Bestand wählen
Merken

Evaluation of Web Processing Service Frameworks

Zitierlink des Filmsegments
Embed Code

Automatisierte Medienanalyse

Beta
Erkannte Entitäten
Sprachtranskript
hello everyone thanks for coming in this presentation i'm going to talk about the result of a quantitative and qualitative evaluation of word processing services the after a
brief introduction and talk about developing standard itself and accountability as servers that we use in this study then I'll describe the methodology presented experimental results and discuss the relative advantages and disadvantages of each WPS server and challenges experience while working with them and the whole summer as our findings in the with the development of
geospatial services and Web Asturias have moved towards the servicer into products so today I'm artificial services can be used to you are effectively support common task of the common tasks that are GAS User Mr. do like it a discovery or service discovery using colored service whether there were Sicily data access using WFS WCS I mean what Peter service with current service and data processing using WBS for the present services in this study we focus on the WPS which is a standard of web service interface for publishing special processes on the web and this special process can be an algorithm is a simple calculation a complex model and which can which are exposed as a Service instance and operating a spatial data With the emergence of just processing on the Web
villages has been used in a wide variety of applications for example from exposure assessment to ecological modelling the increasing use of the WPS instances has also raised the pertinent and quality concerns and mean users and developers alike at concerned about quality of service actually is like performance with reliability capacity security and etc. so there are 2 viewpoints from user's perspective of performance and ease of use are 2 most important factors when choosing and web based application and demo them the main questions that they might ask our how fast is that all our but is it easy to use and decide that has a direct effect on other was actually it's like the capacity is scalability reliability and etc. and in this study we just focus on the performance site from developers perspectives the concerns are twofold from a quantitative perspective performance as 1 of the most important factors that can ensure both users and developers satisfaction from a qualitative perspective a quality concerns such as Software Architecture perceived ease of use of qualities so the quality of the communication community support and this of things are most important stuff when you have a developer and uh this factors can guide developers during selection of WPS that but if it fits a particular application best from some recent works have been done
to validate a spatial with services for example Web Map Services WMS or SOS sensor observation services and there are also few attempts to evaluate web processing services but they mainly focused on the influentials the performance issues w ps protocol and specifications and that compliance and interoperability testing so we need to know there is a need to a valid
WPS functionality and performance but why should we care about the performance of the WPS so it helps developers to identify their strengths and weaknesses of the double server and improved be service to meet both application users and developers requirements to and before going further I'm
going to talk about an adult PS and the WPS a standard and the pieces that we use in this society of other that so as you may know WPS is a standard web service interface for publishing and special algorithms and analyzes on the web from a service interface operas perspective it has 3 mandatory operations get capabilities to get the service metadata this rep process to get the detailed information about a process such as input data and output data parameters or data formats and then execute to invoke the process and the results and from a service implementation perspective it can be developed using a 3 D different should be bindings like uh a peers can be accessed using KTP via but should be get XML to viation pose or so weasel approach and here is the
list of the WBS there is that we use in this study all of them are free and open source the 1st 350 to North degree and user on a Java-based frameworks and develop as service so they can be did not deployed in any sort of 10 10 out of like PY WPS is a Python-based those peer server which runs as a CGI and so you can't deployed as that in any http server like party and 0 about using C C C + + and also runs CGI so can be deployed in any should be server as well and this is the only and server in this list that supports more than 1 programming language to develop a new process and these are the Libya servers that support different final libraries as you can see in this table and from an axis the the viewpoint and 52 North degree MPY WPS can be accessed through as should be get post and soul and deuterons lexical only through a she begin opposed the to evaluate the
performance of those APIs servers we use that what replace transit model which is an axis ability assessment model developed for the blindness this project it can be accessed through the link provided on the screen right based on this model if the user and provide them a location they would like to what from uh the maximum time they're willing to walk average what the speed and maximum time they 1 1 of the 1 0 wait for a transit and the maximum time they wanna travel that transient the system will evaluate at the extent of the area which is accessible using pedestrian and transit infrastructure because what then they want to do as if based on a the POI point of interest like store Park Hospital in center and also based on local crimes in the area and then there and I facilities score and a crime index and the watch it are returned to the client here's suffer development of
methodology that we used to design implement and deploy and test the and framework the what properties framework from a service design perspective using as a bottom-up approach to design its services then depardieu processing services were developed using Python in a way that to be accessible via this should begin and post In this index with special function of where used to perform geometry calculations like uh um calculating distances between pairs of points managing partner polygons calculating in areas of polygons and etc. and then remaining functionality was developed using Python libraries after that presence services where around and exposed using of 5 different no service like 52 nor the producers title appears as you and that the last step is to validate the developed framework from both quantitative and qualitative perspectives here's the architecture we
used to change the developed and dual pricing services which is a workflow manager chain architecture there is enough to think that is to strike a balance between flexibility and is usability and performance so as you can see here there are a 7 model appears instances that involved in this workflow so there are 2 kind of services in this forkful the delegates instances and the back end you processing services the WPS instances were developed using a of vision or the reducer appeal P assigns you uh API and the back end your presence there was ever developed using Python posteriors but I mean the WPS instances act like a gateway to an old D understand the interaction between the client and the back and your person serves as which article is in 5 and here's the you know the sequence diagram that shows how the system works the 1st step is you up the 1st that at the 1st step the clients and then execute requests to the management WPS which manage the whole thing the and then demanded WPS send the request to the watcher WPS so here's the input and output of the process that they wanted the logistics and 7 input parameters and return at a point on which is uh duties on imported the the 2nd is that is to send a request to try the WPS to retrieve all the transit stops within the area and then calculated a what should for each stops
so here's the input and output of the process but the transit WPS and returns a multiple it on feature so but which is also due dates are encoded the the
next the next step is to send a request to Pew RIW-PSO to retrieve all the pure lies in the area the input of this process is a multiple of them returned by transit WPS and the output do of this 1 is the union the union of the PS at so it takes 10 and the multiple it on feature returned by the time the WPS and red and returns the have marriage single a polygon the next step is to send a request appeal RIW-PSO Stewartry all the pure and here's the input and output of the process the input is a single a polygon and then the output is a point set their that is to call crime WPS to get the incident locations within the area again the input is a single for the future and the output is a there's a point set and the last step is to send a request to the aggregation WPS to calculate the score the axis with score and the counter the crime index of the area and then I WPS send the result to the client so years of the input of the process which are the 7 input parameters and output is this and watch it with axis will here's score and a crime index to evaluate the performance of
the WPS servers we designed to to test scenarios now for this that and the purpose of scenario a is to evaluate how the how fast it's a response In this scenario we randomly randomly chosen value as input parameters to generate 45 execute requests and the number of concurrent requests was assumed constant equal to 1 and even parameters were selected using the following criteria for example for the walking start point locations were selected using a stratified random sampling approach within the bounds of the city of Calgary and therefore the the other parameters for example for what is start time it was between 5 and 12 P M and etc. you can see list industry the and for the 2nd scenario would focus on the number of concurrent requests to assess the effect of increased load on the server so in this and as the number of concurrent request was generated using it to power and pattern while variable and was selected with 0 7 so we have 1 2 4 8 to 128 concur requests and 30 delegates requests were generated for cellobiose service and replicated according to the pattern and all for all other criteria were determined using the average for Saturday like what start time working time period and a sister and to deploy and test servers under the same conditions is WPS server was installed on a separate virtual machine and using a harder confirmation of the host machine and the virtual machine on the screen the machines to run the test at the client side was the host machine and to run the test and measure the performance factor that response time response size failure rate and other stuff we use Apache the result of the 1st scenario is listed in this table for area a we measured the average response time and the average response size for WPS server the response time is the time taken for a service call to return all response bytes and this font size is the quantity of data exchanged between the client and server so as shown in the uh in the table the most academic years was degree followed by a GeoServer the future North Zou and puedo PS in terms of response size the Jews WPS return the least amount of data and to decline as PY WPS had the most but the 1 we anova test
showed that all the appears although this service response as similarly and there was no significant differences between them in terms of response time response size it the for the scenario be recorded response time service failure and service through port so we all know what the response time is what the service failure rate represent the number of requests for the server could not handle and server throughput represents the number of requests per unit of time the server handles we use the result for our response time as you can see in the figure of the egregious server and you generally performed similarly and the only difference is that there is an improvement in response time here for 110 AD AT and their request so we can conclude that degree tended to perform better especially on high loads the moreover 52 north and the orderly PS filter response what processing more than 16 and 64 requests respectively this is a failure of the great results in the degree and you server so at the same failure rate pattern P. appears and you also follow the same pattern as it constant between 4 and 64 Denver request and then approached you 100 per cent on the piles as again see in this are figure and a few fit North had difficulty when processing more than 16 and and requests and with more 30 choosing was the failure rate of 52 north approach 100 per cent the the and finally the throughput results as you can see all the WP servers and performed similarity in terms of throughput and throughput was many affected by the number of concurrent requests as you can see in the figure and especially for future north which returned less than 1 successful assesses request per hour once spent request increase of 16 and all servers performs substantially better than when only 1 request was received at a time which is to North achieving a thousand 445 scissors request for our followed by Xu Juicer degree and do it with ps the excuse me so the
quantitative comparison is done so I'm going to have review their relative advantages and disadvantages of each of the via server and challenges experience while working with them In this context of those few servers were evaluated from a qualitative perspective in terms of ease of installation and configuration procedures of you have flexibility to for creating new processes native support for development and languages Quality of the contagion community support so for installation as you all know vision north degree and use are server-based applications of decision processes for that you just need to build a download the lawful and then putting them and put it in a certain general atomic and you're good to go but for puedo ps and Suderman problem is library dependency the lots of libraries that have an should item must install before for training new processes that the 2 North you surrounds you where well documented so a new process was simple to create an easy to configure for example 452 north WPS so we just need to use WPS STK created a job class for the process exporter process as a jar file and then deployed adding to the 50 and 50 North those framework for a dealer PS there was a lack of proper documentation and examples so it was not clear how to build and deploy a process for a purely PS the server occasionally return it should be our 500 and especially after adding a new process so this makes the creation of a new process difficult and to resolve this failure several access permissions settings were required and for development languages except for as you although the lipase from was supporters 1 native did the programming language for developing new process as a superset in programming languages so it's uh adds flexibility for developers to genes and a programming language very comfortable with and in terms of quality of documentation 52 north and GeoServer had a comprehensive documentation which provides clear clear instruction for installation confirmation other node here service along with clear instructions for developing new processes and for community supports except for P. would only PS all other delegates from worse have a large community of users and developers and provide different communication mediums to support them like wikis and as CNN's forums mailing lists get have such the
so to wrap up we used to test scenarios to other WPS service from a quantitative perspective In the 1st scenario we tested the performance of WPS servers which while holding the number of 100 requests constant equal to 1 the results show that on average degree was the most rapid WPS however at 1 know what test and indicates that there is no significant difference between the lobbyists areas and in terms of response on time or some size and for the 2nd scenario load testing was performed by varying the number of concurrent requests and degree and you server performs similarly although the degree tended to perform better under high levels and give it to North had difficulty when a processing more than 60 requests but for from better and more effectively under loads copper 2 other there pieces and to conclude that when selecting an appropriate Olimpia server and I believe it's important to consider both quantitative and qualitative and Criteria and Metrics and the importance of each metric can be a weighted based on the key for application requirements but generally speaking have from users that perspective performance is 1 of the most important factors that should be considered and from a developers perspective quality qualitative factors such as ease of installation and configuration of a variety of developer languages Quality of documentation and accessibility support there are more critical so to choose a model be server to suggest starting with a with an evaluation process with a basic set of questions that are linked to the validation criteria for example the questions might be who is the user of the system and end user application developer how complex are the you processing services how should the system function synchronous or asynchronous and what is the architecture used to design the processing workflow and what is expected number of users so for example for the 1st question um if I want to choose between these of the piscivores if the user our end users like non-expert users I would go with due degree which are easy to set up and has a and have a good performance but if they wanted them to the WPA server for Application Developer I would go with you because it's flexible enough in terms of developing languages and you can have multiple options for other questions as well that to choose a suitable for the best fit to EPS server for a specific application and in the end that the most appropriate WPS server should be selected based on a trade off between the quantitative performance metrics and ease of use qualitative metrics for a specific application or use case and this may lead you all the selection of different Olimpia service for different applications thanks for your
attention if your questions and without the the the by the way with the Apache HTTP verses Tonka reusing stock to packages for those and to do do any performance tuning on them and did you track uh if the background process was using the same amount of load every time verses whether was the web service in between getting the way so for the installation on confirmation the servers we just use the default settings we didn't tune DePaul stress at the back and we didn't and to the other service like multi-threading so we just install them and configure them to the to be working and wouldn't touch anything and it's that it's the same for all the servers and for example the of sort version that we use are here and all of them are the same for all the servers and I know you know if I answered the questions are not this person can Ch and we didn't have configure Apache auction that bite your tongue cancer we didn't touch anything yeah the Stiefel confirmation hello the effect on the alignment of set but or just want to point out about the recent version you server have for support for several scripting languages for that of of machine it so it's possible to write processes in each item j Adobe JavaScript that uh grooving shall uh either remember that all but I I mean it added a bit of a variety of them to their and to the set of languages that 1 can use 2nd school 1 is of clear questioning the uh so the scalability issue toward the end of the graph and in just over we have uh alleviated for the number of threads that cannot on a little Pierce processes to defend the hard work from uh overflowed and by default it's only soul but might affect it if you didn't unit but might have affected the uh that also supports 64th until on I just changed that product you the thousand it's while walking probability and much yet because that uh but it doesn't work with like 8 because at more than like 8 it morning a 8 concurrent because it will crash well we should start to of slowing down the sponsors him in good How was willing right you have to admit this metric about the amount of data returned to I wasn't mob more important if you did a character and was correct to to look at that it is the outcomes actually were correct or otherwise the amount of data is important because when you when use them on animal while or something with limited the bound it you need to have like the tiny WPS like kind useless to have a compressed candidate or with that an XML data of the the least amount of types like necessary tax and I think from that respect from all was competing perspectives can import collaborative if there's too little to exit if it will it will be an incorrect things a even look at that probably commons to that of the RLS ingredient by the Rupiahs we use pretext a multiple this could be this could make a significant difference which was the pool no took the question did you actually contact all the authors of the web processing services to talk to them about how they might like almost all the normal benchmarks he is the quote vendors which cases the authors basically help you to and they set to try to make like I'm telling you this is how we get the best performance ruckus for GeoServer it might be that you have to tune Tomcat for somebody like degree 52 nor the could've been yeah we need this many socket connections open above a power that needs the least this much memory if for the next round of this the be much more helpful if you the same the WMS should out in the tested out the contact the authors using here interview the test platform you Tommy attitude and our and I think that would be a much nicer way to do it actually I didn't like a couple of months ago and nobody uh replied because I'm doing this for like and indeed it while of this and 0 call the blonde and this is called the in this friend comparision like
Chandni
the book is the just
you on the the
the yeah I I think I think I think part of the gradient love of your neck yeah OK the the thank you
Resultante
Beobachtungsstudie
Server
Prozess <Informatik>
Leistungsbewertung
Relativitätstheorie
Kombinatorische Gruppentheorie
Computeranimation
Font
Framework <Informatik>
Server
Textverarbeitung
Leistungsbewertung
Telekommunikation
Web Site
Prozess <Physik>
Kartesische Koordinaten
Computeranimation
Task
Benutzerbeteiligung
Informationsmodellierung
Skalierbarkeit
Web Services
Algorithmus
Perspektive
Trennschärfe <Statistik>
Datenverarbeitung
Elektronischer Programmführer
Softwareentwickler
Dienstgüte
Schnittstelle
Beobachtungsstudie
Benutzerfreundlichkeit
Computersicherheit
Kanalkapazität
Strömungsrichtung
Rechnen
Biprodukt
Teilbarkeit
Arithmetisches Mittel
Einheit <Mathematik>
Softwarearchitektur
Server
Varietät <Mathematik>
Neuronales Netz
Instantiierung
Standardabweichung
Softwaretest
Umwandlungsenthalpie
Lineares Funktional
Softwareentwickler
Server
Prozess <Physik>
Prozess <Informatik>
Protokoll <Datenverarbeitungssystem>
Kartesische Koordinaten
Dienst <Informatik>
Computeranimation
Portscanner
Mapping <Computergraphik>
Physikalisches System
Benutzerbeteiligung
Softwaretest
Funktion <Mathematik>
Softwareschwachstelle
Server
Luenberger-Beobachter
Softwareentwickler
Leistungsbewertung
Schnittstelle
Server
Prozess <Physik>
Implementierung
Kartesische Koordinaten
Dienst <Informatik>
Framework <Informatik>
Computeranimation
W3C-Standard
Metadaten
Benutzerbeteiligung
Algorithmus
Web Services
Perspektive
Programmbibliothek
Softwareentwickler
Schnittstelle
Funktion <Mathematik>
Beobachtungsstudie
Programmiersprache
Internetworking
Parametersystem
Nichtlinearer Operator
Schnelltaste
Prozess <Informatik>
Open Source
Datenmodell
Applet
Wiederkehrender Zustand
Peer-to-Peer-Netz
Mailing-Liste
Marketinginformationssystem
Ein-Ausgabe
Quick-Sort
GRASS <Programm>
Spezialrechner
Minimalgrad
Server
Dateiformat
Information
Standardabweichung
Prozess <Physik>
Punkt
Extrempunkt
Gruppenoperation
Kartesische Koordinaten
Polygon
Räumliche Anordnung
Framework <Informatik>
Computeranimation
Spezielle Funktion
Client
Informationsmodellierung
Perspektive
Programmbibliothek
Flächeninhalt
Abstand
Speicher <Informatik>
Maßerweiterung
Touchscreen
Lineares Funktional
Kategorie <Mathematik>
Physikalisches System
Binder <Informatik>
Rechnen
Flächeninhalt
Automatische Indexierung
Rechter Winkel
Server
Projektive Ebene
Computerarchitektur
URL
Prozess <Physik>
Punkt
Gruppenoperation
Interaktives Fernsehen
Computeranimation
Informationsmodellierung
Client
Datenmanagement
Front-End <Software>
Gateway
Logistische Verteilung
A-posteriori-Wahrscheinlichkeit
Maschinelles Sehen
Sequenzdiagramm
Parametersystem
Benutzerfreundlichkeit
Physikalisches System
Ein-Ausgabe
Ordnungsreduktion
Summengleichung
Verkettung <Informatik>
Flächeninhalt
Server
Dualitätstheorie
Computerarchitektur
Instantiierung
Resultante
Prozess <Physik>
Punkt
Datenparallelität
Kartesische Koordinaten
Inzidenzalgebra
Ähnlichkeitsgeometrie
Computeranimation
Gebundener Zustand
Intel
Client
Softwaretest
Mustersprache
Funktion <Mathematik>
Softwaretest
Parametersystem
Kanal <Bildverarbeitung>
Systemaufruf
Ein-Ausgabe
Bitrate
Frequenz
Teilbarkeit
Menge
Funktion <Mathematik>
Automatische Indexierung
Zahlenbereich
Konditionszahl
Ein-Ausgabe
Client
Server
URL
Transitionssystem
Tabelle <Informatik>
Gruppenoperation
Virtuelle Maschine
Zahlenbereich
Polygon
Term
Virtuelle Maschine
Multiplikation
Zufallszahlen
Mittelwert
Endogene Variable
Stichprobenumfang
Virtuelle Realität
Response-Zeit
Touchscreen
Leistung <Physik>
Soundverarbeitung
Mailing-Liste
Frequenz
Endogene Variable
Abstand
Minimalgrad
Flächeninhalt
Last
Parametersystem
Lie-Gruppe
Resultante
Prozess <Physik>
Datenparallelität
Formale Sprache
Kartesische Koordinaten
Service provider
Computeranimation
Einheit <Mathematik>
Prozess <Informatik>
Mustersprache
E-Mail
Figurierte Zahl
Maschinelles Sehen
Benutzerfreundlichkeit
Ähnlichkeitsgeometrie
Ausnahmebehandlung
Bitrate
Wiki
Kontextbezogenes System
Ranking
Algorithmische Programmiersprache
Entscheidungstheorie
Teilmenge
Menge
Server
Ablöseblase
Ordnung <Mathematik>
Telekommunikation
Subtraktion
Server
Klasse <Mathematik>
Zahlenbereich
Term
Framework <Informatik>
Knotenmenge
Webforum
Perspektive
Endogene Variable
Programmbibliothek
Response-Zeit
Softwareentwickler
Konfigurationsraum
Programmiersprache
Relativitätstheorie
Mailing-Liste
Paarvergleich
Elektronische Publikation
Endogene Variable
Minimalgrad
Last
Eigentliche Abbildung
Resultante
Bit
Prozess <Physik>
Datenparallelität
Leistungsbewertung
Social Engineering <Sicherheit>
Formale Sprache
Versionsverwaltung
Orakel <Informatik>
Kartesische Koordinaten
Euler-Winkel
Synchronisierung
Computeranimation
Last
Web Services
Skalierbarkeit
TUNIS <Programm>
Einheit <Mathematik>
Trennschärfe <Statistik>
Default
Benchmark
Umwandlungsenthalpie
Softwaretest
Lineares Funktional
Prozess <Informatik>
Benutzerfreundlichkeit
Systemaufruf
Fitnessfunktion
Biprodukt
Teilbarkeit
Konfiguration <Informatik>
Konstante
Angewandte Physik
Funktion <Mathematik>
Menge
Rechter Winkel
Server
Socket
Bildschirmsymbol
Schlüsselverwaltung
Normalspannung
Fitnessfunktion
Varietät <Mathematik>
Maschinenschreiben
Server
Subtraktion
Zahlenbereich
Unrundheit
Systemplattform
Term
Spezifisches Volumen
Physikalisches System
Virtuelle Maschine
Multiplikation
Informationsmodellierung
Benutzerbeteiligung
Perspektive
Datentyp
Endogene Variable
Minimalgrad
Thread
Softwareentwickler
Konfigurationsraum
Leistung <Physik>
Leistungsbewertung
Einfach zusammenhängender Raum
Soundverarbeitung
Autorisierung
Architektur <Informatik>
Linienelement
Linienelement
Dämon <Informatik>
Validität
Paarvergleich
Physikalisches System
Quick-Sort
Endogene Variable
Modallogik
System F
Minimalgrad
Flächeninhalt
Komplex <Algebra>
Offene Menge
Last
Computerarchitektur
Orakel <Informatik>
Computeranimation
Mereologie
Orakel <Informatik>
Computeranimation
Gradient

Metadaten

Formale Metadaten

Titel Evaluation of Web Processing Service Frameworks
Serientitel FOSS4G 2014 Portland
Autor Poorazizi, Ebrahim
Hunter, Andrew J.S.
Lizenz CC-Namensnennung 3.0 Deutschland:
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 10.5446/31685
Herausgeber FOSS4G, Open Source Geospatial Foundation (OSGeo)
Erscheinungsjahr 2014
Sprache Englisch
Produzent Foss4G
Open Source Geospatial Foundation (OSGeo)
Produktionsjahr 2014
Produktionsort Portland, Oregon, United States of America

Inhaltliche Metadaten

Fachgebiet Informatik
Abstract In this presentation, I illustrate, and discuss initial results from a quantitative analysis of the performance of WPS servers. To do so, two test scenarios were used to measure response time, response size, throughput, and failure rate of five WPS servers including 52¡North, Deegree, GeoServer, PyWPS, and Zoo. I also assess each WPS server in terms of qualitative metrics such as software architecture, perceived ease of use, flexibility of deployment, and quality of documentation. A case study addressing accessibility assessment is used to evaluate the relative advantages and disadvantages of each implementation, and point to challenges experienced while working with these WPS servers.
Schlagwörter OGC WPS
Geoprocessing
Performance Evaluation
Benchmark

Ähnliche Filme

Loading...
Feedback