Changeset 62c5a55 for doc/theses/thierry_delisle_PhD/thesis/text
- Timestamp:
- Aug 5, 2022, 4:20:16 PM (2 years ago)
- Branches:
- ADT, ast-experimental, master, pthread-emulation
- Children:
- 181036c
- Parents:
- 8040286 (diff), 878be178 (diff)
Note: this is a merge changeset, the changes displayed below correspond to the merge itself.
Use the(diff)
links above to see all the changes relative to each parent. - Location:
- doc/theses/thierry_delisle_PhD/thesis/text
- Files:
-
- 4 edited
Legend:
- Unmodified
- Added
- Removed
-
doc/theses/thierry_delisle_PhD/thesis/text/front.tex
r8040286 r62c5a55 129 129 which begs of the question of how many kernel threads are needed and should the number be dynamically reevaluated. 130 130 Furthermore, scheduling must prevent kernel threads from blocking, otherwise user-thread parallelism drops. 131 When user-threading parallelism does drop, how and when should idle kernel-threads be put to sleep to avoid wast edCPU resources.131 When user-threading parallelism does drop, how and when should idle kernel-threads be put to sleep to avoid wasting CPU resources. 132 132 Finally, the scheduling system must provide fairness to prevent a user thread from monopolizing a kernel thread; 133 133 otherwise other user threads can experience short/long term starvation or kernel threads can deadlock waiting for events to occur on busy kernel threads. … … 138 138 Fairness is handled through preemption and/or ad-hoc solutions, which leads to coarse-grained fairness with some pathological cases. 139 139 140 After testing and selecting specific approaches to these scheduling issues, a complete implementation was created and tested in the \CFA (C-for-all) runtime system.140 After examining, selecting and testing specific approaches to these scheduling issues, a complete implementation was created and tested in the \CFA (C-for-all) runtime system. 141 141 \CFA is a modern extension of C using user-level threading as its fundamental threading model. 142 142 As one of its primary goals, \CFA aims to offer increased safety and productivity without sacrificing performance. -
doc/theses/thierry_delisle_PhD/thesis/text/intro.tex
r8040286 r62c5a55 1 1 \chapter{Introduction}\label{intro} 2 \section{\CFA programming language}3 2 4 The \CFA programming language~\cite{cfa:frontpage,cfa:typesystem} extends the C programming language by adding modern safety and productivity features, while maintaining backwards compatibility. 5 Among its productivity features, \CFA supports user-level threading~\cite{Delisle21} allowing programmers to write modern concurrent and parallel programs. 6 My previous master's thesis on concurrent in \CFA focused on features and interfaces. 7 This Ph.D.\ thesis focuses on performance, introducing \glsxtrshort{api} changes only when required by performance considerations. 8 Specifically, this work concentrates on scheduling and \glsxtrshort{io}. 9 Prior to this work, the \CFA runtime used a strict \glsxtrshort{fifo} \gls{rQ} and no \glsxtrshort{io} capabilities at the user-thread level\footnote{C supports \glsxtrshort{io} capabilities at the kernel level, which means blocking operations block kernel threads where blocking user-level threads whould be more appropriate for \CFA.}. 3 \Gls{uthrding} (M:N) is gaining popularity over kernel-level threading (1:1) in many programming languages. 4 The user threading approach is often a better mechanism to express complex concurrent applications by efficiently running 10,000+ threads on multi-core systems. 5 Indeed, over-partitioning into small work-units with user threading significantly eases load bal\-ancing, while simultaneously providing advanced synchronization and mutual exclusion capabilities. 6 To manage these high levels of concurrency, the underlying runtime must efficiently schedule many user threads across a few kernel threads; 7 which begs of the question of how many kernel threads are needed and should the number be dynamically reevaluated. 8 Furthermore, scheduling must prevent kernel threads from blocking, otherwise user-thread parallelism drops. 9 When user-threading parallelism does drop, how and when should idle kernel-threads be put to sleep to avoid wasting CPU resources. 10 Finally, the scheduling system must provide fairness to prevent a user thread from monopolizing a kernel thread; 11 otherwise other user threads can experience short/long term starvation or kernel threads can deadlock waiting for events to occur on busy kernel threads. 10 12 11 As a research project, this work builds exclusively on newer versions of the Linux operating-system and gcc/clang compilers. 12 While \CFA is released, supporting older versions of Linux ($<$~Ubuntu 16.04) and gcc/clang compilers ($<$~gcc 6.0) is not a goal of this work. 13 This thesis analyses multiple scheduler systems, where each system attempts to fulfill the necessary requirements for \gls{uthrding}. 14 The predominant technique for managing high levels of concurrency is sharding the ready-queue with one queue per kernel-thread and using some form of work stealing/sharing to dynamically rebalance workload shifts. 15 Preventing kernel blocking is accomplish by transforming kernel locks and I/O operations into user-level operations that do not block the kernel thread or spin up new kernel threads to manage the blocking. 16 Fairness is handled through preemption and/or ad-hoc solutions, which leads to coarse-grained fairness with some pathological cases. 17 18 After examining, testing and selecting specific approaches to these scheduling issues, a completely new scheduler was created and tested in the \CFA (C-for-all) user-threading runtime-system. 19 The goal of the new scheduler to offer increased safety and productivity without sacrificing performance. 20 The quality of the new scheduler is demonstrated by comparing it with other user-threading work-stealing schedulers with the aim of showing equivalent or better performance while offering better fairness. 21 22 Chapter~\ref{intro} defines scheduling and its general goals. 23 Chapter~\ref{existing} discusses how scheduler implementations attempt to achieve these goals, but all implementations optimize some workloads better than others. 24 Chapter~\ref{s:CFARuntime} presents the relevant aspects of the \CFA runtime system that have a significant affect on the new scheduler design and implementation. 25 Chapter~\ref{core} analyses different scheduler approaches, while looking for scheduler mechanisms that provide both performance and fairness. 26 Chapter~\ref{s:UserLevelIO} covers the complex mechanisms that must be used to achieve nonblocking I/O to prevent the blocking of \glspl{kthrd}. 27 Chapters~\ref{microbench} and~\ref{macrobench} present micro and macro benchmarks used to evaluate and compare the new scheduler with similar schedulers. 28 13 29 14 30 \section{Scheduling} 15 Computer systems share multiple resources across many threads of execution, even on single 16 On a computer system with multiple processors and work units , there exists the problem of mapping work ontoprocessors in an efficient manner, called \newterm{scheduling}.17 These systems are normally \newterm{open}, meaning new work arrives from an external source or isspawned from an existing work unit.18 On a computer system, the scheduler takes a sequence of work requests in the form of threads and attempts to complete the work, subject to performance objectives, such as resource utilization.19 A general-purpose dynamic-scheduler for an open system cannot anticipate future work requests, so its performance is rarely optimal.20 With complete knowledge of arrive order and work, creating an optimal solution still effectively needs solving the bin packing problem\cite{wiki:binpak}.21 However, optimal solutions are often not required.22 Schedulers do produce excellent solutions, whitout needing optimality, by taking advantage of regularities in work patterns.31 Computer systems share multiple resources across many threads of execution, even on single-user computers like laptops or smartphones. 32 On a computer system with multiple processors and work units (routines, coroutines, threads, programs, \etc), there exists the problem of mapping many different kinds of work units onto many different kinds of processors in an efficient manner, called \newterm{scheduling}. 33 Scheduling systems are normally \newterm{open}, meaning new work arrives from an external source or is randomly spawned from an existing work unit. 34 In general, work units without threads, like routines and coroutines, are self-scheduling, while work units with threads, like tasks and programs, are scheduled. 35 For scheduled work-units, a scheduler takes a sequence of threads and attempts to run them to completion, subject to shared resource restrictions and utilization. 36 A general-purpose dynamic-scheduler for an open system cannot anticipate (future) work requests, so its performance is rarely optimal. 37 Even with complete knowledge of arrive order and work, creating an optimal solution is a bin packing problem~\cite{wiki:binpak}. 38 However, optimal solutions are often not required: schedulers often produce excellent solutions, without needing optimality, by taking advantage of regularities in work patterns. 23 39 24 40 Scheduling occurs at discreet points when there are transitions in a system. … … 27 43 \input{executionStates.pstex_t} 28 44 \end{center} 29 These \newterm{state transition}s are initiated in response to events (\Index{interrupt}s):45 These \newterm{state transition}s are initiated in response to events, \eg blocking, interrupts, errors: 30 46 \begin{itemize} 31 47 \item 32 48 entering the system (new $\rightarrow$ ready) 49 \item 50 scheduler assigns a thread to a computing resource, \eg CPU (ready $\rightarrow$ running) 33 51 \item 34 52 timer alarm for preemption (running $\rightarrow$ ready) … … 36 54 long term delay versus spinning (running $\rightarrow$ blocked) 37 55 \item 38 blocking ends, \ienetwork or I/O completion (blocked $\rightarrow$ ready)56 completion of delay, \eg network or I/O completion (blocked $\rightarrow$ ready) 39 57 \item 40 normal completion or error, \ie segment fault (running $\rightarrow$ halted) 41 \item 42 scheduler assigns a thread to a resource (ready $\rightarrow$ running) 58 normal completion or error, \eg segment fault (running $\rightarrow$ halted) 43 59 \end{itemize} 44 Key to scheduling is that a thread cannot bypass the ``ready'' state during a transition so the scheduler maintains complete control of the system .60 Key to scheduling is that a thread cannot bypass the ``ready'' state during a transition so the scheduler maintains complete control of the system, \ie no self-scheduling among threads. 45 61 46 62 When the workload exceeds the capacity of the processors, \ie work cannot be executed immediately, it is placed on a queue for subsequent service, called a \newterm{ready queue}. 47 63 Ready queues organize threads for scheduling, which indirectly organizes the work to be performed. 48 The structure of ready queues can take many different forms. 49 Where simple examples include single-queue multi-server (SQMS) and the multi-queue multi-server (MQMS). 64 The structure of ready queues can take many different forms, where the basic two are the single-queue multi-server (SQMS) and the multi-queue multi-server (MQMS). 50 65 \begin{center} 51 66 \begin{tabular}{l|l} … … 57 72 Beyond these two schedulers are a host of options, \ie adding an optional global, shared queue to MQMS. 58 73 59 The three major optimization criteria for a scheduler are:74 Once there are multiple resources and ready queues, a scheduler is faced with three major optimization criteria: 60 75 \begin{enumerate}[leftmargin=*] 61 76 \item … … 70 85 Essentially, all multi-processor computers have non-uniform memory access (NUMA), with one or more quantized steps to access data at different levels in the memory hierarchy. 71 86 When a system has a large number of independently executing threads, affinity becomes difficult because of \newterm{thread churn}. 72 That is, threads must be scheduled on multipleprocessors to obtain high processors utilization because the number of threads $\ggg$ processors.87 That is, threads must be scheduled on different processors to obtain high processors utilization because the number of threads $\ggg$ processors. 73 88 74 89 \item 75 \newterm{contention}: safe access of shared objects by multiple processors requires mutual exclusion in some form, generally locking \footnote{76 Lock-free data-structures do not involve locking but incur rsimilar costs to achieve mutual exclusion.}90 \newterm{contention}: safe access of shared objects by multiple processors requires mutual exclusion in some form, generally locking.\footnote{ 91 Lock-free data-structures do not involve locking but incur similar costs to achieve mutual exclusion.} 77 92 78 93 \noindent 79 Mutual exclusion cost and latency increases significantly with the number of processors access ing a shared object.94 Mutual exclusion cost and latency increases significantly with the number of processors access\-ing a shared object. 80 95 \end{enumerate} 81 96 82 Nevertheless, schedulers are a series of compromises, occasionally with some static or dynamic tuning parameters to enhance specific patterns. 83 Scheduling is a zero-sum game as computer processors normally have a fixed, maximum number of cycles per unit time\footnote{Frequency scaling and turbot boost add a degree of complexity that can be ignored in this discussion without loss of generality.}. 84 SQMS has perfect load-balancing but poor affinity and high contention by the processors, because of the single queue. 85 MQMS has poor load-balancing but perfect affinity and no contention, because each processor has its own queue. 97 Scheduling is a zero-sum game as computer processors normally have a fixed, maximum number of cycles per unit time.\footnote{ 98 Frequency scaling and turbo-boost add a degree of complexity that can be ignored in this discussion without loss of generality.} 99 Hence, schedulers are a series of compromises, occasionally with some static or dynamic tuning parameters to enhance specific workload patterns. 100 For example, SQMS has perfect load-balancing but poor affinity and high contention by the processors, because of the single queue. 101 While MQMS has poor load-balancing but perfect affinity and no contention, because each processor has its own queue. 86 102 87 Significant research effort has also looked at load sharing/stealing among queues, when a ready queue is too long or short, respectively.103 Significant research effort has looked at load balancing by stealing/sharing work units among queues: when a ready queue is too short or long, respectively, load stealing/sharing schedulers attempt to push/pull work units to/from other ready queues. 88 104 These approaches attempt to perform better load-balancing at the cost of affinity and contention. 89 Load sharing/stealing schedulers attempt to push/pull work units to/from other ready queues 105 However, \emph{all} approaches come at a cost (zero-sum game), but not all compromises are necessarily equivalent, especially across workloads. 106 Hence, some schedulers perform very well for specific workloads, while others offer acceptable performance over a wider range of workloads. 90 107 91 Note however that while any change comes at a cost, hence the zero-sum game, not all compromises are necessarily equivalent. 92 Some schedulers can perform very well only in very specific workload scenarios, others might offer acceptable performance but be applicable to a wider range of workloads. 93 Since \CFA attempts to improve the safety and productivity of C, the scheduler presented in this thesis attempts to achieve the same goals. 108 \section{\CFA programming language} 109 110 The \CFA programming language~\cite{Cforall,Moss18} extends the C programming language by adding modern safety and productivity features, while maintaining backwards compatibility. 111 Among its productivity features, \CFA supports \gls{uthrding}~\cite{Delisle21} as its fundamental threading model allowing programmers to easily write modern concurrent and parallel programs. 112 My previous master's thesis on concurrency in \CFA focused on features and interfaces~\cite{Delisle18}. 113 This Ph.D.\ thesis focuses on performance, introducing \glsxtrshort{api} changes only when required by performance considerations. 114 Specifically, this work concentrates on advanced thread and \glsxtrshort{io} scheduling. 115 Prior to this work, the \CFA runtime used a strict SQMS \gls{rQ} and provided no nonblocking \glsxtrshort{io} capabilities at the user-thread level.\footnote{ 116 C/\CC only support \glsxtrshort{io} capabilities at the kernel level, which means many \io operations block \glspl{kthrd} reducing parallelism at the user level.} 117 118 Since \CFA attempts to improve the safety and productivity of C, the new scheduler presented in this thesis attempts to achieve the same goals. 94 119 More specifically, safety and productivity for scheduling means supporting a wide range of workloads so that programmers can rely on progress guarantees (safety) and more easily achieve acceptable performance (productivity). 120 The new scheduler also includes support for implicit nonblocking \io, allowing applications to have more user-threads blocking on \io operations than there are \glspl{kthrd}. 121 To complete the scheduler, an idle sleep mechanism is implemented that significantly reduces wasted CPU cycles, which are then available outside of the application. 95 122 123 As a research project, this work builds exclusively on newer versions of the Linux operating-system and gcc/clang compilers. 124 The new scheduler implementation uses several optimizations to successfully balance the cost of fairness against performance; 125 some of these optimizations rely on interesting hardware optimizations only present on modern CPUs. 126 The \io implementation is based on the @io_uring@ kernel-interface, a recent addition to the Linux kernel, because it purports to handle nonblocking \emph{file} and network \io. 127 This decision allowed an interesting performance and fairness comparison with other threading systems using @select@, @epoll@, \etc. 128 While the current \CFA release supports older versions of Linux ($\ge$~Ubuntu 16.04) and gcc/clang compilers ($\ge$~gcc 6.0), it is not the purpose of this project to find workarounds in these older systems to provide backwards compatibility. 129 The hope is that these new features will soon become mainstream features. 96 130 97 131 \section{Contributions}\label{s:Contributions} 98 This work provides the following contributions in the area of user-level scheduling in an advanced programming-language runtime-system:132 This work provides the following scheduling contributions for advanced \gls{uthrding} runtime-systems: 99 133 \begin{enumerate}[leftmargin=*] 100 134 \item 101 135 A scalable scheduling algorithm that offers progress guarantees. 102 136 \item 137 Support for user-level \glsxtrshort{io} capabilities based on Linux's @io_uring@. 138 \item 103 139 An algorithm for load-balancing and idle sleep of processors, including NUMA awareness. 104 140 \item 105 Support for user-level \glsxtrshort{io} capabilities based on Linux's @io_uring@. 141 A mechanism for adding fairness on top of work-stealing through helping (used both for the I/O and ready-queue). 142 \item 143 An optimization of the work-stealing helping-mechanism for load balancing to reduce scheduling costs. 144 \item 145 An optimization for the alternative relaxed-list for load balancing to reduce scheduling costs in embarrassingly parallel cases. 106 146 \end{enumerate} -
doc/theses/thierry_delisle_PhD/thesis/text/io.tex
r8040286 r62c5a55 1 \chapter{User Level \io} 1 \chapter{User Level \io}\label{s:UserLevelIO} 2 2 As mentioned in Section~\ref{prev:io}, user-level \io requires multiplexing the \io operations of many \glspl{thrd} onto fewer \glspl{proc} using asynchronous \io operations. 3 3 Different operating systems offer various forms of asynchronous operations and, as mentioned in Chapter~\ref{intro}, this work is exclusively focused on the Linux operating-system. -
doc/theses/thierry_delisle_PhD/thesis/text/runtime.tex
r8040286 r62c5a55 1 \chapter{\CFA Runtime} 1 \chapter{\CFA Runtime}\label{s:CFARuntime} 2 2 This chapter presents an overview of the capabilities of the \CFA runtime prior to this thesis work. 3 3
Note: See TracChangeset
for help on using the changeset viewer.