- Timestamp:
- Sep 8, 2022, 5:11:19 PM (21 months ago)
- Branches:
- ADT, ast-experimental, master, pthread-emulation
- Children:
- 264f6c9
- Parents:
- e4855f6
- File:
-
- 1 edited
Legend:
- Unmodified
- Added
- Removed
-
doc/theses/thierry_delisle_PhD/thesis/text/existing.tex
re4855f6 r918e0137 8 8 A secondary aspect is how much information can be gathered versus how much information must be given as part of the scheduler input. 9 9 This information adds to the spectrum of scheduling algorithms, going from static schedulers that are well-informed from the start, to schedulers that gather most of the information needed, to schedulers that can only rely on very limited information. 10 Note, this description includes both information about each request, \eg time to complete or resources needed, and information about the relationships among request , \eg whether some requestmust be completed before another request starts.11 12 Scheduling physical resources, \eg in an assembly line, is generally amenable to using well-informed scheduling , since information can be gathered much faster than the physical resources can be assigned and workloads are likely to stay stable for long periods of time.10 Note, this description includes both information about each request, \eg time to complete or resources needed, and information about the relationships among requests, \eg whether some requests must be completed before another request starts. 11 12 Scheduling physical resources, \eg in an assembly line, is generally amenable to using well-informed scheduling since information can be gathered much faster than the physical resources can be assigned and workloads are likely to stay stable for long periods. 13 13 When a faster pace is needed and changes are much more frequent gathering information on workloads, up-front or live, can become much more limiting and more general schedulers are needed. 14 14 15 15 \section{Naming Convention} 16 Scheduling has been studied by various communities concentrating on different incarnation of the same problems.17 As a result, there are no standard naming conventions for scheduling that isrespected across these communities.16 Scheduling has been studied by various communities concentrating on different incarnations of the same problems. 17 As a result, there are no standard naming conventions for scheduling that are respected across these communities. 18 18 This document uses the term \newterm{\Gls{at}} to refer to the abstract objects being scheduled and the term \newterm{\Gls{proc}} to refer to the concrete objects executing these \ats. 19 19 20 20 \section{Static Scheduling} 21 \newterm{Static schedulers} require \ats dependencies and costs be explicitly and exhaustively specified prior to scheduling.21 \newterm{Static schedulers} require \ats dependencies and costs to be explicitly and exhaustively specified prior to scheduling. 22 22 The scheduler then processes this input ahead of time and produces a \newterm{schedule} the system follows during execution. 23 23 This approach is popular in real-time systems since the need for strong guarantees justifies the cost of determining and supplying this information. 24 In general, static schedulers are less relevant to this project because they require input from the programmers that the programming language does not have as part of its concurrency semantic .24 In general, static schedulers are less relevant to this project because they require input from the programmers that the programming language does not have as part of its concurrency semantics. 25 25 Specifying this information explicitly adds a significant burden to the programmer and reduces flexibility. 26 26 For this reason, the \CFA scheduler does not require this information. 27 27 28 28 \section{Dynamic Scheduling} 29 \newterm{Dynamic schedulers} determine \at sdependencies and costs during scheduling, if at all.30 Hence, unlike static scheduling, \at sdependencies are conditional and detected at runtime.29 \newterm{Dynamic schedulers} determine \at dependencies and costs during scheduling, if at all. 30 Hence, unlike static scheduling, \at dependencies are conditional and detected at runtime. 31 31 This detection takes the form of observing new \ats in the system and determining dependencies from their behaviour, including suspending or halting a \at that dynamically detects unfulfilled dependencies. 32 32 Furthermore, each \at has the responsibility of adding dependent \ats back into the system once dependencies are fulfilled. … … 40 40 at best, the scheduler has only some imprecise information provided by the programmer, \eg, indicating a \at takes approximately 3--7 seconds to complete, rather than exactly 5 seconds. 41 41 Providing this kind of information is a significant programmer burden, especially if the information does not scale with the number of \ats and their complexity. 42 For example, providing an exhaustive list of files read by 5 \ats is an easier requirement th en providing an exhaustive list of memory addresses accessed by 10,000 independent \ats.42 For example, providing an exhaustive list of files read by 5 \ats is an easier requirement than providing an exhaustive list of memory addresses accessed by 10,000 independent \ats. 43 43 44 44 Since the goal of this thesis is to provide a scheduler as a replacement for \CFA's existing \emph{uninformed} scheduler, explicitly informed schedulers are less relevant to this project. Nevertheless, some strategies are worth mentioning. … … 49 49 The simplest priority scheduling algorithm is to require that every \at have a distinct pre-established priority and always run the available \ats with the highest priority. 50 50 Asking programmers to provide an exhaustive set of unique priorities can be prohibitive when the system has a large number of \ats. 51 It can therefore be desirable for schedulers to support \ats with identical priorities and/or automatically set ting and adjustingpriorities for \ats.51 It can therefore be desirable for schedulers to support \ats with identical priorities and/or automatically set and adjust priorities for \ats. 52 52 Most common operating systems use some variant on priorities with overlaps and dynamic priority adjustments. 53 53 For example, Microsoft Windows uses a pair of priorities~\cite{win:priority}, one specified by users out of ten possible options and one adjusted by the system. 54 54 55 55 \subsection{Uninformed and Self-Informed Dynamic Schedulers} 56 Several scheduling algorithms do not require programmers to provide additional information on each \at, and instead make scheduling decisions based solely on internal state and/or information implicitly gathered by the scheduler.56 Several scheduling algorithms do not require programmers to provide additional information on each \at, and instead, make scheduling decisions based solely on internal state and/or information implicitly gathered by the scheduler. 57 57 58 58 … … 61 61 This design effectively moves the scheduler into the realm of \newterm{Control Theory}~\cite{wiki:controltheory}. 62 62 This information gathering does not generally involve programmers, and as such, does not increase programmer burden the same way explicitly provided information may. 63 However, some feedback schedulers do allow programmers to offer additional information on certain \ats, in orderto direct scheduling decisions.64 The important distinction beingwhether the scheduler can function without this additional information.63 However, some feedback schedulers do allow programmers to offer additional information on certain \ats, to direct scheduling decisions. 64 The important distinction is whether the scheduler can function without this additional information. 65 65 66 66 67 67 \section{Work Stealing}\label{existing:workstealing} 68 One of the most popular scheduling algorithm in practice (see~\ref{existing:prod}) is work stealing.69 This idea, introduced by \cite{DBLP:conf/fpca/BurtonS81}, effectively has each worker process its local \ats first ,but allows the possibility for other workers to steal local \ats if they run out of \ats.68 One of the most popular scheduling algorithms in practice (see~\ref{existing:prod}) is work stealing. 69 This idea, introduced by \cite{DBLP:conf/fpca/BurtonS81}, effectively has each worker process its local \ats first but allows the possibility for other workers to steal local \ats if they run out of \ats. 70 70 \cite{DBLP:conf/focs/Blumofe94} introduced the more familiar incarnation of this, where each worker has a queue of \ats and workers without \ats steal \ats from random workers\footnote{The Burton and Sleep algorithm has trees of \ats and steals only among neighbours.}. 71 Blumofe and Leiserson also prove worst 71 Blumofe and Leiserson also prove worst-case space and time requirements for well-structured computations. 72 72 73 73 Many variations of this algorithm have been proposed over the years~\cite{DBLP:journals/ijpp/YangH18}, both optimizations of existing implementations and approaches that account for new metrics. … … 86 86 Another example is energy usage, where the scheduler is modified to optimize for energy efficiency in addition/instead of performance~\cite{ribic2014energy,torng2016asymmetry}. 87 87 88 \paragraph{Complex Machine Architecture} Another aspect that has been examined is how well work stealing is applicableto different machine architectures.89 This is arguably strongly related to Task Placement ,but extends into more heterogeneous architectures.88 \paragraph{Complex Machine Architecture} Another aspect that has been examined is how applicable work stealing is to different machine architectures. 89 This is arguably strongly related to Task Placement but extends into more heterogeneous architectures. 90 90 As \CFA offers no particular support for heterogeneous architecture, this is also an area that is less relevant to this thesis. 91 91 Although it could be an interesting avenue for future work. … … 94 94 There is also a large body of research on the theoretical aspects of work stealing. These evaluate, for example, the cost of \glslink{atmig}{migration}~\cite{DBLP:conf/sigmetrics/SquillanteN91,DBLP:journals/pe/EagerLZ86}, how affinity affects performance~\cite{DBLP:journals/tpds/SquillanteL93,DBLP:journals/mst/AcarBB02,DBLP:journals/ipl/SuksompongLS16} and theoretical models for heterogeneous systems~\cite{DBLP:journals/jpdc/MirchandaneyTS90,DBLP:journals/mst/BenderR02,DBLP:conf/sigmetrics/GastG10}. 95 95 \cite{DBLP:journals/jacm/BlellochGM99} examines the space bounds of work stealing and \cite{DBLP:journals/siamcomp/BerenbrinkFG03} shows that for under-loaded systems, the scheduler completes its computations in finite time, \ie is \newterm{stable}. 96 Others show that work stealing is applicableto various scheduling contexts~\cite{DBLP:journals/mst/AroraBP01,DBLP:journals/anor/TchiboukdjianGT13,DBLP:conf/isaac/TchiboukdjianGTRB10,DBLP:conf/ppopp/AgrawalLS10,DBLP:conf/spaa/AgrawalFLSSU14}.96 Others show that work stealing applies to various scheduling contexts~\cite{DBLP:journals/mst/AroraBP01,DBLP:journals/anor/TchiboukdjianGT13,DBLP:conf/isaac/TchiboukdjianGTRB10,DBLP:conf/ppopp/AgrawalLS10,DBLP:conf/spaa/AgrawalFLSSU14}. 97 97 \cite{DBLP:conf/ipps/ColeR13} also studied how randomized work-stealing affects false sharing among \ats. 98 98 … … 101 101 102 102 \section{Preemption} 103 One last aspect of scheduling is preemption ,since many schedulers rely on it for some of their guarantees.103 One last aspect of scheduling is preemption since many schedulers rely on it for some of their guarantees. 104 104 Preemption is the idea of interrupting \ats that have been running too long, effectively injecting suspend points into the application. 105 There are multiple techniques to achieve this effect, but they all aim to guarantee that the suspend points in a \at sare never further apart than some fixed duration.106 While this helps schedulers guarantee that no \ats unfairly monopolize sa worker, preemption can effectively be added to any scheduler.105 There are multiple techniques to achieve this effect, but they all aim to guarantee that the suspend points in a \at are never further apart than some fixed duration. 106 While this helps schedulers guarantee that no \ats unfairly monopolize a worker, preemption can effectively be added to any scheduler. 107 107 Therefore, the only interesting aspect of preemption for the design of scheduling is whether to require it. 108 108 … … 111 111 While these schedulers do not necessarily represent the most recent advances in scheduling, they are what is generally accessible to programmers. 112 112 As such, I believe these schedulers are at least as relevant as those presented in published work. 113 Schedulers that operate in kernel space and inuser space are considered, as both can offer relevant insight for this project.113 Both Schedulers that operate in kernel space and user space are considered, as both can offer relevant insight for this project. 114 114 However, real-time schedulers are not considered, as these have constraints that are much stricter than what is needed for this project. 115 115 … … 117 117 Operating System Schedulers tend to be fairly complex as they generally support some amount of real-time, aim to balance interactive and non-interactive \ats and support multiple users sharing hardware without requiring these users to cooperate. 118 118 Here are more details on a few schedulers used in the common operating systems: Linux, FreeBSD, Microsoft Windows and Apple's OS X. 119 The information is less complete for operating systems with closed source.119 The information is less complete for closed source operating systems. 120 120 121 121 \paragraph{Linux's CFS} … … 124 124 The \at that has used the least CPU time is scheduled. 125 125 It also supports the concept of \newterm{Nice values}, which are effectively multiplicative factors on the CPU time used. 126 The ordering of \ats is also affected by a group 126 The ordering of \ats is also affected by a group-based notion of fairness, where \ats belonging to groups having used less CPU time are preferred to \ats belonging to groups having used more CPU time. 127 127 Linux achieves load-balancing by regularly monitoring the system state~\cite{MAN:linux/cfs/balancing} and using some heuristic on the \gls{load}, currently CPU time used in the last millisecond plus a decayed version of the previous time slots~\cite{MAN:linux/cfs/pelt}. 128 128 129 129 \cite{DBLP:conf/eurosys/LoziLFGQF16} shows that Linux's CFS also does work stealing to balance the workload of each \proc, but the paper argues this aspect can be improved significantly. 130 The issues highlighted stem from Linux's need to support fairness across \ats \emph{and} across users\footnote{Enforcing fairness across users means that given two users, one with a single \at and the other with one thousand \ats, the user with a single \at does not receive one 130 The issues highlighted stem from Linux's need to support fairness across \ats \emph{and} across users\footnote{Enforcing fairness across users means that given two users, one with a single \at and the other with one thousand \ats, the user with a single \at does not receive one-thousandth of the CPU time.}, increasing the complexity. 131 131 132 132 Linux also offers a FIFO scheduler, a real-time scheduler, which runs the highest-priority \ats, and a round-robin scheduler, which is an extension of the FIFO-scheduler that adds fixed time slices. \cite{MAN:linux/sched} … … 135 135 The ULE scheduler used in FreeBSD\cite{DBLP:conf/bsdcon/Roberson03} is a feedback scheduler similar to Linux's CFS. 136 136 It uses different data structures and heuristics but also schedules according to some combination of CPU time used and niceness values. 137 It also periodically balances the load of the system (according to a different heuristic) ,but uses a simpler work stealing approach.137 It also periodically balances the load of the system (according to a different heuristic) but uses a simpler work stealing approach. 138 138 139 139 \paragraph{Windows(OS)} … … 143 143 The scheduler may also temporarily adjust priorities after certain effects like the completion of I/O requests. 144 144 145 In~\cite{russinovich2009windows}, Chapter 1 section 2.3 ``Processes, Threads, and Jobs'' discusses the scheduling policy more in 146 Multicore scheduling is based on a combination of priorities and preferred \proc.145 In~\cite{russinovich2009windows}, Chapter 1 section 2.3 ``Processes, Threads, and Jobs'' discusses the scheduling policy more in-depth. 146 Multicore scheduling is based on a combination of priorities and \proc preferance. 147 147 Each \at is assigned an initial processor using a round-robin policy, called the \at's \newterm{ideal} \proc. 148 148 \Glspl{at} are distributed among the \procs according to their priority, preferring to match \ats to their ideal \proc and then to the last \proc they ran on. 149 This approach is a variation of work stealing, where the stealing \proc restore the \at to its original \proc after running it, but mixed with priorities.149 This approach is a variation of work stealing, where the stealing \proc restores the \at to its original \proc after running it, but mixed with priorities. 150 150 151 151 \paragraph{Apple OS X} … … 160 160 161 161 There is very little documentation on the internals of this scheduler. 162 However, the documentation do describe a feature set that is very similar to the Windows and Linux OS schedulers.162 However, the documentation does describe a feature set that is very similar to the Windows and Linux OS schedulers. 163 163 Presumably, this means that the internals are also fairly similar overall. 164 164 165 165 \subsection{User-Level Schedulers} 166 By comparison, user level schedulers tend to be simpler, gatheringfewer metrics and avoid complex notions of fairness. Part of the simplicity is due to the fact that all \ats have the same user, and therefore cooperation is both feasible and probable.166 By comparison, user-level schedulers tend to be simpler, gather fewer metrics and avoid complex notions of fairness. Part of the simplicity is due to the fact that all \ats have the same user, and therefore cooperation is both feasible and probable. 167 167 168 168 \paragraph{Go}\label{GoSafePoint} 169 169 Go's scheduler uses a randomized work-stealing algorithm that has a global run-queue (\emph{GRQ}) and each processor (\emph{P}) has both a fixed-size run-queue (\emph{LRQ}) and a high-priority next ``chair'' holding a single element~\cite{GITHUB:go,YTUBE:go}. 170 Preemption is present, but only at safe -points,~\cite{go:safepoints} which are inserted detection codeat various frequent access boundaries.170 Preemption is present, but only at safe points,~\cite{go:safepoints} which are detection code inserted at various frequent access boundaries. 171 171 172 172 The algorithm is as follows : … … 184 184 Erlang is a functional language that supports concurrency in the form of processes: threads that share no data. 185 185 It uses a kind of round-robin scheduler, with a mix of work sharing and stealing to achieve load balancing~\cite{:erlang}, where under-loaded workers steal from other workers, but overloaded workers also push work to other workers. 186 This \glslink{atmig}{migration} logic is directed by monitoring logic that evaluates the load a few times per second s.186 This \glslink{atmig}{migration} logic is directed by monitoring logic that evaluates the load a few times per second. 187 187 188 188 \paragraph{Intel\textregistered ~Threading Building Blocks} … … 196 196 \item The successor of t if \textit{t} was its last completed predecessor. 197 197 \item A task popped from the end of the thread's own deque. 198 \item A task with a ffinity for the thread.198 \item A task with an affinity for the thread. 199 199 \item A task popped from approximately the beginning of the shared queue. 200 200 \item A task popped from the beginning of another randomly chosen thread's deque. … … 217 217 While the documentation only gives limited insight into the scheduling and load balancing approach, \cite{apple:gcd2} suggests a fairly classic approach. 218 218 Each \proc has a queue of \ats to run, called \newterm{blocks}, which are drained in \glsxtrshort{fifo}. 219 GCD also has secondary queues, called \newterm{Dispatch Queues} with clear ordering, where executing a block ends -up scheduling more blocks.219 GCD also has secondary queues, called \newterm{Dispatch Queues} with clear ordering, where executing a block ends up scheduling more blocks. 220 220 In terms of semantics, these Dispatch Queues seem to be very similar to Intel\textregistered ~TBB \lstinline{execute()} and predecessor semantics. 221 221 … … 223 223 224 224 \paragraph{LibFibre} 225 LibFibre~\cite{DBLP:journals/pomacs/KarstenB20} is a light -weight user-level threading framework developed at the University of Waterloo.226 Similarly to Go, it uses a variation of work stealing with a global queue that ishigher priority than stealing.225 LibFibre~\cite{DBLP:journals/pomacs/KarstenB20} is a lightweight user-level threading framework developed at the University of Waterloo. 226 Similarly to Go, it uses a variation of work stealing with a global queue that has a higher priority than stealing. 227 227 Unlike Go, it does not have the high-priority next ``chair'' and does not use randomized work-stealing.
Note: See TracChangeset
for help on using the changeset viewer.