Profiling Particular person Queries in a Concurrent System

A superb CPU profiler is value its weight in gold. Measuring efficiency in-situ normally means utilizing a sampling profile. They supply a number of data whereas having very low overhead. In a concurrent system, nonetheless, it’s laborious to make use of the ensuing knowledge to extract high-level insights. Samples don’t embody context like question IDs and application-level statistics; they present you what code was run, however not why.

This weblog introduces trampoline histories, a method Rockset has developed to effectively connect application-level data (question IDs) to the samples of a CPU profile. This lets us use profiles to know the efficiency of particular person queries, even when a number of queries are executing concurrently throughout the identical set of employee threads.

Primer on Rockset

Rockset is a cloud-native search and analytics database. SQL queries from a buyer are executed in a distributed vogue throughout a set of servers within the cloud. We use inverted indexes, approximate vector indexes, and columnar layouts to effectively execute queries, whereas additionally processing streaming updates. Nearly all of Rockset’s performance-critical code is C++.

Most Rockset clients have their very own devoted compute sources known as digital cases. Inside that devoted set of compute sources, nonetheless, a number of queries can execute on the identical time. Queries are executed in a distributed vogue throughout the entire nodes, so which means a number of queries are lively on the identical time in the identical course of. This concurrent question execution poses a problem when making an attempt to measure efficiency.

Concurrent question processing improves utilization by permitting computation, I/O, and communication to be overlapped. This overlapping is particularly vital for top QPS workloads and quick queries, which have extra coordination relative to their elementary work. Concurrent execution can also be vital for lowering head-of-line blocking and latency outliers; it prevents an occasional heavy question from blocking completion of the queries that observe it.

We handle concurrency by breaking work into micro-tasks which can be run by a hard and fast set of thread swimming pools. This considerably reduces the necessity for locks, as a result of we will handle synchronization by way of job dependencies, and it additionally minimizes context switching overheads. Sadly, this micro-task structure makes it troublesome to profile particular person queries. Callchain samples (stack backtraces) may need come from any lively question, so the ensuing profile exhibits solely the sum of the CPU work.

Profiles that mix the entire lively queries are higher than nothing, however a number of guide experience is required to interpret the noisy outcomes. Trampoline histories allow us to assign a lot of the CPU work in our execution engine to particular person question IDs, each for steady profiles and on-demand profiles. This can be a very highly effective software when tuning queries or debugging anomalies.

DynamicLabel

The API we’ve constructed for including application-level metadata to the CPU samples is known as DynamicLabel. Its public interface could be very easy:

class DynamicLabel {
  public:
    DynamicLabel(std::string key, std::string worth);
    ~DynamicLabel();

    template 
    std::invoke_result_t apply(Func&& func) const;
};

DynamicLabel::apply invokes func. Profile samples taken throughout that invocation can have the label connected.

Every question wants just one DynamicLabel. At any time when a micro-task from the question is run it’s invoked by way of DynamicLabel::apply.

One of the crucial vital properties of sampling profilers is that their overhead is proportional to their sampling charge; that is what lets their overhead be made arbitrarily small. In distinction, DynamicLabel::apply should do some work for each job whatever the sampling charge. In some instances our micro-tasks could be fairly micro, so it’s important that apply has very low overhead.

apply‘s efficiency is the first design constraint. DynamicLabel‘s different operations (development, destruction, and label lookup throughout sampling) occur orders of magnitude much less continuously.

Let’s work by some methods we’d attempt to implement the DynamicLabel performance. We’ll consider and refine them with the aim of creating apply as quick as doable. If you wish to skip the journey and bounce straight to the vacation spot, go to the “Trampoline Histories” part.

Implementation Concepts

Thought #1: Resolve dynamic labels at pattern assortment time

The obvious strategy to affiliate utility metadata with a pattern is to place it there from the start. The profiler would search for dynamic labels on the identical time that it’s capturing the stack backtrace, bundling a duplicate of them with the callchain.

Rockset’s profiling makes use of Linux’s perf_event, the subsystem that powers the perf command line software. perf_event has many benefits over signal-based profilers (resembling gperftools). It has decrease bias, decrease skew, decrease overhead, entry to {hardware} efficiency counters, visibility into each userspace and kernel callchains, and the flexibility to measure interference from different processes. These benefits come from its structure, through which system-wide profile samples are taken by the kernel and asynchronously handed to userspace by a lock-free ring buffer.

Though perf_event has a number of benefits, we will’t use it for thought #1 as a result of it could’t learn arbitrary userspace knowledge at sampling time. eBPF profilers have an identical limitation.

Thought #2: Document a perf pattern when the metadata adjustments

If it’s not doable to tug dynamic labels from userspace to the kernel at sampling time, then what about push? We may add an occasion to the profile each time that the thread→label mapping adjustments, then post-process the profiles to match up the labels.

A method to do that can be to make use of perf uprobes. Userspace probes can file operate invocations, together with operate arguments. Sadly, uprobes are too sluggish to make use of on this vogue for us. Thread pool overhead for us is about 110 nanoseconds per job. Even a single crossing from the userspace into the kernel (uprobe or syscall) would multiply this overhead.

Avoiding syscalls throughout DynamicLabel::apply additionally prevents an eBPF resolution, the place we replace an eBPF map in apply after which modify an eBPF profiler like BCC to fetch the labels when sampling.

edit: eBPF can be utilized to tug from userspace when accumulating a pattern, studying fsbase after which utilizing bpfprobelearnconsumer() to stroll a userspace knowledge construction that’s connected to a threadnative. When you’ve got BPF permissions enabled in your manufacturing setting and are utilizing a BPF-based profiler then this various could be a good one. The engineering and deployment points are extra complicated however the end result doesn’t require in-process profile processing. Due to Jason Rahman for pointing this out.

Thought #3: Merge profiles with a userspace label historical past

If it is too costly to file adjustments to the thread→label mapping within the kernel, what if we do it within the userspace? We may file a historical past of calls to DynamicLabel::apply, then be a part of it to the profile samples throughout post-processing. perf_event samples can embody timestamps and Linux’s CLOCK_MONOTONIC clock has sufficient precision to look strictly monotonic (at the very least on the x86_64 or arm64 cases we’d use), so the be a part of can be actual. A name to clock_gettime utilizing the VDSO mechanism is rather a lot quicker than a kernel transition, so the overhead can be a lot decrease than that for thought #2.

The problem with this strategy is the information footprint. DynamicLabel histories can be a number of orders of magnitude bigger than the profiles themselves, even after making use of some easy compression. Profiling is enabled constantly on all of our servers at a low sampling charge, so making an attempt to persist a historical past of each micro-task invocation would shortly overload our monitoring infrastructure.

Thought #4: In-memory historical past merging

The sooner we be a part of samples and label histories, the much less historical past we have to retailer. If we may be a part of the samples and the historical past in near-realtime (maybe each second) then we wouldn’t want to put in writing the histories to disk in any respect.

The commonest approach to make use of Linux’s perf_event subsystem is by way of the perf command line software, however the entire deep kernel magic is on the market to any course of by way of the perf_event_open syscall. There are a number of configuration choices (perf_event_open(2) is the longest manpage of any system name), however when you get it arrange you may learn profile samples from a lock-free ring buffer as quickly as they’re gathered by the kernel.

To keep away from competition, we may keep the historical past as a set of thread-local queues that file the timestamp of each DynamicLabel::apply entry and exit. For every pattern we’d search the corresponding historical past utilizing the pattern’s timestamp.

This strategy has possible efficiency, however can we do higher?

Thought #5: Use the callchains to optimize the historical past of calls to `apply`

We will use the truth that apply exhibits up within the recorded callchains to scale back the historical past measurement. If we block inlining in order that we will discover DynamicLabel::apply within the name stacks, then we will use the backtrace to detect exit. Because of this apply solely wants to put in writing the entry data, which file the time that an affiliation was created. Halving the variety of data halves the CPU and knowledge footprint (of the a part of the work that’s not sampled).

This technique is one of the best one but, however we will do even higher! The historical past entry data a spread of time for which apply was certain to a selected label, so we solely must make a file when the binding adjustments, relatively than per-invocation. This optimization could be very efficient if we now have a number of variations of apply to search for within the name stack. This leads us to trampoline histories, the design that we now have carried out and deployed.

Trampoline Histories

If the stack has sufficient data to search out the best DynamicLabel , then the one factor that apply must do is go away a body on the stack. Since there are a number of lively labels, we’ll want a number of addresses.

A operate that instantly invokes one other operate is a trampoline. In C++ it would seem like this:

__attribute__((__noinline__))
void trampoline(std::move_only_function func) {
    func();
    asm unstable (""); // forestall tailcall optimization
}

Notice that we have to forestall compiler optimizations that might trigger the operate to not be current within the stack, particularly inlining and tailcall elimination.

The trampoline compiles to solely 5 directions, 2 to arrange the body pointer, 1 to invoke func(), and a pair of to wash up and return. Together with padding that is 32 bytes of code.

C++ templates allow us to simply generate a complete household of trampolines, every of which has a novel handle.

utilizing Trampoline = __attribute__((__noinline__)) void (*)(
        std::move_only_function);

constexpr size_t kNumTrampolines = ...;

template 
__attribute__((__noinline__))
void trampoline(std::move_only_function func) {
    func();
    asm unstable (""); // forestall tailcall optimization
}

template 
constexpr std::array makeTrampolines(
        std::index_sequence) {
    return {&trampoline...};
}

Trampoline getTrampoline(unsigned idx) {
    static constexpr auto kTrampolines =
            makeTrampolines(std::make_index_sequence{});
    return kTrampolines.at(idx);
}

We’ve now obtained the entire low-level items we have to implement DynamicLabel:

• DynamicLabel development → discover a trampoline that’s not at the moment in use, append the label and present timestamp to that trampoline’s historical past
• DynamicLabel::apply → invoke the code utilizing the trampoline
• DynamicLabel destruction → return the trampoline to a pool of unused trampolines
• Stack body symbolization → if the trampoline’s handle is present in a callchain, search for the label within the trampoline’s historical past

Efficiency Impression

Our aim is to make DynamicLabel::apply quick, in order that we will use it to wrap even small items of labor. We measured it by extending our current dynamic thread pool microbenchmark, including a layer of indirection by way of apply.

{
    DynamicThreadPool executor({.maxThreads = 1});
    for (size_t i = 0; i 

Maybe surprisingly, this benchmark exhibits zero efficiency affect from the additional degree of indirection, when measured utilizing both wall clock time or cycle counts. How can this be?

It seems we’re benefiting from a few years of analysis into department prediction for oblique jumps. The within of our trampoline seems like a digital methodology name to the CPU. That is extraordinarily frequent, so processor distributors have put a number of effort into optimizing it.

If we use perf to measure the variety of directions within the benchmark we observe that including label.apply causes about three dozen additional directions to be executed per loop. This may sluggish issues down if the CPU was front-end certain or if the vacation spot was unpredictable, however on this case we’re reminiscence certain. There are many execution sources for the additional directions, in order that they don’t truly enhance this system’s latency. Rockset is usually reminiscence certain when executing queries; the zero-latency end result holds in our manufacturing setting as nicely.

A Few Implementation Particulars

There are some things we have carried out to enhance the ergonomics of our profile ecosystem:

• The perf.knowledge format emitted by perf is optimized for CPU-efficient writing, not for simplicity or ease of use. Although Rockset’s perf_event_open-based profiler pulls knowledge from perf_event_open, we now have chosen to emit the identical protobuf-based pprof format utilized by gperftools. Importantly, the pprof format helps arbitrary labels on samples and the pprof visualizer already has the flexibility to filter on these tags, so it was simple so as to add and use the data from DynamicLabel.
• We subtract one from most callchain addresses earlier than symbolizing, as a result of the return handle is definitely the primary instruction that will probably be run after returning. That is particularly vital when utilizing inline frames, since neighboring directions are sometimes not from the identical supply operate.
• We rewrite trampoline to trampoline in order that we now have the choice of ignoring the tags and rendering a daily flame graph.
• When simplifying demangled constructor names, we use one thing like Foo::copy_construct and Foo::move_construct relatively than simplifying each to Foo::Foo. Differentiating constructor varieties makes it a lot simpler to seek for pointless copies. (For those who implement this be sure you can deal with demangled names with unbalanced and >, resembling std::enable_if 4, void>::sort.)
• We compile with -fno-omit-frame-pointer and use body tips to construct our callchains, however some vital glibc features like memcpy are written in meeting and don’t contact the stack in any respect. For these features, the backtrace captured by perf_event_open‘s PERF_SAMPLE_CALLCHAIN mode omits the operate that calls the meeting operate. We discover it by utilizing PERF_SAMPLE_STACK_USER to file the highest 8 bytes of the stack, splicing it into the callchain when the leaf is in a kind of features. That is a lot much less overhead than making an attempt to seize your complete backtrace with PERF_SAMPLE_STACK_USER.