Changes in / [25a1cb0:d3aa64f1]


Ignore:
Location:
doc/theses/thierry_delisle_PhD/comp_II
Files:
3 deleted
5 edited

Legend:

Unmodified
Added
Removed

  • doc/theses/thierry_delisle_PhD/comp_II/Makefile

    r25a1cb0 rd3aa64f1  
    1212
    1313## Define the text source files.
     14
     15SOURCES = ${addsuffix .tex, \
     16comp_II \
     17}
     18
    1419FIGURES = ${addsuffix .tex, \
     20        base \
     21        empty \
    1522        emptybit \
    1623        emptytree \
    1724        emptytls \
    1825        resize \
     26        system \
    1927}
    2028
    2129PICTURES = ${addsuffix .pstex, \
    22         base \
    23         empty \
    24         system \
    2530}
    2631
    … …  
    3237
    3338## Define the documents that need to be made.
    34 all: comp_II.pdf presentation.pdf
    35 comp_II.pdf: ${FIGURES} ${PICTURES}
    36 presentation.pdf: presentationstyle.sty base.dark.pstex empty.dark.pstex system.dark.pstex
    3739
    38 DOCUMENT = comp_II.pdf presentation.pdf
     40DOCUMENT = comp_II.pdf
    3941BASE = ${basename ${DOCUMENT}}
    4042
    … …  
    5052# File Dependencies #
    5153
    52 %.pdf : build/%.ps | ${Build}
     54${DOCUMENT} : ${BASE}.ps
    5355        ps2pdf $<
    5456
    55 build/%.ps : build/%.dvi | ${Build}
    56         dvips $< -o $@
     57${BASE}.ps : ${BASE}.dvi
     58        dvips ${Build}/$< -o $@
    5759
    58 build/%.dvi : %.tex Makefile | ${Build}
     60${BASE}.dvi : Makefile ${GRAPHS} ${PROGRAMS} ${PICTURES} ${FIGURES} ${SOURCES} \
     61                ${Macros}/common.tex ${Macros}/indexstyle ../../../bibliography/pl.bib \
     62                local.bib glossary.tex | ${Build}
    5963        # Must have *.aux file containing citations for bibtex
    60         if [ ! -r ${basename $@}.aux ] ; then ${LaTeX} $< ; fi
    61         -${BibTeX} ${basename $@}
     64        if [ ! -r ${basename $@}.aux ] ; then ${LaTeX} ${basename $@}.tex ; fi
     65        -${BibTeX} ${Build}/${basename $@}
    6266        # Some citations reference others so run again to resolve these citations
    63         ${LaTeX} $<
    64         -${BibTeX} ${basename $@}
     67        ${LaTeX} ${basename $@}.tex
     68        -${BibTeX} ${Build}/${basename $@}
    6569        # Make index from *.aux entries and input index at end of document
    66         -makeglossaries -q -s ${basename $@}.ist ${basename $@}
     70        makeglossaries -q -s ${Build}/${basename $@}.ist ${Build}/${basename $@}
    6771        # Run again to finish citations
    68         ${LaTeX} $<
     72        ${LaTeX} ${basename $@}.tex
    6973
    7074## Define the default recipes.
    … …  
    7377        mkdir -p ${Build}
    7478
    75 %.tex : img/%.fig | ${Build}
     79%.tex : img/%.fig ${Build}
    7680        fig2dev -L eepic $< > ${Build}/$@
    7781
    … …  
    8387        fig2dev -L pstex_t -p ${Build}/$@ $< > ${Build}/$@_t
    8488
    85 ## pstex with inverted colors
    86 %.dark.pstex : img/%.fig Makefile | ${Build}
    87         fig2dev -L pstex $< > ${Build}/$@
    88         sed -i 's/\/col-1 {0 setgray} bind def/\/col-1 {1 setgray} bind def/g' ${Build}/$@
    89         sed -i 's/\/col0 {0.000 0.000 0.000 srgb} bind def/\/col0 {1.000 1.000 1.000 srgb} bind def/g' ${Build}/$@
    90         sed -i 's/\/col7 {1.000 1.000 1.000 srgb} bind def/\/col7 {0.000 0.000 0.000 srgb} bind def/g' ${Build}/$@
    91         fig2dev -L pstex_t -p ${Build}/$@ $< > ${Build}/$@_t
    92 
    9389# Local Variables: #
    9490# compile-command: "make" #

  • doc/theses/thierry_delisle_PhD/comp_II/comp_II.tex

    r25a1cb0 rd3aa64f1  
    6363It aims to add high-productivity features while maintaining the predictable performance of C.
    6464As such, concurrency in \CFA\cite{Delisle19} aims to offer simple and safe high-level tools while still allowing performant code.
    65 \CFA concurrent code is written in the synchronous programming paradigm but uses \glspl{uthrd} in order to achieve the simplicity and maintainability of synchronous programming without sacrificing the efficiency of asynchronous programming.
     65\CFA concurrent code is written in the synchronous programming paradigm but uses \glspl{uthrd} in order to achieve the simplicity and maintainability of synchronous programming without sacrificing the efficiency of asynchronous programing.
    6666As such, the \CFA \newterm{scheduler} is a preemptive user-level scheduler that maps \glspl{uthrd} onto \glspl{kthrd}.
    6767
    … …  
    7575and the cost of scheduling, \ie deciding which thread to run next among all the threads ready to run.
    7676\end{enumerate}
    77 The first cost is generally constant and fixed\footnote{Affecting the constant context-switch cost is whether it is done in one step, after the scheduling, or in two steps, context-switching to a third fixed thread before scheduling.}, while the scheduling cost can vary based on the system state.
    78 Adding multiple \glspl{kthrd} does not fundamentally change the scheduler semantics or requirements, it simply adds new correctness requirements, \ie \newterm{linearizability}\footnote{Meaning, however fast the CPU threads run, there is an equivalent sequential order that gives the same result.}, and a new dimension to performance: scalability, where scheduling cost now also depends on contention.
     77The first cost is generally constant and fixed\footnote{Affecting the constant context-switch cost is whether it is done in one step, after the scheduling, or in two steps, context-switching to a fixed third-thread before scheduling.}, while the scheduling cost can vary based on the system state.
     78Adding multiple \glspl{kthrd} does not fundamentally change the scheduler semantics or requirements, it simply adds new correctness requirements, \ie \newterm{linearizability}\footnote{Meaning however fast the CPU threads run, there is an equivalent sequential order that gives the same result.}, and a new dimension to performance: scalability, where scheduling cost now also depends on contention.
     79
    7980The more threads switch, the more the administration cost of scheduling becomes noticeable.
    8081It is therefore important to build a scheduler with the lowest possible cost and latency.
    … …  
    8384While the illusion of simultaneity is easier to reason about, it can break down if the scheduler allows too much unfairness.
    8485Therefore, the scheduler should offer as much fairness as needed to guarantee eventual progress, but use unfairness to help performance.
    85 In practice, threads must wait in turn but there can be advantages to unfair scheduling, similar to the express cash register at a grocery store.
     86In practice, threads must wait in turn but there can be advantages to unfair scheduling, similar to the the express cash-register at a grocery store.
    8687
    8788The goal of this research is to produce a scheduler that is simple for programmers to understand and offers good performance.
    … …  
    9293\end{quote}
    9394
    94 For a general-purpose scheduler, it is impossible to produce an optimal algorithm as it would require knowledge of the future behaviour of threads.
    95 As such, scheduling performance is generally either defined by the best-case scenario, \ie a workload to which the scheduler is tailored, or the worst-case scenario, \ie the scheduler behaves no worse than \emph{X}.
     95For a general purpose scheduler, it is impossible to produce an optimal algorithm as it would require knowledge of the future behaviour of threads.
     96As such, scheduling performance is generally either defined by the best case scenario, \ie a workload to which the scheduler is tailored, or the worst case scenario, \ie the scheduler behaves no worst than \emph{X}.
    9697For this proposal, the performance is evaluated using the second approach to allow \CFA programmers to rely on scheduling performance.
    9798Because there is no optimal scheduler, ultimately \CFA may allow programmers to write their own scheduler; but that is not the subject of this proposal, which considers only the default scheduler.
    … …  
    100101To achieve the \CFA scheduling goal includes:
    101102\begin{enumerate}
    102         \item producing a scheduling strategy with sufficient fairness guarantees,
    103         \item creating an abstraction layer over the operating system to handle kernel-threads spinning unnecessarily,
    104         \item scheduling blocking I/O operations,
    105         \item and writing sufficient library tools to allow developers to indirectly use the scheduler, either through tuning knobs or replacing the default scheduler.
     103\item
     104producing a scheduling strategy with sufficient fairness guarantees,
     105\item
     106creating an abstraction layer over the operating system to handle kernel-threads spinning unnecessarily,
     107\item
     108scheduling blocking I/O operations,
     109\item
     110and writing sufficient library tools to allow developers to indirectly use the scheduler, either through tuning knobs or replacing the default scheduler.
    106111\end{enumerate}
    107112
    … …  
    114119\paragraph{Correctness} As with any other concurrent data structure or algorithm, the correctness requirement is paramount.
    115120The scheduler cannot allow threads to be dropped from the ready queue, \ie scheduled but never run, or be executed multiple times when only being scheduled once.
    116 Since \CFA concurrency has no spurious wake up, this definition of correctness also means the scheduler should have no spurious wake up.
     121Since \CFA concurrency has no spurious wakeup, this definition of correctness also means the scheduler should have no spurious wakeup.
    117122The \CFA scheduler must be correct.
    118123
    … …  
    125130The \CFA scheduler should offer good performance for all three metrics.
    126131
    127 \paragraph{Fairness} Like performance, this requirement has several aspects : eventual progress, predictability and performance reliability.
     132\paragraph{Fairness} Like performance, this requirement has several aspect : eventual progress, predictability and performance reliability.
    128133\newterm{Eventual progress} guarantees every scheduled thread is eventually run, \ie prevent starvation.
    129 As a hard requirement, the \CFA scheduler must guarantee eventual progress, otherwise the above-mentioned illusion of simultaneous execution is broken and the scheduler becomes much more complex to reason about.
    130 \newterm{Predictability} and \newterm{reliability} mean similar workloads achieve similar performance and programmer execution intuition is respected.
    131 For example, a thread that yields aggressively should not run more often than other tasks.
     134As a hard requirement, the \CFA scheduler must guarantee eventual progress, otherwise the above mentioned illusion of simultaneous execution is broken and the scheduler becomes much more complex to reason about.
     135\newterm{Predictability} and \newterm{reliability} means similar workloads achieve similar performance and programmer execution intuition is respected.
     136For example, a thread that yields aggressively should not run more often then other tasks.
    132137While this is intuitive, it does not hold true for many work-stealing or feedback based schedulers.
    133138The \CFA scheduler must guarantee eventual progress and should be predictable and offer reliable performance.
    … …  
    145150
    146151\begin{enumerate}
    147         \item Threads live long enough for useful feedback information to be gathered.
     152        \item Threads live long enough for useful feedback information to be to gathered.
    148153        \item Threads belong to multiple users so fairness across threads is insufficient.
    149154\end{enumerate}
    … …  
    152157Since \CFA has the explicit goal of allowing many smaller threads, this can naturally lead to threads with much shorter lifetimes that are only scheduled a few times.
    153158Scheduling strategies based on feedback cannot be effective in these cases because there is no opportunity to measure the metrics that underlie the algorithm.
    154 Note, the problem of \newterm{feedback convergence} (reacting too slowly to scheduling events) is not specific to short-lived threads but can also occur with threads that show drastic changes in scheduling, \eg threads running for long periods of time and then suddenly blocking and unblocking quickly and repeatedly.
     159Note, the problem of \newterm{feedback convergence} (reacting too slowly to scheduling events) is not specific to short lived threads but can also occur with threads that show drastic changes in scheduling, \eg threads running for long periods of time and then suddenly blocking and unblocking quickly and repeatedly.
    155160
    156161In the context of operating systems, these concerns can be overshadowed by a more pressing concern : security.
    … …  
    159164In the case of the \CFA scheduler, every thread runs in the same user space and is controlled by the same user.
    160165Fairness across users is therefore a given and it is then possible to safely ignore the possibility that threads are malevolent.
    161 This approach allows for a much simpler fairness metric and in this proposal \emph{fairness} is defined as: when multiple threads are cycling through the system, the total ordering of threads being scheduled, \ie pushed onto the ready queue, should not differ much from the total ordering of threads being executed, \ie popped from the ready queue.
     166This approach allows for a much simpler fairness metric and in this proposal \emph{fairness} is defined as: when multiple threads are cycling through the system, the total ordering of threads being scheduled, \ie pushed onto the ready-queue, should not differ much from the total ordering of threads being executed, \ie popped from the ready-queue.
    162167
    163168Since feedback is not necessarily feasible within the lifetime of all threads and a simple fairness metric can be used, the scheduling strategy proposed for the \CFA runtime does not use per-threads feedback.
    … …  
    167172Another broad category of schedulers are priority schedulers.
    168173In these scheduling strategies, threads have priorities and the runtime schedules the threads with the highest priority before scheduling other threads.
    169 Threads with equal priority are scheduled using a secondary strategy, often something simple like round robin or FIFO.
     174Threads with equal priority are scheduled using a secondary strategy, often something simple like round-robin or FIFO.
    170175A consequence of priority is that, as long as there is a thread with a higher priority that desires to run, a thread with a lower priority does not run.
    171176This possible starving of threads can dramatically increase programming complexity since starving threads and priority inversion (prioritizing a lower priority thread) can both lead to serious problems.
    172177
    173178An important observation is that threads do not need to have explicit priorities for problems to occur.
    174 Indeed, any system with multiple ready queues that attempts to exhaust one queue before accessing the other queues, essentially provide implicit priority, which can encounter starvation problems.
     179Indeed, any system with multiple ready-queues that attempts to exhaust one queue before accessing the other queues, essentially provide implicit priority, which can encounter starvation problems.
    175180For example, a popular scheduling strategy that suffers from implicit priorities is work stealing.
    176181\newterm{Work stealing} is generally presented as follows:
    … …  
    186191
    187192\subsection{Schedulers without feedback or priorities}
    188 This proposal conjectures that it is possible to construct a default scheduler for the \CFA runtime that offers good scalability and a simple fairness guarantee that is easy for programmers to reason about.
     193This proposal conjectures that is is possible to construct a default scheduler for the \CFA runtime that offers good scalability and a simple fairness guarantee that is easy for programmers to reason about.
    189194The simplest fairness guarantee is FIFO ordering, \ie threads scheduled first run first.
    190195However, enforcing FIFO ordering generally conflicts with scalability across multiple processors because of the additional synchronization.
    … …  
    212217Pushing new data is done by selecting one of these underlying queues at random, recording a timestamp for the operation and pushing to the selected queue.
    213218Popping is done by selecting two queues at random and popping from the queue with the oldest timestamp.
    214 A higher number of underlying queues lead to less contention on each queue and therefore better performance.
     219A higher number of underlying queues leads to less contention on each queue and therefore better performance.
    215220In a loaded system, it is highly likely the queues are non-empty, \ie several tasks are on each of the underlying queues.
    216221This means that selecting a queue at random to pop from is highly likely to yield a queue with available items.
    … …  
    219224\begin{figure}
    220225        \begin{center}
    221                 \input{base.pstex_t}
     226                \input{base}
    222227        \end{center}
    223228        \caption{Relaxed FIFO list at the base of the scheduler: an array of strictly FIFO lists.
    224         The timestamp is in all nodes and cell arrays.}
     229The timestamp is in all nodes and cell arrays.}
    225230        \label{fig:base}
    226231\end{figure}
    … …  
    228233\begin{figure}
    229234        \begin{center}
    230                 \input{empty.pstex_t}
     235                \input{empty}
    231236        \end{center}
    232237        \caption{``More empty'' state of the queue: the array contains many empty cells.}
    … …  
    240245Overall performance is therefore influenced by the contention on the underlying queues and pop performance is influenced by the item density.
    241246
    242 This leads to four performance cases for the centralized ready queue, as depicted in Table~\ref{tab:perfcases}.
     247This leads to four performance cases for the centralized ready-queue, as depicted in Table~\ref{tab:perfcases}.
    243248The number of processors (many or few) refers to the number of kernel threads \emph{actively} attempting to pop user threads from the queues, not the total number of kernel threads.
    244249The number of threads (many or few) refers to the number of user threads ready to be run.
    … …  
    316321Sparse global information helps high-contention cases but increases latency in zero-contention-cases, to read and ``aggregate'' the information\footnote{Hierarchical structures, \eg binary search tree, effectively aggregate information but follow pointer chains, learning information at each node.
    317322Similarly, other sparse schemes need to read multiple cachelines to acquire all the information needed.}.
    318 Finally, dense local information has both the advantages of low latency in zero-contention cases and scalability in high-contention cases. However the information can become stale making it difficult to use to ensure correctness.
    319 The fact that these solutions have these fundamental limits suggest to me a better solution that attempts to combine these properties in an interesting way.
     323Finally, dense local information has both the advantages of low latency in zero-contention cases and scalability in high-contention cases, however the information can become stale making it difficult to use to ensure correctness.
     324The fact that these solutions have these fundamental limits suggest to me a better solution that attempts to combine these properties in an interesting ways.
    320325Also, the lock discussed in Section~\ref{sec:resize} allows for solutions that adapt to the number of processors, which could also prove useful.
    321326
    … …  
    323328
    324329How much scalability is actually needed is highly debatable.
    325 \emph{libfibre}\cite{libfibre} has compared favourably to other schedulers in webserver tests\cite{Karsten20} and uses a single atomic counter in its scheduling algorithm similarly to the proposed bitmask.
     330\emph{libfibre}\cite{libfibre} has compared favorably to other schedulers in webserver tests\cite{Karsten20} and uses a single atomic counter in its scheduling algorithm similarly to the proposed bitmask.
    326331As such, the single atomic instruction on a shared cacheline may be sufficiently performant.
    327332
    328333I have built a prototype of this ready queue in the shape of a data queue, \ie nodes on the queue are structures with a single int representing a thread and intrusive data fields.
    329 Using this prototype, I ran preliminary performance experiments that confirm the expected performance in Table~\ref{tab:perfcases}.
     334Using this prototype I ran preliminary performance experiments that confirm the expected performance in Table~\ref{tab:perfcases}.
    330335However, these experiments only offer a hint at the actual performance of the scheduler since threads form more complex operations than simple integer nodes, \eg threads are not independent of each other, when a thread blocks some other thread must intervene to wake it.
    331336
    … …  
    336341\begin{figure}
    337342        \begin{center}
    338                 \input{system.pstex_t}
     343                \input{system}
    339344        \end{center}
    340345        \caption{Global structure of the \CFA runtime system.}
    … …  
    349354This assumption is made both in the design of the proposed scheduler as well as in the original design of the \CFA runtime system.
    350355As such, the proposed scheduler must honour the correctness of this behaviour but does not have any performance objectives with regard to resizing a cluster.
    351 How long adding or removing processors take and how much this disrupts the performance of other threads is considered a secondary concern since it should be amortized over long periods of times.
     356How long adding or removing processors take and how much this disrupts the performance of other threads is considered a secondary concern since it should be amortized over long period of times.
    352357However, as mentioned in Section~\ref{sec:queue}, contention on the underlying queues can have a direct impact on performance.
    353358The number of underlying queues must therefore be adjusted as the number of processors grows or shrinks.
    … …  
    367372This description effectively matches with the description of a reader-writer lock, infrequent but invasive updates among frequent read operations.
    368373In the case of the ready queue described above, read operations are operations that push or pop from the ready queue but do not invalidate any references to the ready queue data structures.
    369 Writes, on the other hand, would add or remove inner queues, invalidating references to the array of inner queues in a process.
     374Writes on the other hand would add or remove inner queues, invalidating references to the array of inner queues in a process.
    370375Therefore, the current proposed approach to this problem is to add a per-cluster reader-writer lock around the ready queue to prevent restructuring of the ready-queue data-structure while threads are being pushed or popped.
    371376
    … …  
    375380
    376381\paragraph{Objectives and Existing Work}
    377 The lock must offer scalability and performance on par with the actual ready queue in order not to introduce a new bottleneck.
     382The lock must offer scalability and performance on par with the actual ready-queue in order not to introduce a new bottleneck.
    378383I have already built a lock that fits the desired requirements and preliminary testing show scalability and performance that exceed the target.
    379384As such, I do not consider this lock to be a risk for this project.
    … …  
    391396give back unneeded CPU time associated with a process to other user processors executing on the computer,
    392397\item
    393 and reduce energy consumption in cases where more idle kernel-threads translate into idle CPUs, which can cycle down.
     398and reduce energy consumption in cases where more idle kernel-threads translate to idle CPUs, which can cycle down.
    394399\end{enumerate}
    395400Support for idle sleep broadly involves calling the operating system to block the kernel thread and handling the race between a blocking thread and the waking thread, and handling which kernel thread should sleep or wake up.
    … …  
    398403This operation is equivalent to the classic problem of missing signals when using condition variables: the ``sleepy'' processor indicates its intention to block but has not yet gone to sleep when another processor attempts to wake it up.
    399404The waking-up operation sees the blocked process and signals it, but the blocking process is racing to sleep so the signal is missed.
    400 In cases where kernel threads are managed as processors on the current cluster, losing signals is not necessarily critical, because at least some processors on the cluster are awake and may check for more processors eventually.
     405In cases where kernel threads are managed as processors on the current cluster, loosing signals is not necessarily critical, because at least some processors on the cluster are awake and may check for more processors eventually.
    401406Individual processors always finish scheduling user threads before looking for new work, which means that the last processor to go to sleep cannot miss threads scheduled from inside the cluster (if they do, that demonstrates the ready queue is not linearizable).
    402407However, this guarantee does not hold if threads are scheduled from outside the cluster, either due to an external event like timers and I/O, or due to a user (or kernel) thread migrating from a different cluster.
    … …  
    408413
    409414Another important issue is avoiding kernel threads sleeping and waking frequently because there is a significant operating-system cost.
    410 This scenario happens when a program oscillates between high and low activity, needing most and then fewer processors.
     415This scenario happens when a program oscillates between high and low activity, needing most and then less processors.
    411416A possible partial solution is to order the processors so that the one which most recently went to sleep is woken up.
    412417This allows other sleeping processors to reach deeper sleep state (when these are available) while keeping ``hot'' processors warmer.
    … …  
    415420
    416421A final important aspect of idle sleep is when should processors make the decision to sleep and when is it appropriate for sleeping processors to be woken up.
    417 Processors that are unnecessarily unblocked lead to unnecessary contention, CPU usage, and power consumption, while too many sleeping processors can lead to suboptimal throughput.
    418 Furthermore, transitions from sleeping to awake and vice versa also add unnecessary latency.
     422Processors that are unnecessarily unblocked lead to unnecessary contention, CPU usage, and power consumption, while too many sleeping processors can lead to sub-optimal throughput.
     423Furthermore, transitions from sleeping to awake and vice-versa also add unnecessary latency.
    419424There is already a wealth of research on the subject\cite{schillings1996engineering, wiki:thunderherd} and I may use an existing approach for the idle-sleep heuristic in this project, \eg\cite{Karsten20}.
    420425
    … …  
    425430It is preferable to block the user thread performing the I/O and reuse the underlying kernel-thread to run other ready user threads.
    426431This approach requires intercepting user-thread calls to I/O operations, redirecting them to an asynchronous I/O interface, and handling the multiplexing/demultiplexing between the synchronous and asynchronous API.
    427 As such, there are three components needed to implement support for asynchronous I/O:
     432As such, there are three components needed to implemented support for asynchronous I/O:
    428433\begin{enumerate}
    429434\item
    … …  
    441446It is sufficient to make one work in the complex context of the \CFA runtime.
    442447\uC uses the $select$\cite{select} as its interface, which handles ttys, pipes and sockets, but not disk.
    443 $select$ entails significant complexity and is being replaced in UNIX operating systems, which make it a less interesting alternative.
     448$select$ entails significant complexity and is being replaced in UNIX operating-systems, which make it a less interesting alternative.
    444449Another popular interface is $epoll$\cite{epoll}, which is supposed to be cheaper than $select$.
    445 However, $epoll$ also does not handle the file system and anecdotal evidence suggest it has problems with Linux pipes and $TTY$s.
     450However, $epoll$ also does not handle the file system and anectodal evidence suggest it has problem with linux pipes and $TTY$s.
    446451A popular cross-platform alternative is $libuv$\cite{libuv}, which offers asynchronous sockets and asynchronous file system operations (among other features).
    447452However, as a full-featured library it includes much more than I need and could conflict with other features of \CFA unless significant effort is made to merge them together.
    448453A very recent alternative that I am investigating is $io_uring$\cite{io_uring}.
    449 It claims to address some of the issues with $epoll$ and my early investigating suggests that the claim is accurate.
    450 $io_uring$ uses a much more general approach where system calls are registered to a queue and later executed by the kernel, rather than relying on system calls to return an error instead of blocking and subsequently waiting for changes on file descriptors.
     454It claims to address some of the issues with $epoll$ and my early investigating suggest that the claim is accurate.
     455$io_uring$ uses a much more general approach where system calls are register to a queue and later executed by the kernel, rather than relying on system calls to return an error instead of blocking and subsequently waiting for changes on file descriptors.
    451456I believe this approach allows for fewer problems, \eg the manpage for $open$\cite{open} states:
    452457\begin{quote}
    453 Note that [the $O_NONBLOCK$ flag] has no effect for regular files and block devices;
    454 that is, I/O operations will (briefly) block when device activity is required, regardless of whether $O_NONBLOCK$ is set.
    455 Since $O_NONBLOCK$ semantics might eventually be implemented, applications should not depend upon blocking behaviour when specifying this flag for regular files and block devices.
     458        Note that [the $O_NONBLOCK$ flag] has no effect for regular files and block devices;
     459        that is, I/O operations will (briefly) block when device activity is required, regardless of whether $O_NONBLOCK$ is set.
     460        Since $O_NONBLOCK$ semantics might eventually be implemented, applications should not depend upon blocking behavior when specifying this flag for regular files and block devices.
    456461\end{quote}
    457462This makes approach based on $epoll$/$select$ less reliable since they may not work for every file descriptors.
    458 For this reason, I plan to use $io_uring$ as the OS abstraction for the \CFA runtime unless further work shows problems I haven't encountered yet.
     463For this reason, I plan to use $io_uring$ as the OS abstraction for the \CFA runtime, unless further work shows problems I haven't encountered yet.
    459464However, only a small subset of the features are available in Ubuntu as of April 2020\cite{wiki:ubuntu-linux}, which will limit performance comparisons.
    460465I do not believe this will affect the comparison result.
    … …  
    485490\section{Discussion}
    486491I believe that runtime system and scheduling are still open topics.
    487 Many ``state of the art'' production frameworks still use single-threaded event loops because of performance considerations, \eg \cite{nginx-design}, and, to my knowledge, no widely available system language offers modern threading facilities.
     492Many ``state of the art'' production frameworks still use single threaded event-loops because of performance considerations, \eg \cite{nginx-design}, and, to my knowledge, no wideyl available system language offers modern threading facilities.
    488493I believe the proposed work offers a novel runtime and scheduling package, where existing work only offers fragments that users must assemble themselves when possible.
    489494
    … …  
    496501\hline November 2020 & March 2021   & Completion of the implementation. \\
    497502\hline March 2021 & April 2021  & Final Performance experiments. \\
    498 \hline May 2021 & August 2021 & Thesis writing and defence. \\
     503\hline May 2021 & August 2021 & Thesis writing and defense. \\
    499504\hline
    500505\end{tabular}

  • doc/theses/thierry_delisle_PhD/comp_II/img/base.fig

    r25a1cb0 rd3aa64f1  
    1 #FIG 3.2  Produced by xfig version 3.2.7b
     1#FIG 3.2  Produced by xfig version 3.2.5c
    22Landscape
    33Center
    44Inches
    5 Letter
     5Letter 
    66100.00
    77Single
    … …  
    13131 3 0 1 0 0 50 -1 20 0.000 1 0.0000 6975 4200 20 20 6975 4200 6995 4200
    1414-6
     156 2400 2100 3000 2700
     161 3 0 1 0 7 50 -1 -1 0.000 1 0.0000 2700 2400 300 300 2700 2400 3000 2400
     172 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
     18         2400 2475 3000 2475
     194 1 0 50 -1 0 11 0.0000 2 120 210 2700 2650 TS\001
     20-6
     216 2400 3000 3000 3600
     221 3 0 1 0 7 50 -1 -1 0.000 1 0.0000 2700 3300 300 300 2700 3300 3000 3300
     232 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
     24         2400 3375 3000 3375
     254 1 0 50 -1 0 11 0.0000 2 120 210 2700 3550 TS\001
     26-6
    15271 3 0 1 0 7 50 -1 -1 0.000 1 0.0000 3900 2400 300 300 3900 2400 4200 2400
    16281 3 0 1 0 7 50 -1 -1 0.000 1 0.0000 3900 3300 300 300 3900 3300 4200 3300
    … …  
    21331 3 0 1 0 7 50 -1 -1 0.000 1 0.0000 6300 3300 300 300 6300 3300 6600 3300
    22341 3 0 1 0 7 50 -1 -1 0.000 1 0.0000 4509 3302 300 300 4509 3302 4809 3302
    23 1 3 0 1 0 7 50 -1 -1 0.000 1 0.0000 2700 3300 300 300 2700 3300 3000 3300
    24 1 3 0 1 0 7 50 -1 -1 0.000 1 0.0000 2700 2400 300 300 2700 2400 3000 2400
    25352 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
    2636         3000 3900 3000 4500
    … …  
    6373        1 1 1.00 45.00 90.00
    6474         6300 4200 6300 3600
    65 2 1 0 1 -1 7 50 -1 -1 0.000 0 0 -1 1 0 2
     752 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 1 0 2
    6676        1 1 1.00 45.00 90.00
    6777         2700 4200 2700 3600
    … …  
    7181        1 1 1.00 45.00 90.00
    7282         4500 4200 4500 3600
    73 2 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
    74          2400 3375 3000 3375
    75 2 1 0 1 0 7 50 -1 -1 0.000 0 0 -1 0 0 2
    76          2400 2475 3000 2475
    77 4 2 -1 50 -1 0 12 0.0000 2 135 630 2100 3075 Threads\001
    78 4 2 -1 50 -1 0 12 0.0000 2 165 450 2100 2850 Ready\001
    79 4 1 -1 50 -1 0 11 0.0000 2 135 180 2700 4450 TS\001
    80 4 2 -1 50 -1 0 12 0.0000 2 165 720 2100 4200 Array of\001
    81 4 2 -1 50 -1 0 12 0.0000 2 150 540 2100 4425 Queues\001
    82 4 1 -1 50 -1 0 11 0.0000 2 135 180 2700 3550 TS\001
    83 4 1 -1 50 -1 0 11 0.0000 2 135 180 2700 2650 TS\001
     834 2 0 50 -1 0 12 0.0000 2 180 660 2100 4200 Array of\001
     844 2 0 50 -1 0 12 0.0000 2 165 600 2100 4425 Queues\001
     854 2 0 50 -1 0 12 0.0000 2 135 645 2100 3075 Threads\001
     864 2 0 50 -1 0 12 0.0000 2 180 525 2100 2850 Ready\001
     874 1 0 50 -1 0 11 0.0000 2 120 210 2700 4450 TS\001

  • doc/theses/thierry_delisle_PhD/comp_II/img/empty.fig

    r25a1cb0 rd3aa64f1  
    1 #FIG 3.2  Produced by xfig version 3.2.7b
     1#FIG 3.2  Produced by xfig version 3.2.5c
    22Landscape
    33Center
    44Inches
    5 Letter
     5Letter 
    66100.00
    77Single
    … …  
    4141        1 1 1.00 45.00 90.00
    4242         2700 4200 2700 3600
    43 4 2 -1 50 -1 0 12 0.0000 2 165 720 2100 4200 Array of\001
    44 4 2 -1 50 -1 0 12 0.0000 2 150 540 2100 4425 Queues\001
    45 4 2 -1 50 -1 0 12 0.0000 2 135 630 2100 3075 Threads\001
    46 4 2 -1 50 -1 0 12 0.0000 2 165 450 2100 2850 Ready\001
     434 2 0 50 -1 0 12 0.0000 2 180 660 2100 4200 Array of\001
     444 2 0 50 -1 0 12 0.0000 2 165 600 2100 4425 Queues\001
     454 2 0 50 -1 0 12 0.0000 2 135 645 2100 3075 Threads\001
     464 2 0 50 -1 0 12 0.0000 2 180 525 2100 2850 Ready\001

  • doc/theses/thierry_delisle_PhD/comp_II/img/system.fig

    r25a1cb0 rd3aa64f1  
    33Center
    44Inches
    5 Letter
     5Letter 
    66100.00
    77Single
    … …  
    1651654 1 -1 0 0 0 10 0.0000 2 105 990 2175 3525 Discrete-event\001
    1661664 1 -1 0 0 0 10 0.0000 2 135 795 2175 4350 preemption\001
    167 4 0 -1 0 0 0 10 0.0000 2 150 1290 2325 4875 generator/coroutine\001
     1674 0 -1 0 0 0 10 0.0000 2 150 1290 2325 4875 genrator/coroutine\001
    1681684 0 -1 0 0 0 10 0.0000 2 120 270 4050 4875 task\001
    1691694 0 -1 0 0 0 10 0.0000 2 105 450 7050 4875 cluster\001
Note: See TracChangeset for help on using the changeset viewer.