation Execution driven simulation is getting harder Micro-architecture complexity –Multiple “in-flight” instructions –Speculative execution –Out-of-order execution Thread-level parallelism –Hardware Multi-threading –Traditional Multi-processing

21. Wisconsin Multifacet Project Methods 21 Challenges to Functional Simulation Commercial workloads have high functional fidelity demands (Simulated) Target System Target Application Database Operating System SPEC Benchmarks Kernels Web Server Application complexity RAM Processor PCI Bus Ethernet Controller Fiber Channel Controller Graphics Card SCSI Controller CD- ROM SCSI Disk … DMA Controller Terminal I/O MMU Controller IRQ Controller Status Registers Serial PortMMU Real Time Clock

22. Wisconsin Multifacet Project Methods 22 Managing Simulator Complexity Functional Simulator Timing Simulator Functional-First (Trace-driven) - Timing feedback + Timing feedback - Tight Coupling - Performance? Timing and Functional Simulator Integrated (SimOS) - Complex Timing-Directed Functional Simulator Timing Simulator  Complete Timing  No? Function  No Timing  Complete Function Timing-First (Multifacet) Functional Simulator Timing Simulator  Complete Timing  Partial Function  No Timing  Complete Function + Timing feedback + Using existing simulators + Software development advantages

23. Wisconsin Multifacet Project Methods 23 Timing-First Simulation Timing Simulator –does functional execution of user and privileged operations –does speculative, out-of-order multiprocessor timing simulation –does NOT implement functionality of full instruction set or any devices Functional Simulator –does full-system multiprocessor simulation –does NOT model detailed micro-architectural timing Timing Simulator Functional Simulator Commit Verify CPU System RAM Network add load Cache CPU Execute

24. Wisconsin Multifacet Project Methods 24 Timing-First Operation As instruction retires, step CPU in functional simulator Verify instruction’s execution Reload state if timing simulator deviates from functional –Loads in multi-processors –Instructions with unidentified side-effects –NOT loads/store to I/O devices Timing Simulator Functional Simulator CPU System RAM Network add load Cache CPU Execute Commit Reload Verify

25. Wisconsin Multifacet Project Methods 25 Benefits of Timing-First Supports speculative multi-processor timing models Leverages existing simulators Software development advantages –Increases flexibility and reduces code complexity –Immediate, precise check on timing simulator However: –How much performance error is introduced in this approach? –Are there simulation performance penalties?

26. Wisconsin Multifacet Project Methods 26 Evaluation Our implementation, TFsim uses: –Functional Simulator: Virtutech Simics –Timing simulator: Implemented less than one-person year Evaluated using OS intensive commercial workloads –OS Boot: > 1 billion instructions of Solaris 8 startup –OLTP: TPC-C-like benchmark using a 1 GB database –Dynamic Web: Apache serving message board, using code and data similar to slashdot.org –Static Web: Apache web server serving static web pages –Barnes-Hut: Scientific SPLASH-2 benchmark

27. Wisconsin Multifacet Project Methods 27 Measured Deviations Less than 20 deviations per 100,000 instructions (0.02%)

28. Wisconsin Multifacet Project Methods 28 If the Timing Simulator Modeled Fewer Events

29. Wisconsin Multifacet Project Methods 29 Sensitivity Results

30. Wisconsin Multifacet Project Methods 30 Analysis of Results Runs full-system workloads! Timing performance impact of deviations –Worst case: less than 3% performance error ‘Overhead’ of redundant execution –18% on average for uniprocessors –18% (2 processors) up to 36% (16 processors) Total Execution Time Timing Simulator Functional Simulator

31. Wisconsin Multifacet Project Methods 31 Performance Comparison Absolute simulation performance comparison –In kilo-instructions committed per second (KIPS) –RSIM Scaled: 107 KIPS –Uniprocessor TFsim: 119 KIPS (Simulated) Target System Target Application Host Computer Out-of-Order MP SPARC V9 SPLASH-2 Kernels 400 MHz SPARC running Solaris Out-of-Order MP Full-system SPARC V9 SPLASH-2 Kernels 1.2 GHz Pentium running Linux match close different TFsimRSIM

32. Wisconsin Multifacet Project Methods 32 Bundled Retires

33. Wisconsin Multifacet Project Methods 33 Timing-First Conclusions Execution-driven simulators are increasingly complex How to manage complexity? Our answer: –Introduces relatively little performance error (worst case: 3%) –Has low-overhead (18% uniprocessor average) –Rapid development time Timing-First Simulation Functional Simulator Timing Simulator  Complete Timing  Partial Function  No Timing  Complete Function

34. Wisconsin Multifacet Project Methods 34 Outline Context Workload Process & Infrastructure Separate Timing & Functional Simulation Cope with Workload Variability –Variability in Multithreaded Workloads –Coping in Simulation –Examples & Statistics Summary

35. Wisconsin Multifacet Project Methods 35 What is Happening Here? OLTP

36. Wisconsin Multifacet Project Methods 36 What is Happening Here? How can slower memory lead to faster workload? Answer: Multithreaded workload takes different path –Different lock race outcomes –Different scheduling decisions (1) Does this happen for real hardware? (2) If so, what should we do about it?

37. Wisconsin Multifacet Project Methods 37 One Second Intervals (on real hardware) OLTP

38. Wisconsin Multifacet Project Methods 38 60 Second Intervals (on real hardware) 16-day simulation OLTP

39. Wisconsin Multifacet Project Methods 39 Coping with Workload Variability Running (simulating) long enough not appealing Need to separate coincidental & real effects Standard statistics on real hardware –Variation within base system runs vs. variation between base & enhanced system runs –But deterministic simulation has no “within” variation Solution with deterministic simulation –Add pseudo-random delay on L2 misses –Simulate base (enhanced) system many times –Use simple or complex statistics

40. Wisconsin Multifacet Project Methods 40 Coincidental (Space) Variability

41. Wisconsin Multifacet Project Methods 41 Wrong Conclusion Ratio WCR (16,32) = 18% WCR (16,64) = 7.5% WCR (32,64) = 26%

42. Wisconsin Multifacet Project Methods 42 More Generally: Use Standard Statistics As one would for a measurement of a “live” system Confidence Intervals –95% confidence intervals contain true value 95% of the time –Non-overlapping confidence intervals give statistically significant conclusions Use ANOVA or Hypothesis Testing – even better!

43. Wisconsin Multifacet Project Methods 43 Confidence Interval Example Estimate #runs to get non-overlapping confidence intervals ROB

44. Wisconsin Multifacet Project Methods 44 Also Time Variability (on real hardware) Therefore, select checkpoint(s) carefully

45. Wisconsin Multifacet Project Methods 45 Workload Variability Summary Variability is a real phenomenon for multi-threaded workloads –Runs from same initial conditions are different Variability is a challenge for simulations –Simulations are short –Wrong conclusions may be drawn Our solution accounts for variability –Multiple runs, confidence intervals –Reduces wrong conclusion probability

46. Wisconsin Multifacet Project Methods 46 Talk Summary Simulations of $2M Commercial Servers must –Complete in reasonable time (on $2K PCs) –Handle OS, devices, & multithreaded hardware –Cope with variability of multithreaded software Multifacet –Scale & tune transactional workloads –Separate timing & functional simulation –Cope w/ workload variability via randomness & statistics References ( www.cs.wisc.edu/multifacet/papers ) –Simulating a $2M Commercial Server on a $2K PC [Computer03] –Full-System Timing-First Simulation [Sigmetrics02] –Variability in Architectural Simulations … [HPCA03]

47. Wisconsin Multifacet Project Methods 47 Other Multifacet Methods Work Specifying & Verifying Coherence Protocols –[SPAA98], [HPCA99], [SPAA99], & [TPDS02] Workload Analysis & Improvement –Database systems [VLDB99] & [VLDB01] –Pointer-based [PLDI99] & [Computer00] –Middleware [HPCA03] Modeling & Simulation –Commercial workloads [Computer02] & [HPCA03] –Decoupling timing/functional simulation [Sigmetrics02] –Simulation generation [PLDI01] –Analytic modeling [Sigmetrics00] & [TPDS TBA] –Micro-architectural slack [ISCA02]

48. Wisconsin Multifacet Project Methods 48 Backup Slides

49. Wisconsin Multifacet Project Methods 49 One Ongoing/Future Methods Direction Middleware Applications –Memory system behavior of Java Middleware [HPCA 03] –Machine measurements –Full-system simulation Future Work: Multi-Machine Simulation –Isolate middle-tier from client emulators and database Understand fundamental workload behaviors –Drives future system design

50. Wisconsin Multifacet Project Methods 50 ECPerf vs. SpecJBB Different cache-to-cache transfer ratios!

51. Wisconsin Multifacet Project Methods 51 Online Transaction Processing (OLTP) DB2 with a TPC-C-like workload. The TPC-C benchmark is widely used to evaluate system performance for the on-line transaction processing market. The benchmark itself is a specification that describes the schema, scaling rules, transaction types and transaction mix, but not the exact implementation of the database. TPC-C transactions are of five transaction types, all related to an order-processing environment. Performance is measured by the number of “New Order” transactions performed per minute (tpmC). Our OLTP workload is based on the TPC-C v3.0 benchmark. We use IBM’s DB2 V7.2 EEE database management system and an IBM benchmark kit to build the database and emulate users. We build an 800 MB 4000-warehouse database on five raw disks and an additional dedicated database log disk. We scaled down the sizes of each warehouse by maintaining the reduced ratios of 3 sales districts per warehouse, 30 customers per district, and 100 items per warehouse (compared to 10, 30,000 and 100,000 required by the TPC-C specification). Each user randomly executes transactions according to the TPC-C transaction mix specifications, and we set the think and keying times for users to zero. A different database thread is started for each user. We measure all completed transactions, even those that do not satisfy timing constraints of the TPC-C benchmark specification.

52. Wisconsin Multifacet Project Methods 52 Java Server Workload (SPECjbb) Java-based middleware applications are increasingly used in modern e- business settings. SPECjbb is a Java benchmark emulating a 3-tier system with emphasis on the middle tier server business logic. SPECjbb runs in a single Java Virtual Machine (JVM) in which threads represent terminals in a warehouse. Each thread independently generates random input (tier 1 emulation) before calling transaction- specific business logic. The business logic operates on the data held in binary trees of java objects (tier 3 emulation). The specification states that the benchmark does no disk or network I/O. We used Sun’s HotSpot 1.4.0 Server JVM and Solaris’s native thread implementation. The benchmark includes driver threads to generate transactions. We set the system heap size to 1.8 GB and the new object heap size to 256 MB to reduce the frequency of garbage collection. Our experiments used 24 warehouses, with a data size of approximately 500 MB.

53. Wisconsin Multifacet Project Methods 53 Static Web Content Serving: Apache Web servers such as Apache represent an important enterprise server application. Apache is a popular open-source web server used in many internet/intranet settings. In this benchmark, we focus on static web content serving. We use Apache 2.0.39 for SPARC/Solaris 8 configured to use pthread locks and minimal logging at the web server. We use the Scalable URL Request Generator (SURGE) as the client. SURGE generates a sequence of static URL requests which exhibit representative distributions for document popularity, document sizes, request sizes, temporal and spatial locality, and embedded document count. We use a repository of 20,000 files (totalling ~500 MB), and use clients with zero think time. We compiled both Apache and Surge using Sun’s WorkShop C 6.1 with aggressive optimization.

54. Wisconsin Multifacet Project Methods 54 Dynamic Web Content Serving: Slashcode Dynamic web content serving has become increasingly important for web sites that serve large amount of information. Dynamic content is used by online stores, instant news, and community message board systems. Slashcode is an open-source dynamic web message posting system used by the popular slashdot.org message board system. We used Slashcode 2.0, Apache 1.3.20, and Apache’s mod_perl module 1.25 (with perl 5.6) on the server side. We used MySQL 3.23.39 as the database engine. The server content is a snapshot from the slashcode.com site, containing approximately 3000 messages with a total size of 5 MB. Most of the run time is spent on dynamic web page generation. We use a multi-threaded user emulation program to emulate user browsing and posting behavior. Each user independently and randomly generates browsing and posting requests to the server according to a transaction mix specification. We compiled both server and client programs using Sun’s WorkShop C 6.1 with aggressive optimization.