From paper to pods: Revolutionised fibre planning process at Deutsche Telekom AG with FOSS4G components
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
| Title |
| |
| Title of Series | ||
| Number of Parts | 295 | |
| Author | ||
| Contributors | ||
| License | CC Attribution 3.0 Germany: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
| Identifiers | 10.5446/43299 (DOI) | |
| Publisher | ||
| Release Date | ||
| Language |
Content Metadata
| Subject Area | ||
| Genre | ||
| Abstract |
| |
| Keywords |
FOSS4G Bucharest 2019284 / 295
15
20
28
32
37
38
39
40
41
42
43
44
46
48
52
54
57
69
72
75
83
85
87
88
101
103
105
106
108
111
114
119
122
123
126
129
130
131
132
137
139
140
141
142
143
144
147
148
149
155
157
159
163
166
170
171
179
189
191
192
193
194
195
196
197
202
207
212
213
214
215
216
231
235
251
252
263
287
00:00
PlanningFiber (mathematics)Connectivity (graph theory)Process (computing)Estimation
01:06
Process (computing)Content (media)Computer animation
01:38
Factory (trading post)Open sourceRankingInternetworkingFiber (mathematics)Internet service providerPlanning
03:44
Fiber (mathematics)MIDIMultiplication signFiber (mathematics)PlanningProjective planeCartesian coordinate systemBitProduct (business)Process (computing)AreaComputer animation
04:42
AutomationProcess (computing)PurchasingDigital signalPlanningMeasurementFactory (trading post)System identificationObject (grammar)WindowBeat (acoustics)Information managementProcess (computing)InformationPlanningPurchasingGraph (mathematics)Physical systemConstructor (object-oriented programming)MultilaterationMereologyDeutscher FilmpreisMeasurementUltimatum gameDigital printing
05:56
PlanningMeasurementFactory (trading post)Process (computing)PurchasingDigital signalAutomationService (economics)Parameter (computer programming)InformationPlanningMereologyGeometryComputer animation
06:24
Service (economics)Data managementCartesian coordinate systemService (economics)Level (video gaming)Web 2.0User interfaceUniform resource locatorProcess (computing)Descriptive statisticsDifferent (Kate Ryan album)SurgeryProgram flowchart
07:04
Water vaporSurfacePlanningBuildingForestField (computer science)InformationArtificial neural networkAcoustic shadowPlanningSurfaceGraph (mathematics)Open setAreaType theoryDigitizingGrass (card game)Different (Kate Ryan album)State of matterDigital photographyOpticsPoint (geometry)Single-precision floating-point formatBitProcess (computing)Theory of everythingService (economics)Row (database)Extension (kinesiology)CausalityFiber (mathematics)Data centerOpen sourceComputer animation
09:49
Network topologySoftwareLine (geometry)Process (computing)
10:10
PlanningCellular automatonArtificial neural networkFactory (trading post)Vector potentialSurfaceCalculationDatabaseImpulse responseElectric generatorExtension (kinesiology)SoftwareDigital photographyAreaSurfaceGraph (mathematics)MereologyInformationRule of inferenceOpen setLevel (video gaming)Artificial neural networkConnectivity (graph theory)Process (computing)DataflowOcean currentLengthOpen sourceDatabaseDigitizingCASE <Informatik>Vector potentialPhysical systemSet (mathematics)Computer animationProgram flowchart
12:39
Factory (trading post)CalculationVector potentialSurfaceWeb 2.0Position operatorInterface (computing)Dot productMereologyOffice suiteCartesian coordinate systemGraph (mathematics)Level (video gaming)Computing platformService (economics)Internet service providerSoftware frameworkMedical imagingData managementProcess (computing)User interfaceComputer animationProgram flowchart
13:40
WebsiteProcess (computing)Interactive televisionSurfaceClient (computing)Fiber (mathematics)PlastikkarteUniform resource locatorLevel (video gaming)Function (mathematics)PlanningGoogle Street ViewPosition operatorForm (programming)Interface (computing)MetreData exchangeGezeitenkraftMereologyComputer animation
14:45
Interface (computing)Direction (geometry)Heat transferFile formatData exchangeLevel (video gaming)Computer animation
15:04
CodePoint cloudOpen setLoginBeat (acoustics)Windows RegistryMereologyPoint cloudFood energyBitProcess (computing)ChainOpen setComputer configurationMedical imagingCartesian coordinate systemGeometryShift operatorCodeTelecommunicationComputer animationDiagramProgram flowchart
16:08
Computer-generated imageryDisintegrationRepository (publishing)Windows RegistryCodeCompilerSoftware testingExecution unitConfiguration spaceBuildingInterior (topology)ChainLevel (video gaming)CodeMereologySoftware developerMedical imagingWindows RegistryUnit testingIntegrated development environmentBitMathematicsOperator (mathematics)Normal (geometry)MUDDifferent (Kate Ryan album)INTEGRALTelecommunicationProduct (business)Computer animation
17:08
CodeCompilerExecution unitComputer-generated imagerySystem programmingIntegrated development environmentBitGraph (mathematics)Ocean currentDifferent (Kate Ryan album)Complex (psychology)Data structurePhysical systemQuicksortOperator (mathematics)TelecommunicationAsynchronous Transfer ModeProjective planeLatent heatMereologyCartesian coordinate systemOverhead (computing)Medical imagingExterior algebraEscape characterINTEGRALEnterprise architectureMultiplication signForcing (mathematics)Software testingExtension (kinesiology)Computer animation
19:09
System programmingComa BerenicesMobile WebComputer animationXMLUML
Transcript: English(auto-generated)
03:23
tenant coverage in Germany and that we provide a faster internet to households and industries. And, um, yeah, so therefore we need to really change the process and speed up all, um, um, the planning work, which is involved with that. Um,
03:45
additionally, or which, uh, uh, underlines this from a board member from Deutsche Telekomage, Dirk Wesner, he stated in public in 2018 that with the beginning of 2021 we will provide, um, FTTH, so fiber to the home, um, for up to two millions per year.
04:04
Um, the project, uh, started mid of 2017 and one may say, uh, okay, that's still a long time until 2021 but that's not really true because if you would like to start building and do the construction work in 2021,
04:24
all the planning has to be done in 2020 which again means that all the it work and all the applications and the processes and tools, they must be ready by the end of this year. So there's not that much time. Um, um, that has not been that much time for the project and all the work we have
04:41
done. So what are we doing? Um, we identified actually certain areas, um, where we can improve, um, uh, the planning process and um, there are certain leverages and, and then measurements, uh,
05:01
which we took and um, we need a lot of data in order to speed up the process and, and to automate a lot of things. We need a 2d and even 3d data, um, to do this. And I will come to this later. Um, with some more details, um, we are having problem with the existing data from Deutsche Telekom.
05:23
So the existing fixed landline, uh, data and Deutsche Telekom needs a correction and confirmation. There's also a Graham here in the audience who's working on that part. And uh, yeah, one goal is to, to automate a lot of processes. So automated planning, and then we have two more,
05:42
it's a simplified purchase process on not covering this. And also a digital plan correction. So by the end, what was really built by the construction teams outside this information has to come back, um, again into the systems. Um, as you can imagine, there's a lot of geo information and geo data involved in this, especially,
06:04
uh, yeah, in the first three parts. And I'm going to focus on the 2d and 3d data collection and also into the automated planning part. Um, yeah, for doing this and for providing the data, um,
06:22
and for also crunching and processing the data, um, we build up a spatial data infrastructure, which is, um, yeah, coming from the description here as doing more or less, uh, what you would expect from a spatial data infrastructure. So we're providing different, uh, kind of services. Of course,
06:40
we have map services with, uh, search services, location services, and we also providing a different applications to the end user. So web GIS, uh, applications, but also management and workflow, um, user interfaces. And of course we're also providing data services and processing services.
07:01
And again, I will focus on those two. Okay. So, um, if we look at the data, what we are taking into this, um, planning process, um, um, yeah, we're, we're taking it from different sources. And one really very important point is that we are getting information about
07:23
the surface types. So why do we need information from the different surfaces? There was also another talk already, uh, this morning, um, by Marcos Nutella who explained a little bit more detailed about this, but we need the surfaces because we need to find or would like to find the
07:42
cheapest possible ways for the trenches. So we would like to find the cheapest way where to put the optical cables and in order to do so we need to have the information from the surfaces. You can imagine that different surfaces, uh, or depending on the surface type, different costs are involved when you would like to put the cable inside there.
08:03
So of course it's cheaper to put the cable in an open land or in grass area than it is, uh, to do this on asphalt or concrete. So we're trying to detect the different surface types. Um, on the one hand you see on the left, um, um,
08:20
we're doing this with a terrestrial LIDAR data, which we are collecting by ourselves, um, with so-called T surface cars. We are collecting LIDAR data and panoramic pictures and also normal pictures. And from this we're generating, um, by the use of an artificial neural network, um, different surface types.
08:43
But the car of course is not, not running around everywhere, not in every single street and small streets. And to some extent, uh, in the LIDAR, um, cannot really record, um, everything because there's a shadow or there are parking cars and, and, uh,
09:01
we don't get information. So we need to supplement this information in order to get the surfaces for the whole area where we're going to, um, to do the planning for the fiber, um, rollout. We're doing this, um, by making use of, um, aerial, um, um, pictures. So digital auto photos and, um, which is again,
09:23
uh, supplemented with LIDAR data when it's available in Germany. Some federal states which have open data policies and we're making use of this. And, um, by making use of this, we're getting as well surfaces for the, for the area. The information is not that detailed as we do get it from the surface car.
09:42
So from the terrestrial LIDAR data, but it's, uh, we are good enough to, to fill the holes and, um, to get the information. And, um, yeah, we're then extracting by a rather complex process which Marcos Netera just explained this morning, um, to extract a topological line network with a potential trenches.
10:04
So where potentially we can put, um, the optical fiber cables. Yeah, this is just again mentioning the talk from this morning. The original idea was that this talk would have been after this one, but this didn't work out. But of course if you're interested to,
10:21
you can still watch it. So this is a very simplified, um, yeah, workflow and to some extent, uh, also showing, uh, the different components which are involved into the process. So the first part, um, you see the incoming data. So on the one hand,
10:43
the terrestrial LIDAR data from the T surface car, um, we're having the auto photos and, um, we're having this artificial neuron network where we are, um, classifying the surface data and the pictures with, and also using actinia to generate, um,
11:01
the surfaces from the digital auto photos. And, um, yeah, it's a hybrid system. It's not only a open source software, which we are using. Uh, we also making use of a FME feature manipulation engine. And in this case of what this workflow, it's, uh, yeah, doing a lot of orchestration of, of all this, uh, workflow. Um,
11:23
so it's always triggering, uh, the different steps and, um, getting the status from all the different processes here. So in the end we're having classified surfaces. And then again, we're making use of actinia for a different process to generate, uh, the potential trenches, um,
11:42
which are in the end stored in the database. Um, for the generation of the trenches, we have a really large set of rules. Um, so for example, um, there are methodologies, um, for the trenches which cannot be applied and, and, uh, in urban areas and cities. Um,
12:02
there are methodologies, um, which need to have a certain length, so that it makes sense to apply them. And, um, there are a lot of more rules, rather complex rules. Um, these are all handled, um, during the process of generating the potential trenches. And, um, we are also making use of catastrophe,
12:21
cadastral data and open street map data, um, in order to get more information and, um, to really apply those rules. Yes. So the first part, so from, from all what you see on the left to storing the data and, uh, and the database that's all automated. Um, the only manual part actually is
12:42
to upload the, um, the data from the surface car. And then, uh, once you can start the process and everything else is handled automatically, which is already a huge step forward. Um, then we have the part with our user interfaces.
13:01
Um, we have different, uh, uh, as I said, uh, web GIS, uh, platforms and we have management platforms. We also have a three D client, which I will, uh, will show you. And for this we're making use of two frameworks. It's two orchestra and a sugar moon. Um, we're using here, uh,
13:21
in order to provide interfaces to the users and provide services for other applications as well. So this is an example of the web GIS and a linkage of a two D and a three D application. So on the left side you see a normal, uh,
13:41
background map and the dots, uh, uh, showing the positions where the surface card took images. So panorama pictures and we have a linkage between the two D and a three D or a client client is actually showing here a panorama pictures. Um, it's kind of imagine it's like a Google street view.
14:02
You can move around in the pictures and uh, we're using this, um, for, for the planning of the cabinets. So the, um, yeah, great cabinets you see, uh, on the, on the side of the streets where we built, um, yeah, which we need to, to, to provide a fiber to the home.
14:22
So, and in the end here, this is the current status. Uh, you could say we're using pots to generate paper again. There's an output from, uh, from paper, which is then, then going to the communities because we need to get the approval for the, all, all the locations we have planned. That's, I hope at least,
14:41
uh, just an intermediate step. So we would like to get, uh, the communities either to work with our interfaces or to have at least a data exchange format where we can, um, um, transfer the data directly. There are already a discussions going on with a, um, for a pilot with the federal stage, um, of Hessen.
15:05
So how are we working? I'm not going to explain all of this picture would be a talk by itself. Um, but uh, yeah, we're using the open telecom cloud here, which is providing an open stack API. And, um,
15:24
as you can see, there are a lot of tools and processes going on, which are not related to any geo processing and geo data handling at all. It's all about setting up our infrastructure. So the first part, at least, um, we're using GitLab and, um, um, infrastructure as code to create our VMs and to provision our VMs and to
15:44
install the applications. And we're using a CICD chain to generate the images and to install the applications and, um, in the cloud and to put the images into the pots of the open shift cluster. And of course we also have a monitoring in place, uh,
16:02
using Grafana and for logging Kibana. Yeah. Um, this is a little bit more detail about the CICD chain. Um, most important part is, uh, that it's really separated between deaths and ops. Um,
16:21
although the people which are doing this, uh, I mean they're all in, working in a team and our team. And um, as you can see, we have certain environments, we have a development environment, integration, staging and production, and we need to bring all the code changes and the developments into the different environments. So the developer, once he checked his code in,
16:44
then, uh, um, the chain is starting, um, to, um, build the image and we're doing automated unit tests as well. Before we built the image and we then put the image into the Magenta trusted registry, which is just the telecom, um, yeah,
17:01
telecom registry. We can compare it to the normal Docker registry. Um, so this, uh, um, this work of the developer by itself to bring everything to the development environment. And at least that's, that's our goal. Currently we handle a little bit differently, but the goal is that coming from this, uh, the operations team is then bringing, um,
17:23
the same image to the different environments. So we can really be sure, um, that the exact, uh, same status is reached in all the different environments. Um, and the workflow is more or less the same. Um,
17:41
also automated to some extent, um, to see if the applications are working correctly and, uh, interacting correctly together. Yeah, of course this is a really huge topic and a complex project and we need a lot of, uh, um, people to work on this and we can't handle it by ourselves.
18:04
That's telecom it or at your telecom. Okay. Okay. Um, there's also a lot of specific knowledge involved which we need to get externally. And here you can see, uh, the companies which are all involved here. There are different parts from Deutsche telecom, um, this systems to systems sort of back here.
18:21
So international team here and also different companies, uh, which some of them, when you're all is the rest has came to camp. They're also present during the phosphor G. We're working, uh, in an agile mode. Um, so we are applying scrum and then working in two weeks, uh, cycles, uh, in order to develop everything.
18:42
And recently starting last year on top of this, we also implemented or make use of safe, which is scaled agile for enterprises, um, which is producing some overhead I would say. But on the other hand it gives structure and helps us to, to move forward.
19:02
Yeah. There was also already a talk, uh, yesterday by Emmanuel Belo from camp to camp. Um, we explained a little bit about the way, how we are interacting and, and, uh, yeah, how we're working together. So with this, I would like to thank you and yeah, I'm happy to hear questions if there are any.