Adam Jorgensen, Executive Vice President Finance and Governance, welcomed the crowd. He pointed out we're providing 391 hours of education and technical training this year. Adam's excited about the future growth of PASS, and the need to develop new leaders throughout the organization. He shared the financial basis for the organization. Revinew for FY is $8.3M, a 10% increase over the previous year. Summit attendees have doubled since 2007. Revenue has grown consistenly since 2008. PASS has delivered 27 '24 Hours of PASS' events since its inception. Sponsorship sales have continued to grow. Since 1999, PASS has grown to over 150,000 members from 163 countries. 78% of the funds that PASS raises goes directly back to the community.
Denise McInerney, Vice President of Marketing, and presented the PASS Outstanding Volunteer award to Bill Graziano, outgoing PASS Immediate Past President, for everything he's done over the years for PASS. She then spoke about the community and our outstanding volunteers. 71 people have been nominated for Outstanding Volunteer since 2012. Ten people were nominated for this year's PASSion award, and it was given to Lance Harra this year.
PASS Summit will be back in Seattle next year, October 25-28, 2016.
The Keynote is Data Management for the Internet of Things (IoT), co-presented by David DeWitt, Technical Fellow, Microsoft Jim Gray Systems Labs, and Rimma Nehme, Principal Software Engineer, Microsoft Jim Gray Systems Labs.
Doctor Nehme came out to thank the PASS community for the new clicker, to replace the one she had that was broken last year. She explained that she'll begin the talk (as the "appetizer"), Dr. DeWitt will present the
IoT involves taking a physical object, add analytics, and you then provide value that the object couldn't have without the combination. There are 2 types of IoT - the consumer type, and the industrial type. Consumer includes things like fitbet, nest, etc. You can use that information to identify credit risk, education, general health, etc., of any of us.
Companies gain value from IoT based on unconventional revenues, inccremental revenues, and operational efficiency. IoT is still in its infancy.
There are four types of IoT capabilities - monitoring, control, optimization and autonomy. We're right at the peak of the "hype" cycle of IoT right now. Around 2008 the number of devices connected to the internet exceeded the number of people in the world, and it keeps growing. The value to customers is huge. With the number of devices connected, a savings of 1% across the board is tremendous in effiency.
IoT - How?
Dr. DeWitt came out to discuss the technical challenges of managing IoT. One of the biggest challenges is device/sensor security. From IoT we want to provide messaging, so we can learn things, and we need to have an easy way to deploy large numbers of implementations. On the consumer side you have to worry about battery life, where on the industrial side the power is not an issue. Consumer side cost is a real issue, and cost isn't really a factor. Consumer side the devices need to be embedded in a device, and wireless, where the industrial side can have standalone devices with wired connections.
IoT today is truly a DIY (do it yourself) process. The state of the art is rather primitive. In the field we've got devices with a sensor and an actuator to adjust the device. In the cloud, there's a event/data aggregator, connected with the device with a D2C (device to cloud) event. The aggregator can feed an application, stored data to storage, and some real-time processing engine, sending data to a cloud-based device controller, which will send commands back to the field device via C2D (cloud to device) events.
Two main components are Azure IoT Hubs and Azure Event Hubs. For data management, we can use Stream Analytics, DocumentDB, HDinsignt, etc. to gather store and analyze the data. Stream Analytics gather the data and stores a vast amount of data, HDinsight allows querying of that data, and Machine Learning can then act on those events.
The IoT hub can manage and control the devices. Sensors push their events into the hubs via endpoints. Hash functions are applied to the event to send the events to proper handlers and the handler can act on those events. Messages can then routed back from the hub back to devices via send endpoints and message queues.
Azure Machine Learning, using data from SQL Azure, Azure storage, etc., will "predict when the boiler will explode". The IoT hub will send events to a real-time query engine which can then process the results. In streaming systems, a sequence of events will be taken in, query the events and send out results across multiple hubs depending on the content. There's no long-term storage of data, it's the queries that are stored long-term.
Field devices require a field gateway to send the sensor readings to the IoT hub. (A Raspberry Pi serves as a good field gateway.) It also makes sense to connect multiple sensors to a single gateway for security ease in configuration. They may be capable of local processing as well. Per-device metadata can be stored in local storage on the device.
Device Security is a major concern. Per-device identities are used to authenticate, and the devices must pull to obtain the C2D commands.
These devices all push to the cloud. But there's some problems with pushing everything to the cloud. Bandwidth, connectivity, latency, data deluge, storage constraints, speed and specifically "non-interesting" events.
So why not exploit the capability of the field gateway? In the field gateway we could put a streaming engine to process events so that only the interesting events get sent to the IoT hub.
Dr. Nehme came back to talk about Fog Computing or Edge Computing.
You never move the data to the computation, you move the computation to the data. The device does some data gathering, the gateway does some data filtering and processing, and the cloud provides the analytics. This gives us better real-time response, scalability, and metadata management.
Polybase for IoT.
We need a Declarative Language, a complex object model, scalable metadata management, discrete and continuous queries, and multi-query processing. She talked about three query types: ExecuteOnce, ExecuteForever, ExecuteAction, where ExecuteOnce sends a query to the device and the device sends a response. ExecuteForever sends a query that continues to send results until given a stop response from the source, and ExecuteAction sends a query with an Action statement, and that action can be done once or done forever, until a stop response is sent.
Why should Data Professionals really care? When new technologies are introduced you can be either the steamroller, or the street. More solutions are required for the growth in IoT that will be coming.
They announced that this will be their last PASS keynote. They've done many keynotes and there are many new speakers that can provide new ideas. Dr. DeWitt will be retiring and Dr. Nehme will be moving on, either within Microsoft, or without.
Thank you, Dr. DeWitt and Dr. Nehme, for the enormous insight you've provide us over the years.