 Good morning, everyone. Very early day fourth session, the start from my NTT data, the title Cloud Platform for IoT, designing, evaluating, large scale data, collecting, and storing platform. My name is Hiroshi Miura, and I come from Japan. And I'm happy to join the OpenStack Summit in Austin. I've been presented this similar title on the Tokyo Summit. And I'd like to share our experience in this half-year's progress. As mentioned in keynote in the Jonathan's one, the OpenStack is coming to be the new stage. And just on the virtual machines, orchestration of the cloud, not only so. The next OpenStack is a big software. And we support the diversity of IT systems. Today, I will share our trial to implement IoT project, IoT platform on the project, and focus on the problem we are faced on the process on our IoT project. This is every time the exclaimers. This is for informative presentation. So about us, who we are. I'm Hiroshi Miura, the project leader, and directing the OpenStack to new IoT platform. This is a memory, Naoto. Naoto is a chief architect on this project and leading the evaluation and the designing of the platform. And Yuji, Yuji is not coming here, but he's a great engineer for implementing everything for his good IT skills. We are also a professional sector in the entity data cooperation. In our floor, there are many OpenStack guides. And I'm centric on the OpenStack shift to implementing the cloud storage of the mobile carrier's mail system or the carrier's OpenStack system. And next to our team, we have the OpenStack Hadoop team. And they have the deep knowledge of the OpenStack Hadoop spark and the young, et cetera. Our main target in this project is cloud technology and accessory for OpenStack and OpenStack shift, SheepDoc, and recently challenging on the talker. And also, we are interested in the automation of the IT platform. As I share our experience in the Kirin delivery companies enterprise system in the last Tokyo Summit, they use OpenStack. And we are helped to deploy the OpenStack with automation system. Then today, I share that challenge for the IoT. In this presentation, the agenda is these three. At first, an overview of the project. And the next, IoT platform we designed is described. And last, the evaluation and IoT platform on OpenStack. As you may know, IoT is buzzing in the market. And everything is connected to the internet. In our vision, everything is connected to the net, but it is sometimes not the internet or the private network, but everything become connected. Almost computers are connected to the internet in nowadays and also smart devices. And then, recent days, many consumer electronics is connected to the internet. And many consumer electronics vendors released a new feature, a connection in the home electronics. Next five or 10 years, a connected vehicle, a connected car, will be become. And this is a big impact for the IoT systems. And recent day, construction machines like Komatsu is very famous for the success story of the IoT system. IoT enables data gathering, covering over the internet. Then, many business users think, oh, let's collect everything, then we can earn more money, we can provide a more good service. But they have a good big issue for realizing IoT platform. Everyone want to do big data analysis with IoT, but gathering data from everything is not easy work. Engineers know, but business person sometimes forgets this. So we are trying to do this issue in our platform. We are already implementing the IoT platform for Azure ASP style, cloud service style for the sensor network. We released the public service on October last year and connect the many sensors for the water processing companies. We use the technology for this sensor network to using OpenStack Swift and Amazon S3 style objects storage and implement on the hybrid cloud, public and private hybrid cloud. Also, last year, buzzing the new keyword industry from Germany and Europe and worldwide every year, the robot in the smart factory generate many, many data for the system. This is a big challenge for us to gathering such a big, big, big data to collecting, gathering, storing the system. We are already experienced in the some limited impact social influence system like a vendor machine connection or a smart factory, some small POC of smart factory or sensor networks. And last year, we helped some utility company, a power or gas company to realize a smart meter for the Japan market. Today, we try to realize the huge impact to social influence system like a smart city, smart grid and connected vehicle. This system would be dealing with so many huge data. So many cars is connected. There are billions of car in the worldwide. It exists now, and that will be depressed to the connected version car vehicles. Every year, millions of car newly connected to the system. This is our target, and we are trying to realize this demand on our system. I tell you just now, there are four key elements on the I4 IoT service. Of course, there are most important things in the business applications to provide value added service for the customer and the users. But to realize the platform system, we need the IoT platform to implement data receiver, data lake, data store, and data analysis platform. Data lake is a new keyword for the storing data like storage, HDFS, Hadoop. But this is a concept of every data stored in some areas of the lake, and then move the necessary data to analyze the platform, then analyze it. So data lake is a store place of everything data on here. To our target and potential customers, the demand is a very huge one. For connection, one million concurrent connection should be dealt in the platform. For the performance, 10,000 or over transaction plus seconds and seconds should be treated in the data receiver. And very surprisingly, 100 petabyte pass over in a month. This would become in the next three or five years in the smart grid of a connected vehicle is released on the consumer market. Our goal is full-stack IoT platform on using the open stack. From data center, lake, network, to upper is middleware, connecting software, and application, and data analysis, and everything, using open stack and related open source softwares. OK, then I change the talk from now to the detail of the platform. Hi, I'm Naoto from NTT Data. So OK, let me talk about IoT platform in the technology perspective. And before talking about IoT platform, I would like to mention about IoT data. In general, IoT data has three key features, velocity, volume, and variety. And as Hiroshi mentioned, the scale of our target is a mission-critical system. And the performance demands are like this one. For example, IoT platform has two process. 10 gig BPS has to establish one mega concurrently connections, has to store 100 petabytes per data, and has to support 100 data formats, and so on. Then let me describe our goal on the next slide. So these are our goal. We have defined three goals along two three key measures, which are based on general demands in our target. The first goal is one mega concurrently connections. IoT platform has to establish MQTT connections between devices at the same time. The second goal is 50-quart transaction per sec. The platform has also received and processed a bunch of MQTT messages without packet losing. It's a big challenge. The last goal is 5 gigabyte per sec. It has the capability of data storage, which can be stored 5 gigabyte per sec, data to itself. We have developed and evaluated IoT platform on our testing environment last fiscal year. From the next slides, let me talk about our testing environment. We have evaluated IoT platform on bare metro and cloud testbed. Objective testing on bare metro testbed is to do hardware benchmarking on physical servers. On the other hand, testing on cloud testbed is to do scale benchmarking on the cloud. Ideally, we should construct and test the IoT platform on full testing environment to achieve our target. However, we have tested on minimal physical and cloud environment like this one due to time and resource limitation, unfortunately. And we have actually chose AWS, which is comparable as OpenStack in this testing. So this is a detail of cloud instances and hardware devices for testing. We have estimated the number of cloud instances and hardware devices based on the result of our hardware sizing. The number of key points of our testing environment is three. The first one is 10 gigabit ether network. The second point is huge memory. The third point is many number of hard disk drives. In our use cases, IoT platform has to process around 50-kilometer messages per stack. That's why it would be a long 10 gigabit ethernet. It's a huge, heavy traffic. And this is a component of our IoT platform, which can be blacked down mainly two elements, data receiver and data like. First, data receiver has three components, connection gateway, worker, and receiver. The second, data like, has also components, cache, converter, and accumulator, and data store. Today, unfortunately, we can't share any testing results on some NDA. But from this slide, we would like to share the knowledge and experience about IoT platform, how to realize IoT platform on OpenStack. Before I'm speaking about how to realize the IoT platform on OpenStack, I'm asking to you guys, IoT is just an use case on OpenStack? What is an OpenStack? OpenStack is just a cloud software. Partly yes. What is IoT? IoT is just application, completely no. We believe IoT must be an infrastructure to support future internet. So OpenStack supports parts of the future infrastructure, right? We have seriously considered how to realize IoT platform powered by OpenStack. So today, we want to share the technical results requests for OpenStack and to do items of OpenStack from next slide. OK, let me share technical requests for OpenStack from our experience. We have five requests. Let's talk about them one by one from next slide. The first one, as we mentioned, we are challenging to realize IoT platform for mission-critical earlier. Some of the characteristics of mission-critical system of our target are high volume and heavy traffic. In our use case, IoT platform has to process around 50 kilo messages per second. As I mentioned, just a while later would be a long 10 gigabit ethernet. So the number of challenges for IoT platform and infrastructure learning IoT platform is two. The one is to receive messages from data source certainly in spite of heavy traffic like DDoS attack. The other one is to store messages to data store laboratory. So we'd like to ask the infrastructure OpenStack with learning IoT platform to be supported IO guaranteed about network and storage. This is first request for OpenStack. The second request is feature of messaging blocking. As you know, OpenStack has already supported messages queue with Decker, but we are happy if OpenStack supports not only support message queue, but also message broker because perhaps based on synchronous IO and feature of retraining tasks if workers fail are powerful to realize high scalability and flexibility. Moreover, we are happy, very happy if the message broker of future OpenStack supports multiple protocol for IoT and QS control. This is the second request. The third request is management integration. In the other IoT use case, we need to care. We need to care how we should manage not only computer resources of IoT platform, but also computer resources of edge devices. We as OpenStack users like to do unified manage of all Open Computer resources by OpenStack due to reduced operational cost. The next, the fourth request is application performance monitoring. In other words, we would like to ask OpenStack to support data gathering and data visualization, data visualizing for application monitoring. As we mentioned, our IoT platform has many features. The first one is data receiver, the second one is data lake, and the third one is data analysis. This means different characteristics application are lining on the same platform. Therefore, it is necessary to be able to monitor how much has happened on the whole of IoT platform and the part of the IoT platform to use hardware resource effectively. So, we'd like to monitor not only hardware and OS status, but also application status. We are happy if OpenStack can provide any framework like elastic search, friendly, or Kibana to gather and visualize an application log on OpenStack. So, last request is easy to manage data stores and ETL. In IoT use case, we have two requests. The one is OpenStack can provide various data stores for data analysis, applications own convenience. So, another is OpenStack can provide any framework to use ETL feature easily. ETL, currently it might be needed for combination data stores. It is a kind of the inventing of the wheel. In other words, it is a waste of time. So, we'd like to ask OpenStack to support any common framework about ETL. We believe this is useful for IoT use case. Okay, these are our requests for OpenStack. And then, we have imagined an IoT infrastructure reflecting these requests. Let me show you the architecture on the next slide. Before I show our new architecture, let me show the old one. This is showed in the last OpenStack time in Tokyo. So, and then, so this is a new architecture reflecting our request for OpenStack. Mainly, update point are PROS, container, connection gateway, edge monitoring, authentication and authorization and ETL part. Okay, about PROS, container and edge management, these are because we are considering how integrate edge computing and ground computing. About connection gateway, this component has to process heavily traffic around 10 gigabit VPS. We thought that virtual router might be accommodate our request for our use case. So, we changed from using virtual router to connection gateway. So, last one is ETL. Yeah, ETL is important in IoT use cases I mentioned. Then, we assured that our architecture mapping on current OpenStack. This is our architecture mapping on OpenStack. PROS container is corresponding to HIT and Magnum, you know, a part of edge gateway and connection gateway are new turn. Broca, current Zaka only supports MQ, not messages Broca. Trouble looks good for our architecture. We are happy if trouble test takes care of ETL future also. So, Hala also looks good for us. Swift is of course, is great for us. It is just fit to our architecture. And then today, we want to share five actual items that OpenStack should do to be used for IoT use cases from our experience. We are happy if we can contribute OpenStack diversity through this presentation. Okay, let me share two to the list for OpenStack to be useful in IoT use cases. We've described five items. Let's get to talk about them one by one from next slide. As I mentioned, the number of challenges for IoT platform and infrastructure running IoT platform is two. The one is to receive messages from data source certainly in spite of having the traffic. The other one is to store messages through data storage laboratory. So, we'd like to ask OpenStack to be supported by your guaranteed flavor about network and storage. The next one is messages broker support. The point is what kind of APIs should be designed for NorthBrand systems and how many messages broker OpenStack should support. From our experience, we ask OpenStack to support these messages broker like, and especially Apache Kafka should be fine because it has high, very high performance, scalability, durability and flexibility. Data streaming project, storm for NT from and so on. The third point is edge device management. This is actually still discussing what is the best idea. But now, we guess these will be to pretend to do edge device management. The one is cloud platform for IoT. It's actually managed the edge gateway. The another one is cloud platform for IoT. It's kicked on API like this figure by mobile carriers system, which is edge gateway management system. How should we manage both of nodes? What is the best architecture to do it? If you guys are interested in this item, we'll discuss this last later. The fourth item is performance monitoring. We would like to monitor not only hardware and OS status, but also application status. We are happy if OpenStack can provide any framework like elastic search, friendly work event to gather and visualize an application log on OpenStack. The last one is ETL as a service. Currently, a bunch of ETLs are needed for combination data storage. It is a kind of reinventing the wheel. If OpenStack supports something like ETL interface like this figure, it can be helpful for IoT use case. These are our total list for OpenStack we thought. Okay, let's go to the summaries of this presentation. IoT technologies enables us to gather huge data over the internet and create new value with data analysis. We updated our IoT architecture and summarize total list to make IoT platform OpenStack. Then we believe IoT powered by OpenStack. That is all about. Any questions? No questions? No questions? Okay, thank you for, yep. I was wondering how long and how much data you're gonna store for a long time, right? Because you are storing lots of data. I think you set up 50 petabyte per month, right? Then do you keep them or just throw away in the month? Just wondering how do you deal with those huge data? Okay, thank you for your question. So actually, it's a depend on our potential customer, but we guess data should be stored a long year, one year from now. No. Any questions? Thank you very much for attending this presentation. Thank you.