Bestand wählen
Merken

Identifying Hotspots in the PostgreSQL Build Process

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
we will see some of the churches the not at all and you don't have to go and we the we it the action is what we want a major and 1 of them at a high everybody on mining chain 1st of all it is is not some opportunity for me because I am a PhD student from Canada and most of the conferences I attend the research this is like up actual stuff is happening the of yeah so things for welcoming me in my talk is about to have identifying hot spots in the post Wilson should before I start here's my Twitter handle
and this would be really awesome for me and my future research how you guys could actually take a look at my Twitter handle I just we a bank and it's a survey devolver survey will take maybe 5 minutes of your time to fill out and it's just about how you interact with build systems in these sorts of things that
really help me during my future work if you can show that for anyways back to the task at hand so we did some work on identifying hot spots in poster and find it it's usually useful for me to take a few minutes to explain what I mean by the old system forest thinking because this kind of overlapping terms everybody thinks something different C so so if we imagine that this part of my goal while blocks is the source code of the software project and this spaceship is what we want wanted to our customers while the build system is the set of
order dependent instructions the take from walks to the actual situation warrants offer terms of makefiles figuration scripts these sorts of things that describe how source code text files documentation files are translated into deliverables in package the and we've been using build systems developers have for quite some time now and since the seventies the original make tool came out in this concept of the developers work cycle of thinking of the current thinking of something like change the editing code to implement the change running to build to integrate the changes that the and then testing what you produce and then you circle back around half the this model all builds were thought of as equals so any time around there was just doing the build step but in reality some builds are more equal than others it's so some builds actually take longer than I was and this 1 reason xk college kind of highlights that saying that 1 of the number 1 reason for developers slacking off this the code is compiled and waiting for the detection and got some quotes from gtk developers who say you know that this is really small and preventing them from being able to get the job done so slowly builds a actually impacting people that's the impression so working people do you tend to address build performance and so refactoring is a contact so if we imagine that we have for file system very simple system made up foresee files in there we notice that this particular file when we change it takes a while to to reintegrate is 1st we need to recompile this object than 3 archive it and then it gets linked into to co before refactoring this source tree for commands trade when you change utilize that's however if we look at this graph and we find out that utilize the goalie ever used in the 1st level it's never used in the 2nd 1 we can actually be factor this going so it's something like this were now if you feudal that is changed only 1 compile commands treated and then 1 deliverables wait so now
only 2 commands are interested in checking should compile time I but the question is now emerging imagine working real system this the where do I focus laughter where should I look for gains to the fact and we thought the most obvious 1 is to look for the file so that is the slowest to you go from the turtles the so we came up with that I use a two-step approach to find the to finding the slope parts of the 1st step we extract the dependency graph from the software system and the way that we do that all explained the next slide the 2nd step is we can always that graph to find out where the decision files so I'm going to each step in the morning and in the 1st thing the 1st step for constructing the dependency graph so we use this tool called my cell which is a tool that will positive debug trace and the builds and they'll understand recursion as well and piece together the dependency graph so what I'm showing here is actually pose 90 2 . 4 which from the ancient for you guys know about and that's the version that we studied and this is the dependency graph and so on the outside edges our out the source files so these are the ones that there's no command to generate and then as you were closer and that's where things are getting compiled and linked in the source of so yet we have this tool to build this graph how in the graph we don't have a cost for energy so we could look at that graph and just say the Costa Rican compiling a file is just all the edges and get tree but instead we decided to do something a little bit more practical and look at how how much each edge costs so how long each compile each link everything costs so we did 10 clean rebuilds most grass and finding each command and then we take the median those and that's the cost for each edge so no doubt a nice way of finding out how how much it costs to rebuild far so we said let's try this so the factory systems and was formed the slowly building up so before I die
then here's some characteristics of those graphs produces the following year highlighted in yellow suppose responded to for we also what did she lived and worked in cheat and I'm showing in blue the number of edges or sorry the number of nodes so these are files this . se files that all files executables libraries these sorts of things and then the red of bars are edges dependencies in the rules that trigger so you can see we looked at 3 systems with 3 different orders of magnitude event so I appeared qt was the largest and number 2 . 7 million edges in the dependency graph posters has been 121 thousand and then dual for circles with smallest she was 120 so yeah so we said now let's go through each 1 of the source file so that the nodes on the outside of that big graph that have no images end traced through the graph to see how much it would cost to rebuild each flight and then we plot the graphs that look like this so on the left hand side of the see following an incentive header files In each point on the graph is 1 file and we sorted according to the hilt cost so the stuff on the far right is the stuff that took longest path so we saw some interesting things here like for example down here is where we would expect most . si files to be trigonal 1 compiled and maybe a couple of links but there were these jumps between the Jedi where some see false actually triggered several of them on the walls in several when commands as well so we dug in and found that they had some code generators there were a few change that's the file and regenerating some code and recompiling rankings and stuff so yeah there was some interesting stuff there on here in the header files you can see there again steps some of these header files were included in the source code generator which meant that they would treat the same sort of the and as you'd expect on the Y axis at the top here see false if you touch them tend to rebuild very maybe 12 seconds this time whereas if the if you touching header files you pay a higher price the this is what PostgreSQL like which was much more of a lot what we expect touching CVOs never really took more than 2 sheets of but we found that if you did touch on some of the bad pedophiles it take more time 3 minutes to recompile and where we tend to Q 2 years is worse and these guys are in trouble is that the there some c files that actually took almost 3 minutes to compile and there's some pedophiles and took over 2 hours to pass so if you if you're unlucky enough to have to touch 1 of those in the the effects whatever you work but in any case we said this is great you know we can sort the staff can find the father really slow rebuild all we have to do to find this stuff you should focus on for refactoring is just draw a simple life I don't wait more than 100 seconds from close to build so everything above this line the I need to look at the that or maybe a more patient in the essay and this the last patients and 50 seconds but either way all we have to do with strong it's a great so let's take an arbitrary line something we felt was bad was 90 seconds that books 90 seconds seems to be how long people say no versatile tool so we said was that these pedophiles or files that took longer than 90 seconds to the you but merely the as of these files do you think any of the elements in the get scary so listen to all of you know none of them are hot spots we don't care what's been going on painful and then we ask them why because none of these files change that often so we went back to the drawing
board originally said with the little files maybe it's not the slow files that of because they might only rarely change so the only really change people don't have to pay the price of the so maybe we should be looking at the files that change the most often the reason OK and I would find the files that change most often returns to and the version
control system and we did some simple might be just with for example for each file the number of changes that have to assumed the files and change the light in the past will continue to change in future so then we said OK with those same grants from line there what we found is that many of these calls already built 3 so we thought OK that's not gonna work either so focusing on the files that change the most often also is insufficient because they may already be optimal so much you can do the factor a file that builds in lessons so what we suggest is that instead of trying 1 of the other you should focus on both so we call these files build hot the the files that we build slowly but also they change frequently so people have to pay that cost the building lot so we took this two-step approach that we originally had and turned it into a three-step so we added a 3rd step we took the version control of stats that mind integrated with the stuff we pulled from penalty graph to get what we call logical this is a pretty simple concept just take to rebuild cost and the number of changes for each file and plot them against each other and that's so each 1 of these dots would represent 1 files in your system and we split into quadrants so based on thresholds this page your project we say I'm not willing to wait more than 90 seconds seconds 20 minutes whatever it is that's use threshold for the cost and then you pick a threshold for the number of changes so 10 changes in the past and what or 100 changes in the past 1 that becomes your threshold and then we just say focus on this red area those the files that you want to those are the files that you want factors so it's a great now we've got a metal detector go try not to the sources and see what kind of hot spots and so for this we have to pick some thresholds like I mentioned we've been playing with 90 seconds because we have these books that that's a pretty good threshold and we also need to take a number of changes to the threshold and for that we take the media so anything that changes more than the median number of changes for filing project we consider that the frequently change so some of the things that we found using quadrant plots for the 3 systems 3 open source this recently been exploring hour so about 7 per cent of GTK this energy lives on codebase was identified as a hotspot using those thresholds so about
65 g the files and some some of things we found that you don't actually have to be factor all 65 sometimes that all bottle neck into 1 place and there were some main culprits reason that you could you could focus but noted here is that the maximum amount of time it takes to rebuild is about 2 and a half to know minutes there from produce well returned to duty aboard organized from the 2 hour 1 build we found that there were a lot of hot spots 732 herself but is roughly 80 per cent of the source code also on par with what we're seeing in uh and she would have as far as the main culprits we started trying to manually analyze this stuff to find something called respectively just 732 files was too much for us as outsiders looking and really familiar with the system so instead what we did is we found that the components of just directories that had the highest hotspot concentrations so like of the files in that directory how many of them to point is hot spots and some of them with really high percentages would call a library for the use XML patterns but again that they had a really bad scenario where you could have files rebuilding very very slow there we turned oppose grass and that we found only 2 per cent of the files you actually hot spots so the job to those of you who were close to problems where you go I 27 hotspots still from 20 so hot spots here and in the worst case it was just just over 3 minutes to go in some of these hotspots and so some of the main culprits IdentiFinder here is far follows so actually if you if you survive the internally goes to the heart of the actions the so can a shift now to what we've learned from doing all these studies enhanced by is it the kind of highlighted this transitive trend for for hot spot in this fucking make turn and so if we imagine that we have of an object file the gets went into a huge list of 2 peoples huge list of a deal Olsen executables and this file changes frequently we were displayed as a hot spot so this this this C file here would be considered a hot spot but if the sea floor actually ended up including header file this header file can actually change his header file can trigger the same amount of elected so it's going end up having to recompile object while we make everything and now this header file also changes frequently it's also going to be a hot spot but that's not it this header file is also included in another header and that the file includes several other his house and so those are also hotspots garbled call a super hot spot is it changing this final lecture truthfully hotspots to be compiled then keeps getting content and the further out you go so the 1st we have here is that you should try to limit the use of these header file Hobson header files that just include the header files because they tend to accumulate builds the activity and if they change frequently now these become hotspots the so just to quickly summarize what I'm talking about so we started out saying what I meant by builds the set of order dependent tasks that have to be executed in order to translate the source code and documentation into something that you customers users and developers reliability and quick so that they can make their edits change their deliverables and test them quickly to keep the feedback fast but there are some files that and taking longer than week and they can really interfere with developers so what we set out to do is come up with a way of finding these files so that you can refactor and improve their performance so we 1st started out trying to find just the files rebuilt slowly that didn't work out so well we tried the files that change the most often which also didn't work out so well when they were combined so we call these files build hotspots in the what are the ones that build slowly but also taken 1 uh change change very frequently and came up with an approach to find these files based on thresholds the finally we did we did some open-source studies using our approach found some hotspots and found some general trends and we think it helps avoid making spots so that's a that's all about the same the hot spots now and happy to take questions from the guys at the same time they remind you again that we put a survey and it's a you can find it might for the length of the really helpful to me and happy to feed back the results to you guys if you could take the 5 minutes of been suggested and should be and without is there any questions for a change I will go on and on the part some yeah this was just single process and 1 of the things we want look at is using this this graph you can actually find how mate with pairwise things you can actually find the problems that the lot of that you you have can fire and not that of all the all the time even when using recursive make that you had any problems with that all the customer OK we call this a lot and I know the present there are the all the the I guess there's 1 other thing before we go into questions any further so my supervisor at the supervisory series at the back end is organizing a workshop on release engineering so maybe he can tell you a couple words about to look the cost of the the the the the the community the the the what is the book that exist in the it have to do use a by the use of the the war the you know that sort of thing and have a year you In this and the other is the the you been the share of the it will also be the on the and I believe so that the there was yeah yeah the problem the group of the journal of the American Society 3 but when you go the part of this the only thing in the of the process of
going on remember the the the of reducing the in the and the the the current in the because of the and in the process of the really small version of progress in your regression tests on the smart enough to know only the status changed only 1 a subset of of each of the new rules of all this work thank a you're going this is the number of reason of aluminum little or no thought experiment where this we we have success all of which will be used to school some parallel and in series and get something about the use of your this and what it is that the the how frequently you can test battery changes number of all we reminiscent of the rules of the schools in the mountains of the from version of this part of the and for around and she'd be to the to growth for the most of the work of the literal from where we have multiple this on the use of a and the and the only way of the and the number of the of and moving in and out of the use of optical imaging and most of the real world and the only thing I can think of of the of the former and all that and the release reduce the you probably about the bigger problem and over all the end of the book of the the and of the will be as and and the and this being the you the rest of the the thing but 1st we were in the list is the so this stuff that we're you know we some of the reasons why the you what you know the the question of the in the you want some of the the so at this stage it's still research which means it's not on the and tweaking would yeah this is the very much the the this study takes a long time and it's just make sure that the cost of each the that's again a tree and you don't want to say that the cost the yes essentially every every step that you might have to be so that the list of things that we will people it's very the and have to go the it and you get value of and and you can and I could be integrated within continuous every change you see what impact this a large part of the the and the this is the sum of the of the of the of the you in in the the rhetoric of some of the 3 years somebody was found to we all of of history with a view of the and when you see the very most you also use the rule of law and we the exchange and you know you know that you and I guess I also of school that time I had some bonus slides which name the so I mentioned defined in the main culprits these are the ones that through the chain of including including including including end up being hot spots in that so I don't know if any would jump at the top of your head is being recently claimed that get to some 512 the trend into the main post this this page to details well each time I indented means another level of say yeah you can see like this one's like 4 levels deep but becomes a hot hotspot because it was needed the also it's change more and more so yeah of the some of you have to do the use of the beauty of the tools used for unnecessary use them is that the and you so whether any other questions professor delta and you come out of this house and opportunity 1 more time serving and the and the soul the next station here will be at 5 and that will be the final and will be announced that from my wife the order of the work and you can load of the user and the elections results in
the theory of radiation and model the In the
you you you you you you know you have a lot of the of the New York remote formed from the Alexander
returning to the heart of the of the of the of the of the of the of the rule of law and
the state of the the world and in the the the of the of the of the of this over here and but
this is what is the value of using the the the
the problem that in the final part of the figure in the in the in the
in the use of the of the of the of the of the of the of the of the of
the of the likelihood of the data in the way of the of the of the of the
of the of the of the function and the and the and the
and the area of the world with the do you have a lot of knowledge of the uh you can find a lot of the history state of
the of the use of the theory about of the of the of the all of the bonds in
the relation to the knowledge of use of the world and you know what I would have to rely on the of the
the the of the of the triple checking the the level of the the so what is the rate of the
of the of the this that and the the the view of the of the of the of the the the the the
the the the the the the the the of the the of the the of the the the
idea of the artist and the use of my space so we don't usually it's the the the the the the
the the the the philosophical problem with the rest of the of the
universe and we did some of the people on the material of the sole support the theory
of of the of the of the volume of the of the of the of the of the of the of the of the of the time support of the
of the of the of the of the of the of the the the president and so on so that you know the the
the the weights and also I find that were expressed the view of the goal of this course but we don't have the whole community
nearly all of the ability of the United but that's the name of the view on the last week we'll kind of girl so in mobilizing point with the label of the inference that we
used to think what we know there but from the source of of the of of slow even in areas with the use of test particularly bring
expensive so what about so really what it is you know you would think that this was this was the the only thing we can do that you know that we're going
from the the of the aboriginal substitutes for oral come for all in little people right out there will be no it's it's because
of from the scale of the things you want this I was in the hope that you're going to have to the use of the we don't always agree unintelligible character of the people in the text book of actually at
the moment will move to war the goal of answers are there are a lot of sense so that's fusion history actually this from 0 to if you will what will think of the prosumers
unit but in the where possible using a platform and goes into 1 of the main problems with your own production work in the reversion test for those of all those
things in the necessary to follow the approach so the because the 1st time so the basis the tests about about which the history of the world
more mindful about it just saying that were assigned to a lot of people don't use it to the talk the the users and some interesting model
similar to what I have to point 1 of the 1st to use it in a little bit of a model of the world in the American so think about
the kinds of thinking you there was 1 of the things we know the president of the median of the sum of foreign aid is because it is you like some of the datasets and you follow this follows from and yes to
figure out where you need to do and that's those and only and another question you know over on the cost of the fuel tank and works of test and by some and
most the work of but those qualities of knowledge about how the features it and show that these is are much more common for most of the
difference of the actual decision theory of related and some of the the the the the the the muscle hand health or relationships with
a view of the world because of we wouldn't wanna excluding things like you know for example if you read the text so there were going to the very 1st so that the breeze
remaining also all the only run this as a source of all sorts of ammonia biographer and friend of the largely we tend to form 2 ammonia
and we still want we will we will not to be the forming of a group of researchers in the in the woods and things of what's called the the quality of life for how use them different
from her and the family course in the yeah the profile of test variations of the
use of the of the quantum effects are expensive and the people in the eye of the given the current report on the of this
is the in the In the world of some fruitful or most of the of the problem so what we're interested in in the
in the company of the of country and of the of of the world little move on to the right of this
and that the evidence of the analysis to the sum of the oxidation form you know of of the
of the of the of the of the of the people of field the of the parts of the world and they have a lot to lose through the
proposals for half an hour of the day and the reason is that the Internet and I think that the value of slope this and this
a lot of the the this all this the the the
the the the the the the the the and
the fact that we also about a lot and in the a bit of the you know the the the the the
the the and at the end of the that the idea of the of the of the of
the of the ASR is the world that at the end of the day in this in the in
the world of the world in of you to all the available data is the focus of people's who was the 1st to do and how
much of the of the of the moment that the the the the the the the the the the the the the the the volatility that has a lot of time 45 minutes as usual with all of the that have the effect of the number of people in the upper part of the of the model and the fact that it's getting to know the noise in the the and all of this was some of the theories we get a hold of the law of the total and then you have to look at it might take 10 to the right of of the of the year and that resulted only members must be used with all of the the visible in previous sessions of actually the stand up for the we did what we did is for that that that the you know so again you made of all of this but would not you just have it all the time the in the 1st of what's the of and that is for some over here yeah but it has a lot of the time you have a and you know the the point of view this is that the use of the of moon at the in the in the and the little the at the end the I the up at the same the we a few points the limitations of the and the last that the the in the in the in the and and the the the the the the the to the and they and and the I the the number of thank the and and the and the correct at the at the the at and the the the in the in the and this is the at the but it the and a it but it it looked at if it's tt but but but but but but but but but but but but but but but but but but but but but but but the but in the but the perspective In this paper you want it to the but the where the the the the the the the way the and the reason is because everything work the only way to do it serves as the the this is quite obviously and the rest of the world the the 2 In the better the the because the the and in the in the the the In the and the and this is a the the it the be at the the the the the it it it it it the the the this the the the and 1 of the the truck in the the movement of goods from 1 side of the what was your membership in the group and the 1st false hope that the address of the corporations but I used the good and of the would call of the and people from borders and languages of those whose doesn't but and present was that the at the the woman who was which the the the 2nd 1 uses a lot of almost instantly all the this time so we're going to do about we just to a the the 1 approach forced the and the don't we can be and is and and FIL and so and in this study and this was the lessons of history and that the market meeting of the latest so it was quite a lot of the the use of is and the of what's a lot out of me a little slower than the constant that was the son of a quorum dance the whole of the In the event of closer to that of the of the go ahead at the end
Verkettung <Informatik>
Gruppenoperation
t-Test
Data Mining
Task
Physikalisches System
Wald <Graphentheorie>
Twitter <Softwareplattform>
Software
Mereologie
Projektive Ebene
Quellcode
Physikalisches System
Sondierung
Term
Quick-Sort
Computeranimation
Bit
Versionsverwaltung
Computeranimation
Übergang
Eins
Netzwerktopologie
Prozess <Informatik>
Softwarewartung
Dateiverwaltung
Skript <Programm>
Figurierte Zahl
Graphentheorie
Softwareentwickler
Strömungsrichtung
Quellcode
Gleichheitszeichen
Teilbarkeit
Entscheidungstheorie
Rechenschieber
Refactoring
Ablaufverfolgung
Mathematisierung
Zahlenbereich
Zellularer Automat
Term
Code
Wiederherstellung <Informatik>
Physikalisches System
Quellcode
Systemprogrammierung
Software
Modelltheorie
Softwareentwickler
Kreisfläche
Elektronische Publikation
Graph
Konvexe Hülle
Physikalisches System
Elektronische Publikation
Binder <Informatik>
Medianwert
Packprogramm
Quick-Sort
Objekt <Kategorie>
Energiedichte
Dreiecksfreier Graph
Mereologie
Turtle <Informatik>
Faktor <Algebra>
GRASS <Programm>
Rekursive Funktion
Subtraktion
Punkt
Stab
Mathematisierung
Versionsverwaltung
Zahlenbereich
Automatische Programmierung
Element <Mathematik>
Whiteboard
Code
Quellcode
Graph
Wechselsprung
Knotenmenge
Rangstatistik
Fokalpunkt
Programmbibliothek
E-Mail
Bildgebendes Verfahren
Gerade
Gammafunktion
Soundverarbeitung
Videospiel
Softwareentwickler
Elektronische Publikation
Kreisfläche
Graph
Zwei
Schlussregel
Physikalisches System
Quellcode
Binder <Informatik>
Elektronische Publikation
Ereignishorizont
Quick-Sort
Generator <Informatik>
Refactoring
Größenordnung
Charakteristisches Polynom
Graphentheorie
Refactoring
Resultante
Prozess <Physik>
Punkt
Extrempunkt
Gemeinsamer Speicher
Versionsverwaltung
Gruppenkeim
Kartesische Koordinaten
Sondierung
Computeranimation
Homepage
Eins
Prozess <Informatik>
Mustersprache
Regelkreis
Plot <Graphische Darstellung>
E-Mail
Gerade
Verschiebungsoperator
Softwaretest
Schwellwertverfahren
Dicke
Reihe
Globale Optimierung
Systemaufruf
Plot <Graphische Darstellung>
Quellcode
Teilbarkeit
Gruppenoperation
Konzentrizität
Menge
Twitter <Softwareplattform>
Kategorie <Mathematik>
Projektive Ebene
Verzeichnisdienst
Rückkopplung
Mathematisierung
Gruppenoperation
Zahlenbereich
Task
Quellcode
Front-End <Software>
Fokalpunkt
Programmbibliothek
Softwareentwickler
Ereignishorizont
Beobachtungsstudie
Elektronische Publikation
Graph
Open Source
Zwei
Mailing-Liste
Statistische Analyse
Physikalisches System
Paarvergleich
Elektronische Publikation
Medianwert
Fokalpunkt
Quick-Sort
Objekt <Kategorie>
Energiedichte
Skalarprodukt
Flächeninhalt
Komponente <Software>
Hypermedia
Mereologie
Wort <Informatik>
Resultante
Offene Menge
Elektromagnetische Schwingung
Prozess <Physik>
Gewichtete Summe
Mathematisierung
Versionsverwaltung
Zahlenbereich
Technische Optik
Gesetz <Physik>
Computeranimation
Übergang
Eins
Homepage
Netzwerktopologie
Open Source
Online-Katalog
Arithmetische Folge
Lineare Regression
Arbeitsplatzcomputer
Modelltheorie
Parallele Schnittstelle
Hacker
Bildgebendes Verfahren
Schreib-Lese-Kopf
Softwaretest
Beobachtungsstudie
Sichtenkonzept
Graph
Speicher <Informatik>
Schlussregel
Mailing-Liste
Knotenmenge
Rechenschieber
Verkettung <Informatik>
Twitter <Softwareplattform>
Mereologie
Offene Menge
Open Source
Schlussregel
Gesetz <Physik>
Hacker
Computeranimation
Gammafunktion
Aggregatzustand
Offene Menge
Open Source
Mereologie
Bildschirmfenster
Figurierte Zahl
Hacker
Computeranimation
Gammafunktion
Offene Menge
Open Source
Lineares Funktional
Flächeninhalt
Likelihood-Funktion
Hill-Differentialgleichung
Hacker
Computeranimation
Aggregatzustand
Offene Menge
Open Source
Uniforme Struktur
Relativitätstheorie
Gewichtete Summe
Bitrate
Hacker
Computeranimation
Übergang
Offene Menge
Open Source
Sichtenkonzept
Gewichtete Summe
Hacker
Raum-Zeit
Computeranimation
Quantisierung <Physik>
Offene Menge
Open Source
Spezifisches Volumen
Grundraum
Lie-Gruppe
Hacker
Physikalische Theorie
Computeranimation
Gammafunktion
Software Engineering
Quellcode
Offene Menge
Open Source
Punkt
Sichtenkonzept
Gewicht <Mathematik>
Inferenz <Künstliche Intelligenz>
Extrempunkt
Gerade
Hacker
Computeranimation
Gammafunktion
Softwaretest
Quellcode
Open Source
Flächeninhalt
Rechter Winkel
Videospiel
Likelihood-Quotienten-Test
Verbandstheorie
Hacker
Computeranimation
Softwaretest
Offene Menge
Open Source
Zentrische Streckung
Momentenproblem
Reverse Engineering
Biprodukt
Systemplattform
Zehn
Hacker
Computeranimation
Offene Menge
Open Source
Bit
Basisvektor
Modelltheorie
Hacker
Computeranimation
Gammafunktion
Softwaretest
Offene Menge
Open Source
Gewichtete Summe
Figurierte Zahl
Medianwert
Hacker
Computeranimation
Offene Menge
Quellcode
Open Source
Subtraktion
Sichtenkonzept
Relativitätstheorie
Quellcode
Hacker
Quick-Sort
Computeranimation
Entscheidungstheorie
Softwaretest
Videospiel
TVD-Verfahren
Offene Menge
Gruppenkeim
Familie <Mathematik>
Profil <Aerodynamik>
Quantenmechanik
Computeranimation
Quellcode
Open Source
Neunzehn
Verkehrsinformation
Hacker
Offene Menge
Open Source
Bildschirmmaske
Gewichtete Summe
Rechter Winkel
Division
ABEL <Programmiersprache>
Hacker
Computeranimation
Analysis
Offene Menge
Open Source
Datenfeld
Mereologie
Hacker
Computeranimation
Internetworking
BEEP
Offene Menge
Open Source
Bit
Einheit <Mathematik>
Videospiel
Hacker
Computeranimation
Soundverarbeitung
Beobachtungsstudie
Offene Menge
Total <Mathematik>
Sichtenkonzept
Punkt
Momentenproblem
Formale Sprache
Adressraum
Güte der Anpassung
Gruppenkeim
Zahlenbereich
Geräusch
Element <Mathematik>
Fokalpunkt
Gesetz <Physik>
Ereignishorizont
Physikalische Theorie
Computeranimation
Open Source
Verbandstheorie
Perspektive
Mereologie
Inverser Limes
Modelltheorie
Hacker

Metadaten

Formale Metadaten

Titel Identifying Hotspots in the PostgreSQL Build Process
Serientitel FOSDEM 2014
Autor McIntosh, Shane
Lizenz CC-Namensnennung 2.0 Belgien:
Sie dürfen das Werk bzw. den Inhalt zu jedem legalen Zweck nutzen, verändern und in unveränderter oder veränderter Form vervielfältigen, verbreiten und öffentlich zugänglich machen, sofern Sie den Namen des Autors/Rechteinhabers in der von ihm festgelegten Weise nennen.
DOI 10.5446/32547
Herausgeber FOSDEM VZW
Erscheinungsjahr 2014
Sprache Englisch

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract Software developers rely on a fast and correct build system to compile their source code changes and produce modified deliverables for testing and deployment. The scale and complexity of the PostgreSQL build process makes build performance an important topic to discuss and address. In this talk, we will introduce a new build performance analysis technique which identifies "build hotspots", i.e., files that are slow to rebuild (by analyzing a build dependency graph), yet change often (by analyzing version control history). We will discuss the identified hotspots in the 9.2.4 release of PostgreSQL

Ähnliche Filme

Loading...
Feedback