1.dbTouch : Analytics at your Fingertips Authors: Stratos Idreos , Erietta L iarou Presenter: Peter Liu
2.Introduction Brief Overlook Motivation: Interactive Data Exploration Vision: dbTouch Interface: Touch Input Challenges: From Gestures to Query Processing Exploiting dbTouch Interactive Exploration Front-end Slide to Explore From Touch to Tuple Identifiers Data Access and Touch Granularity Storing and Accessing Data Interactive Summaries Schema and Storage Layout Gestures Query Plans and Complex Queries dbTouch Prototype Challenges and Opportunities
3.The Big Data Era Exploit the map reduce and the cloud paradigms [1, 12, 34, 2, 38, 32] Database kernels for approximate query processing [3, 31, 40] Adaptive indexing [26, 27, 28, 18, 16, 29] Adaptive data loading [24, 4] High performance column-store [43, 8, 25] Hybrid database kernels [6, 30, 11, 17, 13] Crowdsourcing and anthropocentric systems [35, 15, 39, 14, 44] Crowdscreen : algorithms for filtering data with humans. (39) Crowddb : answering queries with crowdsourcing (15) Exploitation of modern hardware [20, 10, 36, 7] Usability  Energy aware system [21, 33]
4.Motivation: Interactive Data Exploration When we are in search for interesting patterns often not knowing a priori exactly what we are looking for. A n astronomer wants to browse parts of the sky to look for interesting effects. A data analyst of an IT business browses daily data of monitoring streams to figure out user behavior patterns. Identifying candidate features for machine learning models.
5.Vision: dbTouch Core database research + touch interfaces Redefine query plan and data flow Touches and gestures The database system does not have control anymore on the data flow. Challenges?
6.Touch Input IOs from Apple (2008), Android from Google, Windows 8 from Microsoft and WebOS from Palm Simplicity and interactive More people use and interact with touch interfaces. New kinds of applications appear.
7.Challenge: From Gestures to Query Processing Database architectures and visualization Translation: How we translate touch gestures to database query processing algorithms/operators? How we translate different gestures speed or movement to database processing algorithms/operators? Interactive and continuously changing mode The user is now in control of the data flow. Optimization: How we avoid stalling when a user giving interactive requests? Efficiently access, prefetch and precompute
8.Contribution The vision of dbTouch system A lgorithms and functionalities for several basic gestures O ptimizations, challenges and opportunities First implementation, evaluation and demo of an early prototype
9.dbTouch : Interactive Exploration Interactive Exploration Quality of the data Possible patterns and properties Interactive feeling (touching the data) Incremental and adaptive Requirements: Low level query processing action Visualization techniques
10.dbTouch : Front-end Question: what users experience and what they expect when interacting with a dbTouch system? Data Objects Schema-less Querying
11.dbTouch : Slide to Explore Slide What are the challenges here? Data storage and data access decisions Scan and Aggregates Slide speed Data object size Exact area touched Query Processing Which operation in traditional database kernels is equivalent to the slide gesture? The next operation
12.dbTouch : Slide to Explore Slide Example Inspecting Results Challenges How should data be stored and accessed? What happens when the slide patterns such as speed and directions change? Which data tuples exactly do we process with every touch?
13.dbTouch : From Touch to Tuple Identifiers Object view Views are placeholders for visual objects. Master view Properties (size of the view, the location of the view within its master view, what kind of gestures are allowed over the view, etc ) Mapping a touch to a RowID Properties: The number of data entries in the underlying column or table T he data type(s) T he data size, etc. Tuple identifier Rule of Three Tuple identifier: t = touch location o = the size of the data object n = the number of total tuples
14.dbTouch : From Touch to Tuple Identifiers Mapping a touch to a RowID Single column -> height dimension F ull table -> both dimensions vertical slide horizontal slide Rotation
15.dbTouch : Data Access and Touch Granularity Touching sample Physical constraints (what are those?) Finger size Object size Exploration speed Faster slide -> fewer tuples Slower slide -> more tuples Granularity Zoom-in/Zoom-out
16.dbTouch : Storing and Accessing Data Gesture Evolution Physical Layout Fixed-width per attribute (Pros? Cons?) Sample-based Storage Factors: size of the object and speed of the gesture Issue: continuous variation Prefetching Data Caching Data Indexing
17.dbTouch : Interactive Summaries The idea is that when sliding through a data object, dbTouch returns a summary of x items as opposed to simply returning a single data entry which corresponds to the exact touch location. I nspect more data Q uick inspection of properties and patterns
18.dbTouch : Schema and Storage Layout Gestures Pan gesture Group or ungroup Rotate gesture
19.dbTouch : Query Plans and Complex Queries Complex Queries where, select, join Joins Blocking operator Optimization
20.dbTouch Prototype Implementation slide or single tap for scan, aggregation and interactive summaries zoom-in/zoom-out for a more detailed or a more high level exploration. Evaluation Varying Gesture Speed Varying Object Size
21.Challenges and Opportunities Interactive Behavior Data Visualization Remote Processing Alternative Interfaces
22.Related Work Data Exploration Online Aggregation Visual Analytics Polaris: A system for query, analysis, and visualization of multidimensional relational databases. (41) Data3