The Architecture Of Mobile Traffic Map Service
This is a modal window.
Das Video konnte nicht geladen werden, da entweder ein Server- oder Netzwerkfehler auftrat oder das Format nicht unterstützt wird.
Formale Metadaten
| Titel |
| |
| Serientitel | ||
| Anzahl der Teile | 95 | |
| Autor | ||
| Lizenz | CC-Namensnennung - keine kommerzielle Nutzung - Weitergabe unter gleichen Bedingungen 3.0 Unported: Sie dürfen das Werk bzw. den Inhalt zu jedem legalen und nicht-kommerziellen 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 und das Werk bzw. diesen Inhalt auch in veränderter Form nur unter den Bedingungen dieser Lizenz weitergeben | |
| Identifikatoren | 10.5446/15579 (DOI) | |
| Herausgeber | ||
| Erscheinungsjahr | ||
| Sprache | ||
| Produktionsort | Nottingham |
Inhaltliche Metadaten
| Fachgebiet | ||
| Genre | ||
| Abstract |
|
FOSS4G Nottingham 201378 / 95
17
25
29
31
32
34
48
50
56
58
68
69
70
82
89
91
00:00
Mobiles InternetInformationTotal <Mathematik>Dienst <Informatik>MatrizenrechnungARM <Computerarchitektur>Lokales MinimumZeiger <Informatik>PolygonnetzSelbst organisierendes SystemInhalt <Mathematik>Prozess <Informatik>DatenflussMessage-PassingAbstandProgrammierumgebungPhysikalisches SystemMultiplikationsoperatorFamilie <Mathematik>PlastikkarteMAPInteraktives FernsehenZoomPunktOpen SourceCachingSpezialrechnerArchitektur <Informatik>Kanal <Bildverarbeitung>ServerMathematikSQL ServerEDV-BeratungProxy ServerClientBildgebendes VerfahrenDatenbankInformationOrdnung <Mathematik>ComputerarchitekturSelbst organisierendes SystemValiditätTypentheorieMAPSensitivitätsanalyseEchtzeitsystemElementargeometrieÄquivalenzklasseBenchmarkDispersion <Welle>Inhalt <Mathematik>Physikalisches SystemProjektive EbeneGüte der AnpassungFamilie <Mathematik>ServerMigration <Informatik>RuhmasseEigentliche AbbildungVerkehrsinformationGraphfärbungClientOpen SourceWeb SiteKonditionszahlProxy ServerMultiplikationsoperatorURLMapping <Computergraphik>TesselationCachingSmartphoneInterface <Schaltung>Mobiles InternetMathematikMathematische LogikFrequenzBildschirmfensterAggregatzustandDelisches ProblemEinflussgrößeAbstandRoutingWort <Informatik>BeobachtungsstudieSchreib-Lese-KopfDomain <Netzwerk>SondierungFreewareWald <Graphentheorie>Office-PaketComputeranimation
05:52
Architektur <Informatik>Wechselseitige InformationWechselsprungWeitverkehrsnetzBootenOffene MengeNetzwerk-gebundene SpeicherungDreiMAPDatenflussMehrwertnetzServerLokales MinimumBildschirmfensterKanal <Bildverarbeitung>DatenstrukturServerCoxeter-GruppeWeb-SeiteWeg <Topologie>MultiplikationsoperatorDatenbankComputerarchitekturFunktion <Mathematik>MAPElementargeometrieErneuerungstheorieClientTesselationCachingMobiles InternetApp <Programm>Computeranimation
07:02
Physikalisches SystemMultiplikationsoperatorMAPExogene VariableTransaktionStrategisches SpielSkalierbarkeitServerCachingLastROM <Informatik>Mini-DiscTropfenOrdnungsreduktionSQL ServerAbfrageGesetz <Physik>Streaming <Kommunikationstechnik>DatenreplikationElektronischer FingerabdruckGleitendes MittelWiederkehrender ZustandLokales MinimumFahne <Mathematik>Mobiles InternetMathematikKonditionszahlDatenbankDatenverwaltungDynamisches SystemMathematikHalbleiterspeicherTypentheorieSoftwaretestFunktionalMereologiePhysikalisches SystemZahlenbereichServerResponse-ZeitCASE <Informatik>Prozess <Informatik>Strategisches SpielMetropolitan area networkLuenberger-BeobachterWeb-SeiteAuswahlverfahrenDateiformatVerkehrsinformationElektronische PublikationClientSichtenkonzeptSchlüsselverwaltungSondierungMultiplikationsoperatorMapping <Computergraphik>GraphiktablettCachingBenutzerbeteiligungSoftwareentwicklerDatensatzFolge <Mathematik>ComputerarchitekturMAPGenerator <Informatik>SynchronisierungTotal <Mathematik>Einfach zusammenhängender RaumRechenschieberAbfrageKonfigurationsraumZoomMini-DiscProxy ServerTesselationApp <Programm>Computeranimation
13:48
Dienst <Informatik>Total <Mathematik>MAPSkalierbarkeitComputerarchitekturMAPGenerator <Informatik>Physikalisches SystemResultanteServerVerkehrsinformationMultiplikationsoperatorZweiMobiles InternetSoftwareentwicklerWasserdampftafelSondierungComputeranimation
14:53
Open SourceBereichsschätzungProdukt <Mathematik>ServerMobiles InternetE-MailCOMEmulationBereichsschätzungServerCASE <Informatik>MultiplikationsoperatorOffice-PaketProdukt <Mathematik>Projektive EbeneTermE-MailOpen SourceMobiles InternetSoftwareentwicklerComputeranimation
Transkript: Englisch(automatisch erzeugt)
00:00
It is a kind of mobile app, and the Mobile Traffic Map Service is one of these apps. Our background is here. For your information, NTIC is National Transport Information Center. It is a national organization.
00:22
They are collecting all the traffic information in South Korea and synthesizing the contents, and they are providing information to people. The background of this project is this.
00:41
NTIC's requirement is very simple. They want to deliver real-time traffic information to users to disperse traffic on major national holidays. For example, lunatic New Year's Day, New Year's First Day, and Chuseok. Chuseok is the equivalent of Thanksgiving Day.
01:02
At those holiday seasons in Korea, about 30 million people visit their hometowns and families. That amount is nearly 60% of the population of South Korea. The mass migration causes people to be interested in the traffic conditions.
01:25
These days, most people have a smartphone, so they want to get more information through our smartphone.
01:42
Briefly, the main features of the Mobile Traffic Map Service are like this. It supports interactive zooming in and out, and three steps colorize the traffic data. The green color means free traffic, and the red color means congested.
02:03
The reference speed is applied by road types. And in addition, it supports traffic accident information and sensitive on the roads.
02:21
This is the architecture of 2011. Actually, my company wasn't the one who had designed this system. There was a company who originally designed it, but there was a lot of problems. As you can see, it is a Windows-based system, and there are six Geo servers as a map server,
02:50
and one PostGIS Postgres SQL as a database. And there is no cache server, and the client requests a one-size image and a non-tiled.
03:05
So there are lots of problems. So in 2011, they launched this service, so there were more users than they expected.
03:23
And because of the absence of a cache server, the same location information was reportedly requested to the server. And the Geo server was so frequently down at peak times.
03:42
So because of these problems, our customer NTIC suspected the performance of the open source Geo server or PostGIS. So, NTIC is required to us to prepare TrueSoft in September 2012.
04:10
They requested system improvement. So their requirement is they want to support 200,000 users per day,
04:22
and they want to change the database to an SQL server they already had. Because at that time, there are not proper good benchmark sites using open source GIS at South Korea,
04:41
so we failed to persuade the customer. So we suggested solutions like that. On top of that, we insisted the system architecture should be redesigned and the system regarding maps should be redeveloped.
05:08
And in second, we proposed that all map requests on mobile sites should be tiled. In third, a SCID proxy server should be used as a cache server for improving the reusability and performance.
05:25
And first, in order to call out valid traffic information, valid time for cache tile should be determined. And we used WMTS interface and content expired time and custom time tag.
05:41
And last, we suggested making all the tile map data in advance every five minutes. Here is the new architecture in 2012. It consists of three SQL server as database and six Geo server.
06:07
Geo servers are making 256 by 256 tiled map in this architecture. And there are two cache server in front of them. And we put cache maker at the cache server side to produce the traffic map data tiles in advance.
06:31
And the cache maker uses round-robin dispersion principle of output switch to produce tiles. And as we redeveloped the client's mobile apps and mobile web page,
06:49
we also redesigned it to automatically renew traffic map every five minutes and apply open layers cache structure into mobile apps.
07:05
During the Joseph Halliday season in 2012, my colleague Bizhe Zhang was being stood by to monitor this system at NTIC. Reportedly, about one million people downloaded mobile app only during that season.
07:24
As a result, the service was somewhat successfully stable. It was much better than previous system. But frankly speaking, as a developer, it was not satisfied. Mobile app requested tile map every five minutes.
07:43
So unnecessary requests had been increased. And the response time consequently got longer. As a traffic tile, comprised of 10 June rebels, so it had exceeded over one million tiles.
08:03
All the tile map in every level couldn't be updated in advance within five minutes. It's a problem. So to solve this problem, temporarily for nine, ten levels, when clients request to let your server make traffic map data dynamically,
08:24
therefore the connection time got longer and the number of connections increased. And also in this architecture has a scaleability issue.
08:44
The cache server can't share the data. So for processing increased requests, cache server should be added here. But adding more cache servers under this architecture needs more map server
09:02
and according to SQL server. So it causes cost to issue. And then more cache server means more requests from the cache maker at the cache server side.
09:23
So it rose the SQL server n times. It's a burden of this system. So we suggested the new strategy to our customer.
09:45
We persuaded our customer to use PostGIS again, because which provides faster, special query function, we think. And we designed GeoServer Connect PostGIS one by one instead of SQL server.
10:06
And actually the total size of traffic map data, tile map data in whole ten levels was less than four gigabytes. So meaning that data size was not that big.
10:20
So to reduce the cost, we need to speed up data access process. So when adding cache server, we chose memory disk instead of SSD. Furthermore, we changed the system to push tile traffic map data into cache server.
10:48
This slide shows new architecture this year. As you can see in the middle, we physically combined map server and database server.
11:07
And also system generates CSV file there and import PostGIS. And to sync data, we replicate data in Postgres to another Postgres sequence.
11:23
And instead of cache maker at cache server side, we put the tile generation manager at the map and GeoDB server side.
11:45
The tile generation manager is like this. Tile generation managers divide the jobs for each GeoServer clearly to produce web tile data in parallel. And as soon as the jobs finish, they push tile traffic map data into cache server.
12:08
So in case of increasing cache server, the map and GeoDB servers only have a role to produce tile traffic map data.
12:25
And the request of the mobile apps can be handled only by cache server side and don't have influence on web and GeoDB servers. So in case of increasing connections,
12:42
the system can be serviced by just adding cache server here and with some configuration with the tile generation manager. So this architecture is more scalable.
13:01
And we have other ideas to reduce the total amount of data periodically to be updated. The idea is very simple because since the changing data containing traffic information, it would be fine if even if only tiles on passing rows,
13:24
tiles on passing rows is updated. And it also could be fine even if only tiles of changing the traffic information is updated. The zoom level, as a zoom level, this idea is more effective to update data.
13:48
This chart shows the result of improved system compared to 2012 and this year.
14:01
As you can see, the time to update interval of time generation has been shortened to one minute for whole ten levels. Actually, the developer said to me, it takes only ten seconds to update the traffic information.
14:27
And this service is now serviced. This week during FOSS4G conference is a Chuseok holiday in South Korea now. I was told that the mobile traffic information system built by new architecture has been smoothly serviced now.
14:56
So through this project, we came to make sure that open source GIS is definitely equivalent to commercial products in terms of performance now.
15:09
And we got confidence to persuade the customers into adopting open source GIS with no worries. I think this project is a good case adopting open source GIS, especially by public institution sector in South Korea.
15:31
There is a lot of case in South Korea to adopt open source GIS. With this experience, B.J. Jang, he is the original developer.
15:50
He is constructing mobile weather chart service using GeoServer and Postgres SQL at KMA. These are screenshots, the weather chart.
16:02
KMA is a kind of med office in England. Actually, I was not a member of this project. I don't know the technical details.
16:21
So if you have any questions, please ask B.J. Jang via email. Thank you.