Big Data Analytics and Ubiquitous computing

Big Data Analytics and
Ubiquitous Computing
by
Dr.Animesh Chaturvedi
Assistant Professor: LNMIIT Jaipur
Post Doctorate: King’s College London &TheAlanTuring Institute
PhD: IIT Indore

Big Data Analytics
 Big Data Analytics
1. Big Data
2. Spark: Big Data Analytics
3. Resilient Distributed Datasets (RDD)
4. Spark libraries (SQL, DataFrames, MLlib for
machine learning, GraphX, and Streaming)
5. PFP: Parallel FP-Growth
 Ubiquitous Computing

Big Data
 Big data can be described by the following characteristics:
 Volume: size large than terabytes and petabytes
 Variety: type and nature, structured, semi-structured or unstructured
 Velocity: speed of generation and processing to meet the demands
 Veracity: the data quality and the data value
 Value: Useful or not useful
 The main components and ecosystem of Big Data
 DataAnalytics: data mining, machine learning and natural language
processing
 Technologies: Business Intelligence, Cloud computing & Databases
 Visualization: Charts, Graphs etc.

Apache Spark
 Unified analytics engine for large-scale data processing.
 Speed: Run workloads 100x faster.
 Both batch and streaming data, using DirectedAcyclic Graph
(DAG) scheduler, a query optimizer, and a physical execution
engine.
 Ease of Use:Write applications quickly in Java, Scala, Python,
R, and SQL.
 Spark offers 80+ high-level operators to build parallel apps.
https://spark.apache.org/

Spark: Unified Big Data Analytics
 New applications of Big data workloads on Unified Engine of
 Streaming, Batch, and Interactive.
 Composability in programming libraries for big data and
encourages development of interoperable libraries
 Combining the SQL, machine learning, and streaming
libraries in Spark
Zaharia, Matei, et al. "Apache spark: a unified engine for big data processing." Communications
of the ACM 59.11 (2016): 56-65.

Spark: Unified Big Data Analytics
 Spark has MapReduce programming model with extended
data-sharing abstraction called “Resilient Distributed
Datasets,” or RDDs.
Zaharia, Matei, et al. "Apache spark: a unified engine for big data processing." Communications
of the ACM 59.11 (2016): 56-65.

Spark Architecture
 Spark works on the popular Master-Slave architecture.
 Cluster works with a single master and multiple slaves.
 The Spark architecture depends upon two abstractions:
 Resilient Distributed Dataset (RDD)
 Directed Acyclic Graph (DAG)
https://spark.apache.org/docs/latest/cluster-overview.html

Resilient Distributed Datasets (RDD)
 A distributed memory abstraction: perform in-memory
computations on large clusters
 Keeping data in memory can improve performance
 Spark runtime: Driver program launches multiple workers
that read data blocks from a distributed file system and can
persist computed RDD partitions in memory.
Zaharia, Matei, et al. "Resilient distributed datasets: A fault-tolerant abstraction for in-memory
cluster computing." Presented as part of the 9th {USENIX} Symposium on Networked Systems
Design and Implementation ({NSDI} 12). 2012.

 Each box is an RDD, with partitions as shaded rectangles.
 narrow dependencies, where each partition of the parent
RDD is used by at most one partition of the child RDD,
 wide dependencies, where multiple child partitions may
depend on it.

 Direct Acyclic Graph (DAG) to perform a sequence of computations
 Each node is an RDD partition,
 Run an action (e.g.,count or save) on an RDD, the scheduler examines
that RDD’s lineage graph to build a DAG of stages to execute.
 Each stage contains with narrow
dependencies
 The boundaries of the stages are
the shuffle operations required for
wide dependencies.
 Scheduler computes missing
partitions until it computed RDD.

Apache Spark
 Combine SQL, streaming, and complex analytics. Spark libraries
 SQL and DataFrames,
 MLlib for machine learning,
 GraphX, and
 Spark Streaming.
 Spark runs on Hadoop,Apache Mesos, Kubernetes, standalone, or
in the cloud.
 Run Spark using its standalone cluster mode, on EC2, on Hadoop
YARN, on Mesos, or on Kubernetes.
 Access data in HDFS,Alluxio,Apache Cassandra,Apache
HBase,Apache Hive, and hundreds of other data sources.
https://spark.apache.org/

Spark Code Example
 Word Count
 Pi Estimation

Spark SQL
 Working with structured data.
 Integrated: SQL queries with Spark programs.
results = spark.sql("SELECT * FROM people")
names = results.map(lambda p: p.name)
Apply functions to results of SQL queries.
 Uniform DataAccess: Connect to data sources, including Hive,
Avro, Parquet, ORC, JSON, and JDBC.
spark.read.json("s3n://...").registerTempTable("json")
results = spark.sql("""SELECT * FROM people JOIN json ...""")
 Query and join different data sources
 Hive Integration Spark HiveQL
 Standard Connectivity: Connect through JDBC or ODBC.
 Business intelligence tools to query big data.
https://spark.apache.org/sql/

DataFrame API Examples
 Collection of data organized into named columns
 Use DataFrame API to perform various relational operations
 Automatically optimized by Spark’s built-in optimizer

Spark Streaming
 Build scalable fault-tolerant streaming applications.
 Write streaming jobs -- SameWay --Write batch jobs
 Counting tweets on a sliding window
TwitterUtils.createStream(...)
.filter(_.getText.contains("Spark"))
.countByWindow(Seconds(5))
 Reuse the same code for batch processing
 Find words with higher frequency than historic data:
stream.join(historicCounts).filter {
case (word, (curCount, oldCount)) =>
curCount > oldCount
}
https://spark.apache.org/streaming/

Spark GraphX
 Spark'sAPI for graphs and graph-parallel computation
graph = Graph(vertices, edges)
messages = spark.textFile("hdfs://...")
graph2 = graph.joinVertices(messages) {
(id, vertex, msg) => ...
}
 Fast Speed for graph algorithms
 GraphX graph algorithms
 PageRank
 Connected components
 Label propagation
 SVD++
 Strongly connected components
 Triangle count
https://spark.apache.org/graphx/

Spark MLlib
 Spark's scalable machine learning library
 Spark MLlib algorithms
 Classification: logistic regression, naive Bayes,...
 Regression: generalized linear regression, survival regression,...
 Decision trees, Random forests, and Gradient-boosted trees
 Recommendation:Alternating Least Squares (ALS)
 Clustering: K-means, Gaussian mixtures (GMMs),...
 Topic modeling: Latent Dirichlet Allocation (LDA)
 Frequent itemsets,Association rules, and Sequential
pattern mining
https://spark.apache.org/graphx/

FP-Growth for recommendation
 “FP” stands for Frequent Pattern in a Dataset of transactions
1. calculate item frequencies and identify frequent items,
2. a suffix tree (FP-tree) structure to encode transactions, and
3. frequent itemsets can be extracted from the FP-tree.
 Input:Transaction database
 Intermediate Output: FP-Tree
 Output: {f, c, a  a, m p}, {f, c, a  b, m}
Han Jiawei, Jian Pei, and YiwenYin. "Mining frequent patterns without candidate generation." ACM
SIGMOD Record 29.2 (2000): 1-12.

PFP: Parallel FP-Growth
 In Spark ML-Library (MLLib), a parallel version of FP-
growth called PFP: Parallel FP-Growth
 PFP distributes the work of growing FP-trees based on the
suffixes of transactions.
 More scalable than a single-machine implementation.
 PFP partitions computation, where each machine executes an
independent group of mining tasks
Li, Haoyuan, et al. "PFP: Parallel FP-Growth for query recommendation." Proceedings of the 2008 ACM
Conference on Recommender systems. 2008.

Example of MapReduce FP-Growth: Five transactions
composed of lower-case alphabets representing items
Li, Haoyuan, et al. "PFP: Parallel FP-Growth for query recommendation." Proceedings of the 2008 ACM
Conference on Recommender systems. 2008.

 Sharding: Divide DB into successive parts
and storing the parts (as a Shard) on P
different computers.
 Parallel Counting: MapReduce counts the
support of all items that appear in DB. Each
mapper inputs one shard of DB. The result is
stored in F-list.
 Grouping Items: Dividing all the items on
F-List into Q groups of a list (G-list).
 Parallel FP-Growth: A MapReduce
- Mapper: Each mapper uses a Shard. It reads a
transaction in the G-list and outputs one or more
key-value pairs, where each key is a group-id and
value is a group-dependent transaction.
- For each group-id, the MapReduce groups all
group-dependent transactions into a shard.
- Reducer: Each reducer processes one or more
group-dependent Shard. For each shard, a reducer
builds a local FP-Tree and discover patterns.
 Aggregating: Aggregate the results
generated as final result.
Li, Haoyuan, et al. "PFP: Parallel FP-Growth for query
recommendation." ACM Conf.on Recommender systems. 2008.

 FP-Growth implementation takes the following (hyper-)parameters
 minSupport: the minimum support for an itemset to be identified as
frequent e.g., if an item appears 3 out of 5 transactions, it has a support of
3/5=0.6.
 minConfidence: minimum confidence for generatingAssociation Rule e.g., if
in the transactions itemset X appears 4 times, X andY co-occur only 2 times,
the confidence for the rule X =>Y is then 2/4 = 0.5.
 numPartitions: the number of partitions used to distribute the work.
 FP-Growth model provides:
 freqItemsets: frequent itemsets in the format of DataFrame(“items”[Array],
“freq”[Long])
 associationRules: association rules generated with confidence above
minConfidence, in the format of DataFrame(“antecedent”[Array],
“consequent”[Array], “confidence”[Double]).

Ubiquitous Computing
 Big DataAnalytics (Apache Spark (SQL, Mllib, GraphX))
 Ubiquitous Computing
1. Edge Computing
2. Cloudlet
3. Fog computing
4. Internet ofThings (IoT)
5. Virtualization
6. Virtual Conferencing
7. Virtual Events (2D, 3D, and Hybrid)

Ubiquitous computing
 MarkWeiser:Three basic ubiquitous computing devices:
 Tabs: a wearable device that is approx in centimeters
 Pads: a hand-held device that is approximately a decimeter in
size
 Boards: an interactive larger display device that is approximately
a meter in size
 computing is made to appear anytime and everywhere
 any device, in any location, and in any format
Weiser, Mark. "The computer for the 21st century." ACM SIGMOBILE mobile computing and
communications review 3.3 (1999): 3-11.
https://en.wikipedia.org/wiki/Ubiquitous_computing

Edge computing
 Distributed computing paradigm
 Computation and data storage closer to the user location
 Improve response times and save bandwidth
 Cloud computing operates on big data, whereas Edge
computing operates on “instant data”
 Content Delivery Network or Content Distribution Network
(CDN) (Refer to Unit 4)
 Akamai CDN (Refer to Unit 4)
 Akamai-Facebook’s Photo-Serving Stack (Refer to Unit 4)
https://en.wikipedia.org/wiki/Edge_computing

Cloudlet
 First coined by Mahadev Satyanarayanan (Satya),Victor Bahl,
Ramón Cáceres, and Nigel Davie
 It is a mobility-enhanced small-scale cloud datacenter that is
located at the edge of the Internet.
 It work as a data center in a box which brings the cloud closer.
 Support resource-intensive and interactive mobile
applications by providing powerful computing resources to
mobile devices with lower latency.
https://en.wikipedia.org/wiki/Cloudlet

Fog computing
 Architecture that distributes computing, storage, control and
networking functions closer to the users along a cloud-to-thing.
 Fog computing is often erroneously called edge computing, but
there are key differences.
 Fog works with the cloud, whereas edge is defined by the
exclusion of cloud.
 Fog is hierarchical where edge tends to be limited to a small
number of layers.
 Cloud computing deal with Big Data, whereas Fog computing
deals with real-time data generated by sensors or users.
1 IEEE StandardAssociation. "IEEE 1934-2018-IEEE Standard for adoption of OpenFog reference
architecture for fog computing." (2018).
https://en.wikipedia.org/wiki/Fog_computing

Internet of Things (IoT)
 The network of physical objects —“things”— embedded with
sensors, software, and other technologies for the purpose of
connecting and exchanging data with other devices and systems
over the Internet.
 Example:
 “Smart Home” devices and appliances,
 “Smart city” equipment and facilities,
 Real-Time Data Analytics
 Information explosion or Data Deluge
 due to data flood or information flood
 ever-increasing amount of electronic data exchanged per time unit
 unmanageable amounts of data growthV/S power of data processing
https://en.wikipedia.org/wiki/Internet_of_things

Virtualization
 Virtual computer hardware platforms, storage devices,
and computer network resources.
 Hardware virtualization or platform virtualization refers to the
creation of aVirtual Machine (VM).
 Full virtualization – complete simulation of the actual hardware to
allow software environments, guest operating system and its apps.
 Paravirtualization – the guest apps are executed in their own isolated
domains.
 Application virtualization andWorkspace virtualization: isolating
individual apps from the underlying OS and other apps
 Service virtualization: emulating the behavior ofAPI, Cloud and
SOA
https://en.wikipedia.org/wiki/Virtualization

Virtualization
 Memory virtualization: aggregating RAM resources from
networked systems into a single memory pool
 Virtual memory: giving an app the impression that it has
contiguous working memory, isolating the underlying physical
memory
 Storage virtualization: the process of completely abstracting
logical storage from physical storage
 Distributed file system: any file system that allows access to files
from multiple hosts sharing via a computer network
 Virtual file system: an abstraction layer on top of a more concrete
file system
 Storage hypervisor: combines physical storage resources into one
or more flexible pools of logical storage

Virtualization
 Virtual disk: emulates a disk drive such as a hard disk drive or
optical disk drive
 Data virtualization: the presentation of data as an abstract
layer, independent of underlying database systems, structures
and storage
 Database virtualization: the decoupling of the database layer,
which lies between the storage and application layers
 Network virtualization: creation of a virtualized network
addressing space within or across network subnets
 Virtual private network (VPN): replaces the actual wire or
other physical media in a network with an abstract layer

Virtual Conferencing
 Teleconference: Phone lines, Landlines or Cellular devices
 Video conference:
 Webcam, Microphone, Speaker, Internet
 hardware, software, devices is dedicated for this
 Web Conference: Internet and Cloud supported
 Web 2.0
 Well-Known: Google Meet, Skype, and MicrosoftTeam
 Multi-Communications from Many sender to Many receivers
 Webinars ("web seminars"),Webcasts (live media presentation),
Podcast (audio presentation), and web meetings
 Virtual Events
https://en.wikipedia.org/wiki/Internet_of_things

Virtual Events
 An online event involves people interacting in a virtual
environment on the web, instead of physical meeting.
 Multi-session online events often feature webinars and
webcasts.
 Aim to create similar experience as physical meeting.
 Live-streaming the event online or on-demand video.
 Issues
 Echo of voice
 Audio andVideography logistics
 Network Bandwidth of conferencing server and users
https://en.wikipedia.org/wiki/Virtual_event

2D Virtual Events: Edited Collage

2D Virtual Events: Virtual Degree

2D Virtual Events: Virtual Address

3D Virtual Event: Live Video
Virtual HLF 2020 - Heidelberg Laureate Forum

3D Virtual Event:

3D Virtual Event: Poster Sessions

3D Virtual Event: Virtual Gathering

3D Virtual Event: Recreations

3D Virtual Event: Virtual Dancing

Live Streaming

Hybrid Event
 Combines a “physical" in-person event with a "virtual"
online component
 Tradeshow, Conference, Seminar,Workshop, Convocation
https://en.wikipedia.org/wiki/Hybrid_event

Hybrid Event: Help Chat FrontEnd

Hybrid Event: Help Chat BackEnd

Hybrid Event: Blowing Flower Effect

Hybrid Event: Remote Connectivity

References
 https://spark.apache.org/
 Zaharia, Matei, et al. "Apache spark: a unified engine for big data
processing." Communications of theACM 59.11 (2016): 56-65.
 Zaharia, Matei, et al. "Resilient distributed datasets:A fault-tolerant abstraction for in-
memory cluster computing." Presented as part of the 9th {USENIX} Symposium on
Networked Systems Design and Implementation ({NSDI} 12). 2012.
 https://spark.apache.org/docs/latest/cluster-overview.html
 https://spark.apache.org/
 https://spark.apache.org/sql/
 https://spark.apache.org/streaming/
 https://spark.apache.org/graphx/
 Han Jiawei, Jian Pei, and YiwenYin. "Mining frequent patterns without candidate
generation." ACM SIGMOD Record 29.2 (2000): 1-12.
 Li, Haoyuan, et al. "PFP: Parallel FP-Growth for query recommendation." Proceedings of
the 2008 ACM Conference on Recommender systems. 2008.

References
 Weiser, Mark. "The computer for the 21st century." ACM
SIGMOBILE mobile computing and communications review 3.3 (1999):
3-11.
 https://en.wikipedia.org/wiki/Ubiquitous_computing
 https://en.wikipedia.org/wiki/Edge_computing
 https://en.wikipedia.org/wiki/Cloudlet
 IEEE StandardAssociation. "IEEE 1934-2018-IEEE Standard for
adoption of OpenFog reference architecture for fog computing."
(2018).
 https://en.wikipedia.org/wiki/Fog_computing
 https://en.wikipedia.org/wiki/Virtual_event
 Virtual HLF 2020 - Heidelberg Laureate Forum
 https://en.wikipedia.org/wiki/Hybrid_event

Thank You
Japanese
Hebrew
English
Merci
French
Russian
Danke
German
Grazie
Italian
Gracias
Spanish
Obrigado
Portuguese
Arabic
Simplified
Chinese
Traditional
Chinese
Tamil
Thai
Korean
https://sites.google.com/site/animeshchaturvedi07

Big Data Analytics and Ubiquitous computing

More Related Content

What's hot (19)

Similar to Big Data Analytics and Ubiquitous computing (20)

More from Animesh Chaturvedi (20)

Recently uploaded (20)

Big Data Analytics and Ubiquitous computing