Meta tags:
Headings (most frequently used words):
2013, monday, thursday, blog, lock, mechanical, sympathy, may, 2014, 26, august, tuesday, 16, july, 27, june, 14, february, search, this, discussion, group, about, me, training, consulting, popular, posts, archive, followers, simple, binary, encoding, based, vs, free, concurrent, algorithms, java, garbage, collection, distilled, printing, generated, assembly, code, from, the, hotspot, jit, compiler, cpu, cache, flushing, fallacy,
Text of the page (most frequently used words):
the (683), and (216), for (182), can (111), are (101), this (91), that (91), with (80), memory (69), cache (67), objects (61), when (59), from (53), have (47), generation (47), #collection (46), not (43), which (43), java (41), latency (40), space (39), concurrent (38), will (38), application (38), old (37), collector (37), time (36), then (34), throughput (33), object (31), all (30), minor (30), more (29), message (29), collections (29), garbage (28), size (28), you (27), data (27), line (27), sbe (27), heap (26), very (26), each (26), some (26), collectors (26), promotion (26), code (25), these (25), live (25), system (24), performance (24), cms (24), new (24), tenured (23), may (22), other (22), has (22), than (22), often (21), parallel (21), run (21), allocation (21), fields (21), messages (20), also (20), world (20), stop (20), been (19), systems (19), lock (19), applications (19), pauses (19), free (19), most (18), major (18), store (18), thread (18), significant (18), cost (17), access (17), your (17), cpu (17), need (17), number (17), read (17), threads (17), such (17), interthreadlatency (17), regions (17), many (16), single (16), share (16), caches (16), core (16), socket (16), algorithms (16), int (16), car (16), use (15), survivor (15), our (15), state (15), used (15), hotspot (15), note (15), encoding (14), how (14), should (14), available (14), instructions (14), being (14), buffer (14), pongrunner (14), final (14), references (14), append (14), main (13), sub (13), version (13), case (13), out (13), over (13), test (13), static (13), but (13), region (13), large (13), stubs (13), because (12), into (12), like (12), known (12), young (12), eden (12), spaces (12), between (12), must (12), long (12), get (12), while (12), generated (12), rate (12), exec (12), thrpt (12), ops (12), times (11), make (11), any (11), see (11), work (11), cores (11), shared (11), design (11), variable (11), possible (11), cycle (11), order (11), value (11), instruction (11), thus (11), buffers (11), set (11), events (11), generational (11), implementation (11), binary (10), back (10), high (10), multiple (10), one (10), within (10), local (10), allocated (10), write (10), full (10), implementations (10), using (10), mark (10), point (10), low (10), less (10), hlt (10), decoding (10), fullgc (10), schema (10), gpb (10), simple (9), load (9), its (9), could (9), algorithm (9), off (9), real (9), tuning (9), via (9), example (9), same (9), effectively (9), both (9), only (9), written (9), protocol (9), does (9), greater (9), compiler (9), allows (9), method (9), volatile (9), sometimes (9), assembly (9), found (9), target (9), would (9), reduced (9), compaction (9), rates (9), keep (8), different (8), gen (8), public (8), hardware (8), older (8), achieve (8), per (8), across (8), registers (8), class (8), copy (8), changes (8), bus (8), ordering (8), require (8), based (8), much (8), what (8), jit (8), jvm (8), due (8), targets (8), strings (8), support (8), pause (8), roots (8), first (8), failure (8), results (8), market (8), protobuf (8), zing (7), financial (7), writer (7), even (7), flush (7), fallacy (7), them (7), server (7), similar (7), larger (7), try (7), consistent (7), controller (7), another (7), ring (7), second (7), issue (7), issues (7), efficient (7), avoid (7), operation (7), generations (7), optimisation (7), few (7), r10 (7), rax (7), here (7), end (7), cannot (7), running (7), footprint (7), collecting (7), during (7), safepoint (7), copying (7), promoted (7), reachable (7), good (7), concurrently (7), length (7), spaceship (7), coordinates (7), xml (7), fix (7), encode (7), decode (7), ncar (7), 2013 (6), requests (6), interesting (6), fast (6), done (6), numbers (6), about (6), perm (6), london (6), martin (6), thompson (6), location (6), x86 (6), bridge (6), processors (6), following (6), qpi (6), modified (6), track (6), return (6), fence (6), addition (6), address (6), two (6), however (6), significantly (6), cycles (6), depending (6), they (6), model (6), figure (6), necessary (6), plugin (6), below (6), mov (6), pointer (6), r11d (6), pong (6), ping (6), amount (6), mostly (6), therefore (6), expect (6), occurs (6), runtime (6), triggered (6), purpose (6), stampedlock (6), encoded (6), codec (6), carbenchmark (6), testdecode (6), testencode (6), marketdatabenchmark (6), message_header (6), blog (5), follow (5), since (5), bytes (5), native (5), language (5), fences (5), important (5), vast (5), takes (5), structures (5), highly (5), having (5), labels (5), pinterest (5), facebook (5), blogthis (5), email (5), comments (5), posted (5), execution (5), programs (5), intel (5), processor (5), sandy (5), non (5), follows (5), coherent (5), never (5), increased (5), requirements (5), previous (5), connected (5), network (5), made (5), there (5), variables (5), loop (5), register (5), achieved (5), user (5), observed (5), context (5), switch (5), required (5), reference (5), indicates (5), level (5), sets (5), additional (5), accessing (5), private (5), entry (5), useful (5), particular (5), appropriate (5), performed (5), makes (5), rather (5), rsp (5), start (5), fly (5), those (5), discussed (5), give (5), reduce (5), mean (5), job (5), milliseconds (5), efficiency (5), result (5), before (5), moved (5), sizes (5), contention (5), tlab (5), occur (5), byte (5), collect (5), sufficient (5), compared (5), mutation (5), seen (5), incremental (5), little (5), tens (5), general (5), typical (5), tool (5), api (5), methods (5), position (5), presentation (5), languages (5), directbuffer (5), schemaid (5), july (4), august (4), under (4), until (4), bit (4), anything (4), channels (4), azul (4), article (4), barriers (4), miss (4), sending (4), compact (4), build (4), principle (4), operations (4), serial (4), next (4), logic (4), coherence (4), common (4), attempt (4), function (4), cpus (4), multi (4), independent (4), process (4), capable (4), allow (4), small (4), executed (4), cross (4), keeping (4), source (4), know (4), separate (4), provides (4), clean (4), writing (4), reasons (4), compilers (4), generate (4), reads (4), sequentially (4), simply (4), setting (4), tlb (4), cases (4), later (4), way (4), usually (4), around (4), enable (4), without (4), come (4), runs (4), take (4), jvms (4), oracle (4), standard (4), linux (4), measure (4), associated (4), above (4), normally (4), followed (4), best (4), 0x0 (4), compiled (4), total (4), repetitions (4), focus (4), commercial (4), great (4), worst (4), frequent (4), trading (4), tend (4), resulting (4), longer (4), tends (4), collected (4), suffer (4), reach (4), array (4), event (4), phase (4), directly (4), proportional (4), fail (4), threshold (4), fixed (4), spent (4), needs (4), increasing (4), uses (4), default (4), suitable (4), consider (4), become (4), compacting (4), fragmentation (4), marking (4), logs (4), remembered (4), jrockit (4), ratio (4), tests (4), mode (4), readers (4), principles (4), header (4), template (4), protocols (4), bufferoffset (4), templateid (4), 2014 (3), pressure (3), doing (3), programming (3), patterns (3), people (3), terms (3), last (3), had (3), trying (3), cause (3), flushing (3), hundreds (3), popular (3), posts (3), view (3), modern (3), limited (3), groups (3), mechanical (3), sympathy (3), search (3), latest (3), resource (3), their (3), huge (3), together (3), hierarchy (3), opportunity (3), means (3), exists (3), taking (3), effective (3), expense (3), amd (3), segment (3), held (3), containing (3), exclusive (3), always (3), relatively (3), telling (3), after (3), issuing (3), put (3), ensure (3), part (3), believe (3), classes (3), flushed (3), whole (3), expensive (3), atomic (3), nehalem (3), forward (3), copies (3), associative (3), hash (3), associativity (3), trade (3), now (3), pre (3), average (3), split (3), segments (3), designed (3), mob (3), floating (3), allocate (3), advantage (3), understand (3), decisions (3), applied (3), install (3), targeting (3), was (3), printing (3), print (3), optimised (3), able (3), field (3), xchg (3), oop (3), 0x00007fed81060a00 (3), stub (3), 0x00007fed810609d0 (3), cmp (3), 0x00007fed81060980 (3), void (3), string (3), pongthread (3), program (3), amd64 (3), find (3), tradeoffs (3), something (3), comes (3), batch (3), seconds (3), maintain (3), short (3), whereas (3), immortal (3), almost (3), age (3), operate (3), managed (3), specific (3), where (3), were (3), meet (3), card (3), table (3), grow (3), bump (3), providing (3), call (3), lists (3), increases (3), gets (3), too (3), adding (3), approach (3), updating (3), maintaining (3), influence (3), range (3), compactions (3), frequency (3), extended (3), constrained (3), websphere (3), advantages (3), millisecond (3), needed (3), least (3), three (3), dynamic (3), spikes (3), update (3), better (3), reader (3), dimensional (3), reentrantreadwritelock (3), provide (3), complex (3), representation (3), move (3), param (3), delta (3), year (3), tag (3), working (3), community (3), todd (3), codecs (3), input (3), decoded (3), blobs (3), root (3), struct (3), repeating (3), metadata (3), sbetool (3), wrap (3), abstraction (3), direct (3), extension (3), alignment (3), overview (3), specification (3), feedback (3), benchmarks (3), 29193 (3), 600 (3), 34078 (3), 646 (3), 11657 (3), 190 (3), 10436 (3), 476 (3), 619 (3), serialnumber (3), modelyear (3), october (2), december (2), 2012 (2), february (2), june (2), respond (2), accepting (2), response (2), applying (2), quite (2), ever (2), turn (2), stream (2), largest (2), penalty (2), communicate (2), receiving (2), formats (2), detailed (2), implications (2), lot (2), quest (2), limitation (2), scalability (2), writers (2), experienced (2), technologists (2), hear (2), talk (2), certain (2), safepoints (2), distilled (2), training (2), consulting (2), profile (2), https (2), google (2), com (2), forum (2), group (2), seems (2), explain (2), execute (2), techniques (2), ends (2), nanoseconds (2), hide (2), exercise (2), others (2), slower (2), developing (2), lost (2), considered (2), stale (2), whereby (2), place (2), works (2), bandwidth (2), tracks (2), finite (2), states (2), employs (2), mesif (2), sent (2), traffic (2), sockets (2), lanes (2), inclusive (2), either (2), worry (2), visibility (2), appear (2), firstly (2), periods (2), repeatedly (2), updates (2), aware (2), ahead (2), sequential (2), impact (2), want (2), think (2), subsequent (2), indexing (2), entries (2), request (2), numa (2), matches (2), transitions (2), involves (2), plus (2), stage (2), tables (2), ways (2), hashed (2), levels (2), uniform (2), maybe (2), remote (2), hop (2), westmere (2), aggregated (2), link (2), individual (2), accesses (2), through (2), addresses (2), acts (2), queue (2), contains (2), 32k (2), already (2), waiting (2), searched (2), existing (2), transferred (2), writes (2), stores (2), expected (2), files (2), integers (2), accessible (2), constitute (2), fastest (2), given (2), located (2), thursday (2), debugging (2), previously (2), installed (2), programmer (2), sufficiently (2), reading (2), situation (2), once (2), again (2), reordered (2), add (2), stack (2), semantics (2), action (2), efficiently (2), oopmap (2), jmpq (2), runtime_call (2), 0x00007fed81060a0a (2), handler (2), exception (2), iload_1 (2), eax (2), rip (2), 0x00007fed8ba77000 (2), rbp (2), 0x10 (2), 102 (2), jne (2), 0x00007fed810609c0 (2), r9d (2), goto (2), nopl (2), lang (2), constants (2), 0x00007fed81060850 (2), 0x00007fed81060b28 (2), 0x00007fed81060b30 (2), scopes (2), 0x00007fed81060a78 (2), 0x00007fed81060a30 (2), 0x00007fed81060a18 (2), unlockdiagnosticvmoptions (2), 1000 (2), throws (2), duration (2), runtest (2), pingthread (2), pingrunner (2), nanotime (2), implements (2), runnable (2), busy (2), spin (2), hsdis (2), disassembler (2), flags (2), tune (2), describing (2), every (2), choosing (2), usage (2), play (2), missed (2), acceptable (2), regardless (2), workloads (2), occasionally (2), period (2), behavior (2), lived (2), magnitude (2), likely (2), sparse (2), survived (2), generates (2), spending (2), portion (2), happens (2), living (2), lower (2), distinct (2), checks (2), executing (2), reached (2), potentially (2), eventually (2), basis (2), behind (2), wrong (2), adjust (2), avoiding (2), cheap (2), commences (2), fit (2), differently (2), dead (2), include (2), internal (2), survive (2), potential (2), 512 (2), stored (2), cards (2), scan (2), empty (2), area (2), copied (2), doubling (2), increase (2), heaps (2), starting (2), greatest (2), fills (2), effort (2), prevent (2), accommodate (2), fullgcs (2), experience (2), linearly (2), perform (2), 7u4 (2), extent (2), tolerate (2), compacted (2), gain (2), goal (2), driven (2), worse (2), taken (2), grows (2), phases (2), check (2), requires (2), introduced (2), supported (2), humongous (2), moving (2), referenced (2), evacuating (2), problem (2), well (2), scenarios (2), alternative (2), still (2), truly (2), higher (2), side (2), analysis (2), histogram (2), jhiccup (2), lifetimes (2), printgcapplicationstoppedtime (2), maxgcpausemillis (2), reset (2), sweep (2), updated (2), initial (2), maxtenuringthreshold (2), xmx (2), maximum (2), concurrency (2), session (2), optimistic (2), current (2), fact (2), indirection (2), built (2), harness (2), pattern (2), inlining (2), employed (2), variance (2), produce (2), especially (2), interface (2), provided (2), attempts (2), xdelta (2), axis (2), ydelta (2), monday (2), feeds (2), gives (2), ascii (2), presentations (2), complain (2), cme (2), who (2), format (2), grouped (2), sniffer (2), bitsets (2), enums (2), arrays (2), type (2), body (2), offsets (2), block (2), optional (2), supporting (2), unsafe (2), carry (2), otherwise (2), aligned (2), parts (2), structure (2), otf (2), developers (2), don (2), welcome (2), 1316 (2), 123 (2), 2088 (2), 998 (2), 433 (2), 711 (2), 467 (2), msg (2), 1242 (2), 252 (2), 1148 (2), 050 (2), 326 (2), 018 (2), 462 (2), 817 (2), thr (2), cnt (2), sec (2), error (2), units (2), messagetemplateversion (2), actingblocklength (2), blocklength (2), actingversion (2), stringbuilder (2), sbetemplateid (2), sbeschemaversion (2), jar (2), powered, blogger, followers, september, november, 2011, march, april, january, archive, sustained, deadly, hockey, overloaded, leaving, lmax, neglecting, opposi, fun, nirvana, wish, serialisation, discuss, fundamental, technique, computing, said, post, tuples, scalable, biggest, contend, item, course, www, complete, technology, geek, exploring, capabilities, development, services, twitter, mjpt777, united, kingdom, discussion, subscribe, atom, home, heirarchy, illustrating, interacts, fulfil, interact, concrete, dive, accessed, 10s, 100s, 100ns, 0ghz, 1200, retiring, ipc, employ, capacity, persistent, pondering, 500, double, interconnects, circa, servers, decomposed, truth, mainstream, fetched, master, style, management, evicted, blocks, reduces, variant, transition, series, effect, greatly, days, buses, module, integrated, graphics, rapid, identification, snooping, keeps, owns, wants, snooped, guaranteed, why, modifications, visible, qualifying, beware, inadequate, reorder, shortly, likes, overcome, let, accompanied, preventing, progress, drained, marker, boundary, vega, opposed, returning, safely, say, drain, policy, arm, armv6, did, tags, requiring, kept, bottom, switching, topic, pollution, lazyset, peterson, dekker, acknowledge, snoop, hypertransport, opteron, specialised, designated, unused, invalid, exclusively, ownership, rfo, invalidate, broadcast, dirty, moesi, l1d, masking, bits, handle, collision, slot, slots, hold, storing, power, 40ns, step, limit, raised, 4gt, 0gt, eliminating, bottleneck, delivered, choke, fetch, dram, 65ns, extremely, columns, row, queuing, effects, refresh, conflict, pipelining, 2mb, 20mb, inclusiveness, intercept, removing, burden, 256k, hidden, pipelined, comprised, flight, fully, queued, blocking, disambiguate, compliance, published, loads, ordered, constructed, according, 160, 144, arguments, subset, expands, ability, streams, enabled, hyperthreads, hyperthreading, architectural, asm, profiling, determining, affect, knowing, emitted, rules, bugs, various, illustrates, command, options, obtain, debug, sun, steps, executes, hot, details, optimise, honour, statement, forest, trees, generally, twice, osr, red, highlighted, lines, optimisations, staging, inserting, chosen, enforce, idiom, ideal, correctly, wonder, compromises, portable, architectures, observation, oopmapset, oopmaps, 0x00007fed81060a17, 0x00007fed81060a16, 0x00007fed81060a15, 0x00007fed81060a14, 0x00007fed81060a0f, 0x00007fed81038c00, subq, 0x5, 0x00007fed81060a05, callq, deopt, 0x00007fed8105eaa0, no_reloc, 0x00007fed810609ff, 0x00007fed810609fe, 0x00007fed810609fd, 0x00007fed810609fc, 0x00007fed810609fb, 0x00007fed810609fa, 0x00007fed810609f9, retq, poll_return, 0x00007fed810609f3, 0xaa16607, 0x00007fed810609f2, pop, 0x00007fed810609ee, 0x00007fed810609ec, if_icmpeq, 0x00007fed810609e5, 0x5f5e100, 0x00007fed810609e2, inc, iinc, 0x00007fed810609dd, addl, putstatic, 0x00007fed810609d9, 0x78, 0x00007fed810609d7, 0x00007fed810609d4, 000, 0x74, getstatic, 0x00007fed810609c6, nopw, poll, 0xaa1663a, 0x00007fed810609bb, 0x00007fed810609b9, jmp, 0x00007fed810609af, 0x7ad0fcbf0, 0x00007fed810609ac, xor, 0x00007fed810609a7, synchronization, 0x00007fed810609a0, 0x18, verified, 0x00007fed8106099c, 0x00007fed81060994, 0x00007fed81060991, 0x00007fed8106098b, 0x00007fed81037a60, 0x00007fed81060988, 0x00007fed81060984, shl, 0x3, 0x8, rsi, r10d, 0x20, caller, dependencies, pcs, 176, oops, 128, relocation, 0x00007fed81060970, 736, 10531, compilecommand, printassembly, import, 100, args, printf, dns, avg, interruptedexception, join, library, java_home, jre, lib, rename, libhsdis, download, kenai, projects, base, downloads, installation, startup, baffle, alone, documentation, feels, man, pages, aircraft, knob, dial, explained, nowhere, guide, workload, openjdk, ones, towards, illustrate, alternatives, wise, folk, silently, continue, warning, margin, distribution, interpret, term, lowest, refers, deterministic, processing, matter, paused, overall, completed, sooner, virtually, human, facing, interactive, goes, unresponsive, spell, disaster, worthwhile, physical, fronts, expectation, percentage, theory, iterators, experimentation, shown, ubiquitously, separating, newly, scavenges, record, consistently, fairly, filter, frequently, practical, engineering, periodically, points, varying, durations, bring, signaling, contents, returns, edges, optimized, away, places, making, dynamically, rare, cloning, monotonic, counted, bound, ttsp, consideration, surfaced, enabling, flag, undergo, scheduling, resume, released, reliance, explore, organised, initially, temporary, described, collectively, enough, knows, unfortunately, loading, ongoing, motivating, assumption, interned, treat, virtual, statistics, accordingly, assigned, allocates, tlabs, scale, bumps, roughly, platforms, cheaper, malloc, exhausted, filled, accommodated, created, handles, own, section, becomes, tenuring, overflows, iteratively, frames, graphs, graph, processed, sure, identifying, existence, corresponding, barrier, indicate, exist, represent, alternate, roles, beginning, includes, dominated, dealt, halved, doubled, traded, constant, rapidly, becoming, majority, variety, filling, predict, fill, begin, passed, promoting, unwound, padding, promotions, resizing, avoided, simplest, option, smallest, forms, multiprocessor, affected, retain, fewer, port, employing, collects, five, benefits, allocating, shame, feature, trigger, multistep, unreachable, reclaimed, sized, hole, holes, logged, reducing, brought, performing, sees, extremes, reduction, carried, fails, caused, late, addressed, might, officially, partially, tries, smaller, minimize, plague, organizes, dividing, 2000, contiguous, increments, costly, date, challenge, remember, marked, slow, down, scanned, patched, 200ms, efforts, futile, tendency, fragmented, regional, fall, just, evacuation, evacuated, challenging, dealing, heavily, choice, mid, called, look, clear, regularly, occurring, unsuitable, strict, reactive, interfaces, ibm, limitations, author, shifts, related, contained, digit, including, allowing, apply, fear, rule, thumb, budget, warranted, today, seldom, behaving, settings, nature, launch, jvisualvm, visual, understanding, applcations, representative, grips, configurations, experiments, perspective, capturing, outside, correlate, determine, causing, idle, hours, surprised, invest, combined, skilled, verses, purchasing, disruptor, hdrhistogram, chewiebug, verbose, xloggc, printgcdetails, printgcdatestamps, printtenuringdistribution, printgcapplicationconcurrenttime, monitoring, overflow, useg1gc, failures, failed, reclaiming, occupied, capture, remarking, useconcmarksweepgc, usenuma, useparalleloldgc, useparallelgc, useserialgc, xms, promotedpadding, pretenuresizethreshold, permgen, organisation, weak, hypothesis, infant, mortality, managing, controlling, inducing, amortized, active, retained, parameter, responding, impacted, equating, gctimeratio, nothing, tuesday, locks, week, attended, review, excellent, unconference, arise, attending, couple, things, occurred, thought, reviewed, status, secondly, although, looks, jdk, solution, compare, favour, atomically, transaction, cleaner, factoring, immutable, create, minimal, easily, exercised, coarsening, unrolling, subjected, threading, profiles, 0_25, quad, 2ghz, ivy, 3632qm, measured, repeated, warm, throughputs, averaged, approximate, deployment, affinity, isolation, operating, raw, charts, surprise, beyond, balance, takeaways, nice, seeing, appearing, strategy, evidenced, teaching, conclusion, offer, careful, measurement, reentrantlock, counts, discovered, synchronised, easy, mistakenly, locking, actions, improvement, megamorphic, dispatch, ship, readposition, jcrete, heinz, kabutz, jsr166, big, really, lets, quantify, finance, industry, exchanges, typically, emitting, thousands, aggregate, peak, million, volumes, growing, crazy, slightly, sane, markets, commit, sin, income, ultra, parsers, member, commissioned, 29west, lbm, fame, myself, aimed, addressing, coming, atonement, sins, past, worked, helped, net, amazing, cool, technical, team, dream, layer, spend, parsing, transforming, json, business, adhering, features, offered, consists, environment, log, viewer, drive, ensures, streamed, backtracking, unnecessary, underestimated, reclamation, applies, totally, predictable, 25ns, 1000ns, orders, sweet, spot, structured, restrictions, usability, users, suited, preserve, streaming, backtrack, insert, pointers, types, indirected, extra, losing, prefetchers, pure, identifies, nest, finally, define, characters, composites, repeat, currently, output, defined, flyweight
Text of the page (random words):
hat will not have been compiled with the stubs a full example of the ir being used can be found here direct buffers sbe via agrona provides an abstraction to java with the mutabledirectbuffer class to work with buffers that are byte heap or direct bytebuffer buffers and off heap memory addresses returned from unsafe allocatememory long or jni in low latency applications messages are often encoded decoded in memory mapped files via mappedbytebuffer and thus can be be transferred to a network channel by the kernel thus avoiding user space copies c and c have built in support for direct memory access and do not require such an abstraction as the java version does a directbuffer abstraction was added for c to support endianess and encapsulate the unsafe pointer access message extension and versioning sbe schemas carry a version number that allows for message extension a message can be extended by adding fields at the end of a block fields cannot be removed or reordered for backwards compatibility extension fields must be optional otherwise a newer template reading an older message would not work templates carry metadata for min max null timeunit character encoding etc these are accessible via static class level methods on the stubs byte ordering and alignment the message schema allows for precise alignment of fields by specifying offsets fields are by default encoded in little endian form unless otherwise specified in a schema for maximum performance native encoding with fields on word aligned boundaries should be used the penalty for accessing non aligned fields on some processors can be very significant for alignment one must consider the framing protocol and buffer locations in memory message protocols i often see people complain that a codec cannot support a particular presentation in a single message however this is often possible to address with a protocol of messages protocols are a great way to split an interaction into its component parts these parts are then often composable for many interactions between systems for example the ir implementation of schema metadata is more complex than can be supported by the structure of a single message we encode ir by first sending a template message providing an overview followed by a stream of messages each encoding the tokens from the compiler ir this allows for the design of a very fast otf decoder which can be implemented as a threaded interpreter with much less branching than the typical switch based state machines protocol design is an area that most developers don t seem to get an opportunity to learn i feel this is a great loss the fact that so many developers will call an encoding such as ascii a protocol is very telling the value of protocols is so obvious when one gets to work with a programmer like todd who has spent his life successfully designing protocols stub performance the stubs provide a significant performance advantage over the dynamic otf decoding for accessing primitive fields we believe the performance is reaching the limits of what is possible from a general purpose tool the generated assembly code is very similar to what a compiler will generate for accessing a c struct even from java regarding the general performance of the stubs we have observed that c has a very marginal advantage over the java which we believe is due to runtime inserted safepoint checks the c version lags a little further behind due to its runtime not being as aggressive with inlining methods as the java runtime stubs for all three languages are capable of encoding or decoding typical financial messages in tens of nanoseconds this effectively makes the encoding and decoding of messages almost free for most applications relative to the rest of the application logic feedback this is the first version of sbe and we would welcome feedback the reference implementation is constrained by the fix community specification it is possible to influence the specification but please don t expect pull requests to be accepted that significantly go against the specification support for javascript python erlang and other languages has been discussed and would be very welcome update 08 may 2014 thanks to feedback from kenton varda the creator of gpb we were able to improve the benchmarks to get the best performance out of gpb below are the results for the changes to the java benchmarks the c gpb examples on optimisation show approximately a doubling of throughput compared to initial results it should be noted that you often have to do the opposite in java with gpb compared to c to get performance improvements such as allocate objects rather than reuse them before gpb optimisation mode thr cnt sec mean mean error units exec u c r protobuf carbenchmark testdecode thrpt 1 30 1 462 817 6 474 ops ms exec u c r protobuf carbenchmark testencode thrpt 1 30 1 326 018 2 972 ops ms exec u c r protobuf marketdatabenchmark testdecode thrpt 1 30 1 1148 050 17 194 ops ms exec u c r protobuf marketdatabenchmark testencode thrpt 1 30 1 1242 252 12 248 ops ms exec u c r sbe carbenchmark testdecode thrpt 1 30 1 10436 476 102 114 ops ms exec u c r sbe carbenchmark testencode thrpt 1 30 1 11657 190 65 168 ops ms exec u c r sbe marketdatabenchmark testdecode thrpt 1 30 1 34078 646 261 775 ops ms exec u c r sbe marketdatabenchmark testencode thrpt 1 30 1 29193 600 443 638 ops ms after gpb optimisation mode thr cnt sec mean mean error units exec u c r protobuf carbenchmark testdecode thrpt 1 30 1 619 467 4 429 ops ms exec u c r protobuf carbenchmark testencode thrpt 1 30 1 433 711 10 364 ops ms exec u c r protobuf marketdatabenchmark testdecode thrpt 1 30 1 2088 998 60 619 ops ms exec u c r protobuf marketdatabenchmark testencode thrpt 1 30 1 1316 123 19 816 ops ms throughput msg ms before gpb optimisation test protocol buffers sbe ratio car encode 462 817 10436 476 22 52 car decode 326 018 11657 190 35 76 market data encode 1148 050 34078 646 29 68 market data decode 1242 252 29193 600 23 50 throughput msg ms after gpb optimisation test protocol buffers sbe ratio car encode 619 467 10436 476 16 85 car decode 433 711 11657 190 26 88 market data encode 2088 998 34078 646 16 31 market data decode 1316 123 29193 600 22 18 posted by martin thompson at 20 01 43 comments email this blogthis share to x share to facebook share to pinterest labels big data low latency performance location london uk monday 26 august 2013 lock based vs lock free concurrent algorithms last week i attended a review session of the new jsr166 stampedlock run by heinz kabutz at the excellent jcrete unconference stampedlock is an attempt to address the contention issues that arise in a system when multiple readers concurrently access shared state stampedlock is designed to perform better than reentrantreadwritelock by taking an optimistic read approach while attending the session a couple of things occurred to me firstly i thought it was about time i reviewed the current status of java lock implementations secondly that although stampedlock looks like a good addition to the jdk it seems to miss the fact that lock free algorithms are often a better solution to the multiple reader case test case to compare implementations i needed an api test case that would not favour a particular approach for example the api should be garbage free and allow the methods to be atomic a simple test case is to design a spaceship that can be moved around a 2 dimensional space with the coordinates of its position available to be read atomically at least 2 fields need to be read or written per transaction to make the concurrency interesting interface to a concurrent representation of a ship that can move around a 2 dimensional space with updates and reads performed concurrently public interface spaceship read the position of the spaceship into the array of coordinates provided param coordinates into which the x and y coordinates should be read return the number of attempts made to read the current state int readposition final int coordinates move the position of the spaceship by a delta to the x and y coordinates param xdelta delta by which the spaceship should be moved in the x axis param ydelta delta by which the spaceship should be moved in the y axis return the number of attempts made to write the new coordinates int move final int xdelta final int ydelta the above api would be cleaner by factoring out an immutable position object but i want to keep it garbage free and create the need to update multiple internal fields with minimal indirection this api could easily be extended for a 3 dimensional space and require the implementations to be atomic multiple implementations are built for each spaceship and exercised by a test harness all the code and results for this blog can be found here the test harness will run each of the implementations in turn by using a megamorphic dispatch pattern to try and prevent inlining lock coarsening and loop unrolling when accessing the concurrent methods each implementation is subjected to 4 distinct threading scenarios that result in different contention profiles 1 reader 1 writer 2 readers 1 writer 3 readers 1 writer 2 readers 2 writers all tests are run with 64 bit java 1 7 0_25 linux 3 6 30 and a quad core 2 2ghz ivy bridge i7 3632qm throughput is measured over 5 second periods for each implementation with the tests repeated 5 times to ensure sufficient warm up the results below are throughputs averaged per second over 5 runs to approximate a typical java deployment no thread affinity or core isolation has been employed which would have reduced variance significantly note other cpus and operating systems can produce very different results results figure 1 figure 2 figure 3 figure 4 the raw data for the above charts can be found here analysis the real surprise for me from the results is the performance of reentrantreadwritelock i cannot see a use for this implementation beyond a case whereby there is a huge balance of reads and very little writes my main takeaways are stampedlock is a major improvement over existing lock implementations especially with increasing numbers of reader threads stampedlock has a complex api it is very easy to mistakenly call the wrong method for locking actions synchronised is a good general purpose lock implementation when contention is from only 2 threads reentrantlock is a good general purpose lock implementation when thread counts grow as previously discovered choosing to use reentrantreadwritelock should be based on careful and appropriate measurement as with all major decisions measure and make decisions based on data lock free implementations can offer significant throughput advantages over lock based algorithms conclusion it is nice seeing the influence of lock free techniques appearing in lock based algorithms the optimistic strategy employed on read is effectively a lock free algorithm at the times when a writer is not updating in my experience of teaching and developing lock free algorithms not only do they provide significant throughput advantages as evidenced here they also provide much lower and less variance in latency posted by martin thompson at 11 48 50 comments email this blogthis share to x share to facebook share to pinterest labels algorithms concurrency java lock free locks performance location london uk tuesday 16 july 2013 java garbage collection distilled serial parallel concurrent cms g1 young gen new gen old gen perm gen eden tenured survivor spaces safepoints and the hundreds of jvm startup flags does this all baffle you when trying to tune the garbage collector while trying to get the required throughput and latency from your java application if it does then do not worry you are not alone documentation describing garbage collection feels like man pages for an aircraft every knob and dial is detailed and explained but nowhere can you find a guide on how to fly this article will attempt to explain the tradeoffs when choosing and tuning garbage collection algorithms for a particular workload the focus will be on oracle hotspot jvm and openjdk collectors as those are the ones in most common usage towards the end other commercial jvms will be discussed to illustrate alternatives the tradeoffs wise folk keep telling us you do not get something for nothing when we get something we usually have to give up something in return when it comes to garbage collection we play with 3 major variables that set targets for the collectors throughput the amount of work done by an application as a ratio of time spent in gc target throughput with xx gctimeratio 99 99 is the default equating to 1 gc time latency the time taken by systems in responding to events which is impacted by pauses introduced by garbage collection target latency for gc pauses with xx maxgcpausemillis n memory the amount of memory our systems use to store state which is often copied and moved around when being managed the set of active objects retained by the application at any point in time is known as the live set maximum heap size xmx n is a tuning parameter for setting the heap size available to an application note often hotspot cannot achieve these targets and will silently continue without warning having missed its target by a great margin latency is a distribution across events it may be acceptable to have an increased average latency to reduce the worst case latency or make it less frequent we should not interpret the term real time to mean the lowest possible latency rather real time refers to having deterministic latency regardless of throughput for some application workloads throughput is the most important target an example would be a long running batch processing job it does not matter if a batch job is occasionally paused for a few seconds while garbage collection takes place as long as the overall job can be completed sooner for virtually all other workloads from human facing interactive applications to financial trading systems if a system goes unresponsive for anything more than a few seconds or even milliseconds in some cases it can spell disaster in financial trading it is often worthwhile to trade off some throughput in return for consistent latency we may also have applications that are limited by the amount of physical memory available and have to maintain a footprint in which case we have to give up performance on both latency and throughput fronts tradeoffs often play out as follows to a large extent the cost of garbage collection as an amortized cost can be reduced by providing the garbage collection algorithms with more memory the observed worst case latency inducing pauses due to garbage collecting can be reduced by containing the live set and keeping the heap size small the frequency with which pauses occur can be reduced by managing the heap and generation sizes and by controlling the application s object allocation rate the frequency of large pauses can be reduced by concurrently running the gc with ...
|
