Big Data Architech
the term ___________ refers to a technology environment comprised of Big Data mechanisms and technology artifacts that serves as a platform for developing Big Data solutions
The ___________ concentrate on the design of the underlying technology architecture of common Big Data platforms
Other memory-resident datasets can also be incorporated for performing analytics that provide context-aware results
In the context of software systems, a _________ represents the fundamental design of a software system
Design patterns
are primarly applied through the implementation of big data mechanisms
A _________ can be considered a proven design solution to a common design problem
Relational Source
File-Based Source
Streaming Source
Data Size Reduction
Alternative, the member patterns that comprise a __________ can represent a set of related features provided by a particular environment. In this case, a coexistent application of patterns establishes a "solution environment" that may be realized by a combination of tools and technologies
The _________ compound pattern represents a fundamental solution evironment comprised of a processing ______ with data ingress, storage, processing and egress capabilities
Dataset Decomposition
Streaming Storage
High Volume Binary Storage
High Volume Hierarchical Storage
Relational Sink
File-Based Sink
Processing Abstraction
Automatic Data Replication and Reconstruction
Automatic Data Sharding
Large-Scale Batch Processing
To provide efficient storage of such data, the ____________ pattern can be applied to stipulate the use of a storage device in the form of a key-value NoSQL database that services insert, select and delete operations
Complex Logic Decomposition
Automated Dataset Execution
High Volume Tabular Storage
High Volume Linked Storage
represent proven solutions to common problems
Big data ________ are (partially or entirely) applied by implementing different combinations of Big Data Mechanism
security layer
Note that the analysis layer may also encompass some level of data visualization features
Compound Pattern
is a coarse-grained pattern comprised of a set of finer-grained patterns singled out in this catalog are some of the more common and important combinations of the patterns, each of which is classified as a __________
Each _________ is represented by a hierarchy comprised of core (required) member patterns and extension (optional) patterns
However, with the passage of time, as more Big Data solutions are built and their complexity increases, additional Big Data mechanisms are introduced
Random Access Storage
Core patterns are connected via solid lines, and extension patterns are connected via dashed lines
document the effects of applying multiple patterns together
A _________ can represent a set of patterns that are applied together in order to establish a specific set of design characteristics. This would be referred to as joint application
Nested ___________ also exist. The Random Access Storage and Streaming Access Storage _________ are part of the Poly Storage pattern
Big Data Pipeline
Intermediate Results Storage (optional)
Poly Sink
Poly Source
Poly Storage
The architecture of a set of Big Data Mechanisms assembled into a solution
Big Data Processing Environment
are comprised of specific combinations of core (required) and extension(optional) member patterns
It helps define policies for: acquiring data from internal and external sources, which fields need to be anonymized/removed/encrypted, what consitutes personally identifiable information, how processed data should be persisted, the publication of the analytics' results and how long the data should be stored
When a event data transfer engine is used, the ingested data can normally be filtered out in-flight via the removal of unwanted or corrupt data
The extend to which an enterprise can benefit from a Big Data Solution is limited when it is deployed in isolation from the rest of the traditional enterprise systems
Big Data technology artifacts
Each pattern is asociated with one or more mechanism that represent common_______
visualization layer
Correct access levels need to be consistently configured across all required resources, such as a storage device or a processing engine
It is important to assess the interoperability and extensibility of each ________ so that upgrading a single ____________ does not impact the funcionality of other ______
Pattern-Mechanism Associations
A mechanism may either be implemented when the pattern is applied, or it may be directly affected by the application of the pattern
Not all of ________ may be required to apply the pattern. Sometimes one or more associated mechanisms may act as an alternative to others for how the pattern is applied
For regular exports, the file data transfer engine can be configurated via a workflow engine to run at regular intervals
Correspondingly, an ____________ (apart from other specifications) generally includes the specifications of multiple component architectures
The application of a pattern may not be limited to the use of its associated mechanisms. Other required components or artifacts are explained as part of the pattern descriptions
Note also that mechanisms are not associated with compound patterns
However, when performing deep analytics, such as in the case of predictive and prescriptive analytics, an analytics engine also exists
Due to its enterprise-wide focus, the _____________ can also be considered as a reference point for understanding what constitutes the enterprise infraestructure
Mechanisms
represent technology artifacts that can be combined to form Big Data architectures
A ________ provides the ability to compress and decompress data in big data platform
As a result, the corresponding architecture of the __________ also gets complicated
The ___________ provides an opportunity for developing an understanding of the analysis results in a graphical manner
Big Data Mechanisms
Serialization Engine
Compression Engine
Visualization Engine
Security Engine
Cluster Manager
Data Governance Manager
Productivity Portal
Serialization
is the process of transforming objects or data entities into bytes for persistence (in memory or on disk) or transportation from one machine to another over a network
In Big Data platforms, ________ is required for establishing communication between machines by exchanging messages between them, and for persisting data
To make data processing easier by not having to deal with the intricacies of processing engines, the ________ pattern can be applied, which uses a query engine to abstract away the underlying processing engine
The ________ bytes can either be encoded using a binary format or a plain-text format
Deserialization
The opposite transformation process from bytes to objects or data entities is called _________
The _______ pattern is normally applied together with the Large-Scale Batch Processing pattern as part of a complete solution
The _________ pattern is primarily implemented by using an event data transfer engine that is built on a publish-subscribe model and further uses a queue to ensure availability and reliability
A documentation of the _________ further helps to ascertain which maturity level of the analytics the enterprise is currently at
The ________ provides the ability to serialize and deserialize data in a Big Data platform
Different ____________ may provide different levels of speed, extensibility and interoperability
For example, a data processing job needs to be partitioned first to act on sub-groups of data. After processing, the results from each partition need to be consolidated together
A file data transfer engine is generally used to implement this design pattern that can further be encapsulated via the productivity portal
Ideally, a __________ should serialize/deserialize data at a fast speed, be amenable to future changes and work with a variety of data producers and consumers
These goals are achieved in part by serializing and deserializing data into and out of non-proprietary formats, such as XML, JSON and BSON
Generally, these multiple processing runs are connected together using the provided functionality within the processing engine or through further application of the Automated Dataset Execution pattern
The ___________ pattern is associated with the data transfer engine (relational), storage device, workflow engine and productivity portal mechanisms
Compression
is the process of compacting data in order to reduce its size, whereas decompression is the process of uncompacting data in order to bring the data back to its original size
The _________ compound pattern represents a part of a Big Data solution environment capable of storing high-volume and high-variety data and make-it available for indexed, ___________
The ______ consists of the storage device(s) that store the acquired data and generally consist of a distributed file system and a least one NoSQL database
A ____________ represents the design of a single software program, representing a module, is structured within a modular/distributed software environment
In Big Data environments, there is a requirement to acquire and store as much data as possible in order to derive the largest potential value from analysis.
governance layer
However, if data is stored in an uncompressed form, the available storage space may not be effienciently utilized
As a result, data compression can be used to effectively increase the storage capacity of disk/memory space. In turn, this helps to reduce storage cost
Due to the aforementioned reasons, a Big Data solution is generally integrated with the rest of the enterprise IT systems in order to provide maximum value
By using a point and click interface to work with the Big Data platform, the _______ makes it easier and quicker to populate the Big Data platform with the required data, manage and process that data and export the processed results
A Big Data platform is a combination of multiple resources, which are provided by different mechanism, each with its own unique security configuration, This requires developing and maintaining separate security policies and API-based integration
In a clustered environment, this can become cumbersome and difficult to maintain, especially where data is accessed across the enterprise with varying levels of authorization
However, a __________ goes beyond data curation and includes data processing, analysis and visualization technologies as well
Instead of configuring access for each resource individually, a ___________ can be used
A ______ acts as a single point of contact for securing a Big Data platform, providing authentication, authorization and access auditing features
It acts as a perimeter guard for the cluster with centralized security policy declaration and management, enabling role-based security
For enhanced data access security, a ________ may provide fine-grained control over how data is accessed from a range of storage devices and assist with addressing regulatory compliance concerns
A _______ may also integrate with enterprise identity and access management (IAM) systems to enable single sign-on (SSO)
Futhermore, _________ provide data confidentiality by enabling data encryption for at-rest (data stored in a storage device) and in-motion (data in transit over a network) data
A Big Data platform generally exists as a cluster-based environment ranging from a few to a large number of nodes
Due to the multi-node nature of such an environment, the provisioning, configuration,day-to-day, management and health monitoring of a cluster can be a daunting task
Cloud-based Big Data Storage (optional)
A ________ provides centralized management of a cluster, enabling streamlined deployment of core services over the cluster and their subsequent monitoring
In the context of a Big Data platform, a service, such as MapReduce (a processing engine) or HDFS (a distributed file system), refers to a background process that executes a Big Data mechanism
Instead of individually installing, managing and monitoring services on each node, a ________ provides a dashboard from where these task can be centrally performed via simple mouse click instead of the authoring and running of command line scripts
A _______ provides a centralized view to monitor cluster health, services status and resource utilization. It also provides for the configuration of various node-level and cluster-level alerts
the ______ supports the deployment of new services and the addition of nodes to a cluster
the _______ further help reduce cluster administration overhead and makes diagnosis more efficient
Note that the ____________ pattern also covers the acquisition of semi-structured data, such as XML or JSON-formatted data
Based on the type and location of the data sources, this layer may consist of more than one data transfer engine mechanism
For instance, the _______ makes it quicker and easier to find out why a particular service responsible for a specific storage device is not running
The net effect is streamlined _____________ that minimizes cluster downtime and enables reliable and timely Big Data analysis
Futhermore, the _________ can integrate with other infraestructure management tools to provide a unified view and the ability to perform performance tuning
This pattern is associated with the storage device (column-family) and serialization engine mechanisms
controls the management of the data lifecycle to ensure that quality data is available in a controlled, secure and timely fashion
helps ensure regulatory compliance, risk management and the establishment of data lineage
The _________ provides an understanding of the range of datasets used by different Big Data solutions
One of the main differentiating characteristics of Big Data environments when compared with traditional data processing environments is the sheer amount of data that needs to be processed
in a Big Data environment, the variety characteristic coupled with unknown access scenarios can make __________ a challenging task
A _________ is a tool with features for performing a range of common ____________ task in a centralized manner
A _________ can provide information on where the dataset resides, who the data owner/steward is, what the format of the data is, when the dataset was acquired, the source of the dataset, expiry date(if any), schema information via metadata search, a lineage viewer for establishing provenance
A _________ supports data lifecycle management through: the authoring of data retention and eviction policies, the establishment of security policies that specify the condition under which encryption is applied to a dataset or specific fields of a dataset, the creation of policies that establish disaster recovery management procedures
Futhermore, a ___________ can provide information on the level of trust and sensitivity of data
This information includes whether or not the data can be stored in a cloud environment, as well as any geographical limitations for data persistence
To ensure enhanced data confidentiality and privacy within a cluster, an advanced _________ may further enable fine-grained control over data storage by specifying which nodes can store which types of datasets
Automatic Data Replication and Reconstruction (core)
To make sense of large amounts of data and to perform exploratory data analysis in support of finding meaningful insights, it is important to correctly interpret the results obtained from data analysis
This interpretation is dependent upon the Big Data platform's ability to present data in visual form
A ________ graphically plots large amounts of data using traditional visualization techniques, including the bar chart, line graph and pie chart, alongside contemporary, Big Data-friendly visualization techniques, such as heat maps, word clouds, maps and spark charts
Additionally, a __________ may allow the creation of dashboards with filtering, aggregation, drill-down and what-if analysis features, along with the exportation of data for specific views
A ________ greatly enhances the productivity of data scientists and business analysts
A ________ provides a foundation for creating self-service visualizations for business intelligence (BI) and analytics
A Big Data platform provides a range of features, including data import, storage, processing and analysis, as well as workflow creation, via various mechanisms
Interacting with each of these mechanisms using their default interfaces can be difficult and time-consuming due to the mechanisms non-uniform natures. Further tools may need to be installed in order to make this interaction easier and to make sense of the processing results
Multiple __________ are generally deployed in an enterprise for fulfilling different business requirements
As a result, it takes longer to get from data import to data visualization, which further impacts the productivity and the overall value attributed to Big Data exploration and insight discovery (the value characteristic)
A ________ provides a centralized graphical user interface (GUI) for performing key activities that are a part of working with Big Data, including importing and exporting data, manipulating data storage, running data processing jobs, creating and running workflows, querying data, viewing schemas and performing searches
By understanding the makeup of each _______ the corresponding data ingestion requirements are better understood
The ________ pattern can be applied to automatically export data from the Big Data platform as a delimited or a hierarchical file
A ________ establishes a unified interface for configuring and managing the underlying mechanisms of the Big Data solution environment, such as establishing settings for the security engine
Additionally, a __________ may encapsulate a visualization engine in order to provide more meaningful, graphical views of data
Shared-Everything Architecture
is a machine-level architecture where multiple processors (CPUs) share memory and disk storage
can be implemented in two different ways: simetric multiprocessing and distributed shared memory
This is due to the fact that the same _________ can be physically implemented in a number of ways. The physical implementation is dependent upon the underlying technology that is used
A ________ represents the high-level components of a system, their functionality and how they are connected with one another
It should be noted that SMP and DSM apply to a single machine
is suitable for transactional workloads where data being processed is small and can be stored on a single machine
As all the resources (processor, memory and disk) exist within the boundaries of a single machine, data exchange only occurs within those boundaries
Therefore, transactional data can be processed quickly without any latency using a simple programming framework
Since all of the resources are tightly coupled together in a ___________, scalability becomes an issue
A storage area network (SAN) or a network-attached storage (NAS) solution can possibly be attached to a high-end multiprocessor machine to process large amounts of data
However, the network then becomes a bottleneck, with the data transfer taking longer than the actual data processing because the data needs to be transferred across the network
A ________ augmented with an SAN, which geatly increases storage capacity. However, the data now needs to be transferred across the network for processing, which adds to the data processing latency
With a ___________ , in order to cope with greater resource demands for CPU and/or disk space, the only option to scale up by replacing existing machines with higher-end (expensive) machines. Scaling up allows more processing and offers greater storage
However, any type of architecture that relies upon vertical scaling has an upper limit due to technology constraints such as maximum number of processors or memory limitations
Once the limit is reached, the only option is to scale out. Scaling out is a Big Data processing requirement that is not supported by _________
In the case that stored value conforms to a structure, such as a log file, the field names, along with field type information, are also recorded
Simetric Multiprocessing (SMP)
memory is pooled and shared between all processors. ______ is also known as uniform memory access (UMA)
In most cases, the ingested data is first stored on the distributed file system in a compressed form (apart from removal of unwanted and corrupt data)
the adquisition of new hardware resources
Although expensive, _________ databases provide atomicity, consistency, isolation and durability (ACID) compliance while supporting the querying of data using Structured Query Language (SQL)
Distributed Shared Memory (DSM)
multiple pools of memory exist. Thus memory is not shared between processors. ____ is also know as non-uniform memory access (NUMA)
For ________, apart form documenting the attributes and types of each entity, the possible connections (the edges) between entities are also recorded
Shared-Nothing Architecture
A ________ is a type of distributed architecture that consists of fully independent machines. The machines each have their own processors, memory, disk and operating system and are networked together as a single system
The ________ is self-sufficient and is free of any shared resources. For this reason, it is a highly scalable architecture that provides scale-out support, meaning extra machines can be added as required.
A ____________ is somewhat similar to a component architecture. However, in practice a single mechanism may itself be comprised of more that one component
Although highly scalable, this architecture approach requires the use of complex distributed programming frameworks
This design pattern is generally applied together with the Stream Access Storage pattern
Usual data processing techniques employed within a ___________ include data sharding and replication where large datasets are divided and replicated across multiple machines
This work well for Big Data where a single dataset may be divided across several machines due to its volume
In this way, with Big Data processing, data and processing resources can be co-located, thereby reducing data transfer frecuency and volume
the __________ achieves this functionality via the data _________ manager
Masive parallel processing
is an architecture that can be applied to distributed query processing in a shared-nothing architecture
is mainly employed by high-end databases and database appliances like IBM Netezza and Teradata
An __________ describes the design used to integrate two or more applications, and further encompasses the technology architectures of the integrated applications.
the architecture of a single Big Data mechanism
Databases based on ______ architecture generally use high-end hardware and a proprietary interconnect to link machines in order to enable the throughput required for high-speed analytics
databases generally require data to exist in a structured format at the time of loading the data into the database. In other words, a schema needs to exist
This prior knowledge about the structure of the data makes ___________ databases very fast at querying large datasets
The structured format requirement introduces the need for an extra ETL step to be performed before unstructured data can be loaded into the __________ database
It is advisable to develop an inventory of all _________ to avoid duplication of datasets. This also helps with identifying relevant datasets when performing exploratory analysis
MapReduce
Similar to MPP, __________ (a batch processing engine) is a distributed data processing framework that requires a shared-nothing architecture
makes use of commodity hardware where machines are generally networked using Local Area Network (LAN) technology
A productivity portal normally encapsulates a relational data transfer engine for point-and-click import
___________ - based processing platforms, such a Hadoop, do not require knowledge of data structure at load time.
Therefore ______ is ideal for processing semi-structured and unstructured data in support of executing analytical queries
However, without any knowledge of the data structure, data processing is slower with ________ as compered to MPP due to the inability to optimize query execution
Both MPP databases and __________ make use of the shared-nothing architecture and are based on the divide-and-conquer principle
Both MPP systems and __________ can be used for Big Data processing. However, from scalability point of view, MPP systems, when compare with ________, provide limited support for scaling out, as they are generally appliance-based
MPP systems are also a costlier option than ___________, which leverages inexpensive commodity hardware
_________, a framework for processing data, requires interaction via a general purpose programming language, such as java
A ________ provides raw storage where the value (the stored data) can be of any type, such as a file or a image, and is accessible via a key
Technology Infraestructure
In the context of software systems, a _________ represents the underlying environment that enables the design and execution of a software system
A ____________ defines the overall processing and storage capabilities of an IT enterprise. As well, a ____________ set the constraints within which the technology architecture needs to be designed
Efficient processing of large amounts of data demands an offline processing strategy, as dictated by the ___________ design pattern
Technology Architecture
A __________ can be defined for varying levels of software artifacts ranging from a single software library to the set of sotfware systems across the entire IT enterprise
With a reasonable amount of data acquisition, IT spending only increases slightly with the passage of time. As the amount of acquired data increases exponentially, there is a tendency for IT spending to increase exponentially as well
In a Big Data solution environment, quite often data needs to be imported from relational databases into the Big Data platform for various data analysis tasks
Traditional Architecture Types
component architecture
application architecture
storage layer
integration architecture
enterprise technology architecture
Feature-wise, a module is inherently different from a software program, as the former only provides a specific set of functionality for performing a subset of operations when compared against the complete software program
The modules are dependent on other modules for providing the full set of functionality and hence are designed to be composable
An __________ represents the design and structure of a complete software system that can be deployed on its own
In a modular/distributed software environment, an _____________ generally consists of a number of modules and some storage
This generally involves connectors, middleware and any custom developed components
A documented _____________ provides a point of reference for ensuring continued integration in the face of changes to the integrated applications' architecture
Futhermore, the capabilities of this layer indicate the kinds of a Big Data solutions that can be built
enterprise technology
An _________ architecture represents an ____________ landscape including its respective architectures
In the case of continuosly arriving data, data is first accumulated to create a batch of data and only then processed
In Big Data environment, large volume not only refers to tall datasets (a large number of rows) but also to wide datasets (a large number of columns)
In contrast to the other three technology architectures, which can be documented before their development the ______________ can is generally documented once other architectures are in place
The scope of the _____________ encompasses component, application and integration architectures
Big Data Mechanisms Architecture
The ___________ refers to the technology architecture of an individual ___________ that provides a specific functionality, such as a data transfer engine or a query engine
However, if internal data is also integrated, the same solution will provide business-specific results
Unlike traditional component architecture, the architectures of many ___________ are available due to the many active, open source projects that create and sustain them
A _____________ is generally a complete software package that can exist on its own but only truly realices its full potential when it is combined with other __________
For example, a storage device and a processing engine can exist on their own. However, the real value of these two is accomplished when the processing engine retrieves data from the storage device and processes the data to obtain meaningful results
For example, a resource manager should be compatible with different types of processing engines (batch and realtime) and should provide extension points for integrating future, processing-specific processing engines
Similarly, when upgraded, the resource manager should provide backward compatibility with older processing engines to ensure disruption-free operation
Big Data Solution Architecture
The ___________ represents a solution environment built to address a specific Big Data problem, such as realtime sensor data analysis or a recommendation system
The __________ is associated with the storage device mechanism
Such a solution environment represents a set of multiple Big Data mechanism that collectively provide the required business functionality
In a Big Data environment, the term __________ is similar to the term application architecture. However, in the domain of Big Data, it is the collective application of Big Data mechanisms that results in the creation of a __________. This is different that the concept of a traditional, package software application
A ___________ is generally a Big Data pipeline comprising multiple stages where complex processing is broken down into modular steps called tasks
Each task in a Big Data pipeline can make use of a processing engine mechanism, such as MapReduce or Spark, or a query engine mechanism, such as Hive or Pig to perform operations on Data
Complex ___________ may involve more than one data pipeline, for example, one for realtime data processing and the other for batch data processing
A documentation of the corresponding architectures provides an understanding of the utilization levels of common Big Data mechanism
This helps with establishing the scalability requirements of each Big Data mechanism and determining any potential performance bottlenecks
Big Data Integration Architecture
The architecture that consists of integrating a Big Data solution with the traditional enterprise systems
The ________ pattern is associated with the query engine, processing engine, storage device, resource manager and coordination engine mechanisms
With a ___________ architecture, in order to cope with greater resource demands for CPU and/or disk space, the only option to scale up by replacing existing machines with higher-end (expensive) machines. Scaling up allows more processing and offers greater storage
A Big Data solution that is integrated with other parts of an enterprise ecosystem provides maximum value because the data it contains or the analytics results it generates can be usedby other traditional enterprise systems, such as the enterprise data warehouse or an ERP system
Similarly, a Big Data Solution that only makes use of external datasets may produce generic or out-of-context results that are of little value to the business
The resulting architecture is known as the __________, which includes the architecture of the Big Data solution, any connected enterprise systems and integration components
With respect to a Big Data solution, there are generally two integration points: one for importing the raw data that needs to be processed and the other for exporting the results or, in some cases, exporting the ingested cleansed data
One prominent area within the field of pattern identification is the analysis of connected entities. Due to the large volume of data in Big Data environments, efficient and timely analysis of such data requires specialized storage
Big Data solutions demand opposing access requirements when it comes to raw versus processed data
For this, multiple data transfer engines or connectors, such as ODBC, are employed
Instead of using multiple point-to-point connections (connectors) between each Big Data solution and the traditional system, a single data bus can be used that provides a standardized integration approach across multiple Big Data solutions
Big Data Platform Architecture
The architecture of the entire ____________ that enables the execution of multiple Big Data solutions
The __________ is the underlying technology architecture that supports the execution of multiple Big Data solutions
Furthermore, the storage device automatically detects when a replica becomes unavailable and recreates the lost replica from one of the available replicas
This type of architecture documents the underlying Big Data mechanism that have been assembled in different combinations to construct multiple Big Data solutions
This generally represents a layered architecture where each top layer makes use of the successive bottom layer
The typical makeup of a __________ includes the storage layer, processing layer, analysis layer and visualization layer
At the start of a Big Data initiative, the _________ may only consist of a rudimentary set of Big Data mechanism for supporting a simple Big Data Solution
For example, the existence of an analytics engine in addition to a query engine indicates that the enterprise employs some level of predictive analytics
An enterprise generally starts off or is already at the descriptive or diagnostic analytics maturity level and aims to move towards the predictive or predictive analytics maturity level
A _________ can also be considered a superset of the traditional data architecture, as the former includes the development of data architecture for both raw and processed data
Dataset Location: the location from where the data will be available, which can be internal or external to the enterprise, including the cloud
It further includes decisions based around which storage technologies to use, what should be the format and structure of processed data and dictionary of datasets available for developing Big Data solutions
For _________, the description and type of the entity being stored is documented, such as product image and png
Logical Architecture
The term logical emphasizes the fact that the description of the architecture does not bear any resemblance to the physical implementation of the system
For health monitoring purposes, the cluster manager gathers metrics from various components running within different layers, such as the storage , processing and analysis layers, and displays their current status using a dashboard
Big Data Analytics Logical Architecture
A __________ defines the logical components required for the implementation of a Big Data analytics solution
It is a specialized form of a Big Data platform architecture that defines the Big Data mechanisms at each of its different layers, the responsibility of each of these layers and the generic flow of data between these layers
Data is ingested generally via file and/or relational data transfer engines, saved to the disk-based storage device and then processed using a ___________ engine
Devising such an architecture provides an easy-to-understand reference point for both the Big Data architects and the Big Data engineers
data sources layer
data adquisition layer
processing layer
analysis layer
utilization layer
management layer
The ____ comprises all the _______ that had been identified during the Data Identification stage of the Big Data analysis lifecycle
However, rather than consisting of ________ for a single big Data solution, the _________ encompasses ___________ across all Big Data Solutions
Dataset Type: the underlying format of the data produced by the source (structured, unstructured or semi-structured)
It is important to understand that this layer does not form part of the physical Big Data analytics architecture, as the data is produced by a source, such as an API, a database or web location, that is part of a separate system
Processing Big Data datasets involves the use of processing engines that need programmatic skills in order to work with them
Access Type: has the data open or restricted access?
Access Method: is data available via a simple connection or does it need scraping from a web resource?
Access Cost: is the data available freely or is there a cost associated with this acquisition, such as from a data market?
Data Production Speed: the rate at which the data source generates the data
data acquisition layer
The _______ provides functionality for acquiring data from the sources in the data sources layer
For internal structured data sources, a relational data transfer engine can be used
For semi-structured and unstructured data sources, whether internal or external, an event or file data transfer engine can be used
In the case of realtime processing of data or stream analysis, an event data transfer engine is generally used
Similarly, in the case of a relational data transfer engine, corrupt or wanted data can be filtered out at the source by the specification of a constrained selection criteria
However, in the case of a file transfer engine, the file needs to be ingested before it can be examinated for the filtration process
The ________ provides an easy-to-interact interface for analyzing data in the storage layer and consists of the query and analytics engines
this layer also includes mechanism for automatically appending metadata to the ingested data for assuring quality and maintaining provenance and compression of data
The data adquisition and Filtering stage of the Big Data analysis lifecycle is supported by the ________
Under certain circumstances, acquiring data may require API integration, which further warrants the development of custom (code) libraries or service development, which reside in this layer
the distributed file system automatically splits a large dataset into multiple smaller datasets that are then spread across the cluster
Note that in the case of realtime data processing, the ________ also consist of in -memory storage technologies that enable fast analysis of high velocity data as it arrives
Furthermore, in a production environment, the complete cycle needs to be repeated over and over again
This is because a distributed file system provides the most inexpensive form of storing large volumes of data
From the distributed file system, data can be pre-processed and put into a more structured form using an appropiate NoSQL storage device
A structured (but not necessarily relational) form is required because the exploratory analysis of data and the derivation and application of statistical and machine learning models require data whose attributes can be accessed in a standardized manner
Although the conversion to a structured form may not seem obvious in the case of applying semantic analysis techniques, even techniques such as text analytics first convert a document into a structured form before performing clustering, classification or a searching
Data that has undergone transformation, validation and cleasing operations is generally stored in one of the NoSQL databases, namely key-value, column-family, document and graph NoSQL databases
key-value database
the _______ pattern is associated with the storage device (distributed file system) and processing engine (batch) mechanisms.
document database
A ________ is capable of storing each record in a hierarchical form that can be accessed via a key, imitating a physical document that can have multiple selections
For ________, the hierarchical structure of the different documents being stored, along with their types, is documented
column-family database
A ________ is like a relational database that stores data in rows and columns. However, rather than storing a value per column, multiple key-value pairs can be stored inside a single column
For ________, the field names of each entity and any sub-fields within each field, along with their data types, are recorded. Also, based on the analysis requirements, it is important to decide between storing data as wide-rows or as tall-columns
In an actual implementation, the __________ consists of Open Source libraries, third-party business intelligence (BI) or analytics software
graph database
A ________ stores data in the form of connected entities where each record is called a node or a vertex and the connection between the entities is called the edge, which can be one-way or two-way
Before data can be stored in a NoSQL database, a data modelling exercise is generally undertaken
However, unlike the relational data modeling activity where entity names, attributes and relationships are documented, the nature of the NoSQL data modelling activity is different due to its non-relational nature and further depends on the type of the NoSQL database
In a NoSQL database, the emphasis is more on the structure of the individual aggregate, which is a self-contained record that has no relationships with other records
Apart from different storage devices, this layer also houses serialization and compression engines for storing data in an appropriate format and reducing storage footprint, respectively
The ________ provides a range of processing capabilities that play a pivotal role in generating value from a variety of voluminous data arriving at a high velocity in a meaningful time period
Apart from resource manager and coordination engines, although this layer can contain both the batch and the realtime processing engine mechanisms, based on the type of analytics performed, only one processing engine, such as the batch processing engine, may actually be present
Data Ingress/Egress - a data transfer engine may utilize a processing engine for transfering data
This pattern requires the use of a storage device implemented via a document NoSQL database servicing insert, select, update and delete operations. The document NoSQL database generally automatically encodes the data using a binary or a plain-text hierarchical format, such as JSON, before storage
Data Wrangling - data pre-processing activities, including data validation, cleansing and joining
Data Analysis - analytical activities, including querying, exploratory data analysis and model generation
This layer can further be divided into the batch and realtime ________
The ________ compound pattern represents a part of a Big Data solution environment capable of egressing high-volume, high-velocity and high-variety data out of the Big Data solution environment
Batch Processing
involves a _______ engine that processes large amounts of data stored on disk-based storage device in batches
This is the most common form of data processing employed in a Big Data environment for data wrangling operations, exploring data and developing and executing statistical and machine learning models
This layer represents the functionality as required by the Utilization of Analysis Results stage of the Big Data analysis lifecycle
Due to its nature of processing, the processing results are not available instantaneously
Although the __________ is mainly concerned with policies that guide data management activities, it may further provide functionality for managing other aspects of the Big Data platform
Realtime Processing
involves a _______ engine that processes continuosly arriving data (streams) or data arriving at intervals (events) as it arrives
Instead of persisting the data to a disk-based storage device, _________ persists the data to a memory-based storage device
Although providing instantaneous results, setting up such as capability is not only complex but also expensive due to the reliance on memory-based storage (memory is more expensive that disk)
Data is ingested via an event data transfer, saved to a memory-based storage device and then processed using a ___________ engine
Note that although in-memory storage is initially used, data is also saved to disk-based storage for deep analysis or future use
For providing maximun value, a __________ layer should provide low latency, high throughput, high availability and high fault tolerance
Event Stream Processing (ESP)
Complex Event Processing (CEP)
During ________, multiple streams or events that generally originate form disparate sources and are spread out over different time intervals are analyzed simultaneously for finding correlations, patterns, anomalous behavior and error conditions
generally refers to processing event-based data
However, the execution of data queries, which requires an instant response, on already persisted data acquired via batch import also falls in the domain of _________
The application of the __________ pattern further increases the reach of the Big Data soltution environment to non-IT users, such as data analysts and data scientists
Event Stream Processing
During ________, the incoming stream of data or events, which generally originates from a single source and is ordered by time, is continuosly analyzed via the application of algorithms or query execution
In simple use cases, ____________ involves data cleasing, transformation and the generation of some statistics, such as sum, mean, min or max, which are then fed to an operational dashboard
To counter this issue, the dataset is horizontally broken into smaller parts as prescribed by the ___________ pattern
In complex use cases, statistical or machine learning algorithms with fast execution times can be executed to detect a pattern or an anomaly or to predict the future state
Although the processing results can be directly utilized (a dashboard or an application), they can act as a trigger for another application that performs a preconfigured action, such as making computational adjustments, or further analyses
In general, ___________ focuses more on speed than complexity. The operation needs to be executed in a comparatively simple manner in order to aid faster execution. Also, it is easier to set up than CEP but provides less value
Complex Event Processing
Like ESP, the objective is to aid in making realtime decisions either automatically or through human intervention the moment data is received
Storing very large dataset where they are accessed by a number of users simultaneously can seriously affect the data access performance of the underlying database
When compared with ESP, __________ provides more value but is harder to set up, as it involves connecting with multiple data sources and executing complex logic
Complex correlation and pattern identification algorithms are applied, and business logic and KPIs are also taken into account for discovering cross-cutting ________ patterns
can be considered a superset of ESP. Oftentimes, both approaches can be deployed together such that the synthetic events generated as the output of ESP can become input for __________
provides rich analytics. However, due to its complex nature, time-to-insight may be adversely affected
Depending on the type of amalytics being performed, this layer may only consist of a query engine, such as in the case of descriptive and diagnostic analytics
Due to the contemporary nature of these processing engines and the specialized processing frameworks they follow, programmers may not be conversant with the APIs of each processing engine
This is the layer that converts large amounts of data into information that can be acted upon
This layer abstracts the processing layer with a view of making data analysis easier and further increasing the reach of the Big Data platform to data scientists and data analysts
Activities supported by this layer include: data cleasing, data mining, exploratory data analysis, preparing data for statistical/machine learning model development, model development, model evaluation and model execution
The functionality provided by this layer corresponds to the data analysis stage of the Big Data analysis lifecycle
In case of Open Source libraries, interaction is mostly command-line-based, with a basic graphical user interface (GUI) in some cases
However, third party software provides a GUI with point-and-click functionality for statistical/machine learning model development and other general data querying
The __________ hosts the visualization engine and provides functionality as required by the Data Visualization stage of the Big Data analysis lifecycle
The ________ pattern is associated with the compresssion engine, storage device, data transfer engine and processing engine mechanisms
However, the nature od such visualizations is different and more analysis-specific
Visualizations are generally utilized by data scientists and analysts to help them understand the data and the output of various analysis techniques
The visualization features provided by the __________ are mainly geared towards business users so that they can easily interpret the insights obtained from the data analysis exercise
However, based on the physical implementation, the third-party tools used at the analysis layer may provide the ability to create visualizations and publish them for enterprise-wide use so that different information workers and business users can turn the published information into knowledge for making informed decisions
This layer is also fundamental to the concept of self service BI, where business users can access enterprise data directly without first requesting it from the IT team, can perform the required analyses and can create the required reports and dashboards themselves
To ensure longevity of the Big Data analytics platform, the compatibility of the visualization engine, with respect to the types of data sources it can connect to, needs to be assessed, as the analysis results are normally persisted to a storage device, such as a NoSQL database
However, if the maximum benefit is to be gained from these results, they need to be incorporated into the enterprise in some shape and form
The __________ provides the analysis results so that an enterprise can take advantage of an opportunity or mitigate a risk in a proactive manner
The query engine provides an easy-to-interact-with interface where the user specifies a script that is automatically converted to low-level API calls for the required processing engine
The functionality provided by the ____________ will vary based on the utilization pattern of the analysis results
This includes exporting results to dashboard and alerting applications (online portal), operational systems (CRM, SCM, ERP and e-commerce systems) and automated business processes (Business Process Execution Language-based processes)
At other times, data products are provided that enable the generation of computational results, such as outlier detection, recommendations, predictions or scores that can be used to optimize business operations
In almost all cases, one or more data transfer engines are present that enable the export of the analysis results form the storage device(s) to downstream systems or applications
To automate the entire export process, a workflow engine that resides in the management layer is used in conbination with a data transfer engine
This enables automatic access to the analysis results stored in the disk-based or memory-based storage devices
the _________ is tasked with the automated and continuous monitoring as well as maintenance of a Big Data platform for ensuring its operational integrity
The functionality supported by this layer relates to the operational requirements of a Big Data platform, including cluster setup, cluster expansion, system and software upgrades across the cluster and fault diagnosis and health monitoring of the cluster
the __________ provides functionality that ensures that the storage and access to data within the Big Data platform are managed throughout the lifespan of the data
The ________ achieves the provisioning of the required functionality by hosting a cluster manager
For cluster maintenance, such as adding a new node to the cluster, taking a node offline or installing a new service, and disaster ___________ task a graphical user interface (GUI) is used
Due to the requirement of integrating with multiple types of components, an inteorperable and extensible cluster manager should be chosen
This ensures continuous monitoring and _________ of the Big Data platform, as new components are added across multiple layers in response to the ever-changing analytics requirements of an enterprise
Apart from operational ________, this layer also provides data processing and data __________ functionality through workflow engine and productivity portal mechanisms
The __________ is responsible for securing various components operating within other layers of the Big Data platform
This layer provides functionality for authentication, authorization and confidentiality via the encryption of at-rest and in-motion data
Storage of unstructured data generally involves access scenarios where partial updates of data are not performed and specific data items (records) always accessed in their entirely, such as an image or user session data
As ilustrated in the upcoming diagram, the ____________ houses the security engine. The security features provided by this layer are primarily used to secure the data acquisition layer, storage layer, processing layer and analysis layer
This layer provides functionality for authoring, applying and managing security policies as well as monitoring resource access via auditing
A _________ is a tool with features for performing a range of common ___________ task in a centralized manner
One of the main objectives of this layer is to ensure that only the intended user with the correct access level can access the requested resources, such as the storage device or the processing engine
The _________ intercepts access requests, which are made from enterprise application and systems using different security schemes, for resources within the Big Data platform
By acting as an intermediary, the __________ provides seamless access to the Big Data platform in a secure manner without the need for custom integration
what kind of encryption should be used for data at-rest and in-motion
the integration of new components or tools within the Big Data platform
the evolution of the Big Data platform
Big Data platform
A __________ is a set of technologies that collectively provide Big Data storage and processing capabilities
Instead of directly using the data transfer engine, it can be indirectly invoked via a productivity portal which normally denotes ad-hoc usage
A ___________ is a data-driven workflow consisting of task where each task involves the input , operation and output
To achieve low latency data access, a memory based storage device can be used instead, however, this increases the cost of setting up the Big Data platform
Data Pipeline
Each _________ consists of multiple tasks joined together in a sequencial manner such that the output of the previous task becomes the input of the next task. Such a combination of tasks denotes an single stage
In such a situation, the _________ pattern can be applied, which requires dividing the ________ into multiple simple steps. This is executed over multiple processing runs
A __________ can be very simple, consisting of a single stage or very complex, consisting of multiple stages
Data flowing in at high speeds needs to be captured instantly so that it can be processed without any delay for obtaining maximum value
The entire set of activities, from data ingestion to data egress, can be thought of as a _______ involving a range of operations from data cleasing to the computation of a statistic
Depending on the required functionality, a _______represents a partial or a complete Big Data solution in support of Big Data analysis
The __________ compound pattern represents a part of a Big Data solution environment capable of ingesting high-volume and high-velocity data from a range of structured, unstructured and semi-structured data sources
Relational Source (core)
File-Based Source (core)
Streaming Source (core)
Fan-in ingress (optional)
Fan-out ingress (optional)
this can be enabled through the application of the ___________ design pattern, which involves the use of a relational data transfer engine
such storage devices implement functionality that automatically creates replicas of a dataset and copies them on multiple machines
a relational data transfer engine is used to extract data from the relational database based on an SQL Query that internally uses connectors for connecting to different relational databases
The ___________ design pattern is generally applied when data needs to be extracted from internal OLTP systems, operational systems, such as CRM, ERP and SCM systems, or data warehouses
File-based Source
Finding hidden insights generally involves analyzing unstructured data, such as textual data, from internal as well as external data sources
Acquisition of large amounts of unstructured data from a variety of data sources can be automated through the application of the _____________ design pattern
Apart from textual files, images, audio and video files can also be imported through the application of this design pattern
The ___________ pattern is associated with the data transfer engine (file), storage device, workflow engine and productivity portal mechanisms
The _________ pattern covers both human and machice-generated data and deals exclusively with unstructured and semi-structured data
The Realtime Access Storage design pattern is often applied in combination with the __________ design pattern when high velocitity data needs to be analyzed in realtime
Based on the supported feature set, an event data transfer engine may provide some level of in-flight data cleansing and simple statistic computation, such as count, min, max functionality
The _________ pattern is associated with the data transfer engine (event), storage device (in-memory) and productivity portal mechanisms
The _________ compound pattern represents a part of a Big Data solution environment capable of storing high-volume, high-velocity and high-variety data for both streaming and random access
Random Access Storage (core)
Streaming Access Storage (core)
Realtime Access Storage (core)
Data Size Reduction (optional)
Confidential Data Storage (optional)
A Big Data platform generally consists of a cluster environment built using commodity hardware, which increases the chances of hardware failure
An entire dataset can be lost in the machine that saves the dataset becomes inavailable due to a hardware failure
To make sure that data is not lost and clients can still have access if there are hardware failures, the __________ pattern can be applied, which requires the use of eithera distributed file system or a NoSQL database
The __________ pattern is also applied whenever the Dataset Decomposition and Automatic Data Sharding patterns are applied
In a Big Data solution environment where large amounts of data get accumulated in a short amount of time, storing data get accumulated in a short amount of time, storing data in its raw form can quickly consume available storage and may require continuous addition of storage devices to keep increasing storage capacity
Also the requirements of keeping all data online and maintaining redundant storage for fault-tolerance entail more storage space
The _________ pattern is applied in these situations to reduce the storage footprint of data, make data transfer faster and decrease data storage cost
The application of the __________ pattern mainly requires the use of a compression engine
Although reducing storage footprint, the application of this pattern can increase the overall processing time, as data first needs decompressing. Hence, an efficient compression engine needs to be employed
However, the storage capacity does not need to be increased proportionally if a data compression engine is introduced. As a result, IT spending only increases slightly
The complete data processing cycle in Big Data environments consists of a number of activities, from data ingress to the computation of results and data egress
Although raw data is normally accessed in sequential manner, processed data requires non-sequential access such that specific records, identified via a key or a field, can be accessed individually
To enable random write and read of data, The __________ pattern can be applied, which involves the use of a storage device in the form of a NoSQL database
Such data can be treated as a BLOB that is only accessible via a unique key
By acting as an intermediary, the ________ provides seamless access to the Big Data platform in a secure manner without the need for custom integration
Note that some level of post-processing may be required to put the file in the required format before it can be copied over the target location
The ___________ pattern is associated with the storage device (key-value) and serialization engine mechanisms
In some cases, each column may itself contain a number of other columns
A relational database cannot be used in such circumstances due to a limit on columns and the inability to store more than one value in a column
The ___________ pattern can be applied to store such data, which stipulates the use of a storage device implemented via a column-family NoSQL database servicing insert, select, update and delete operations
The use of a column-family database enables storing data in more traditional, table-like storage, where each record may further consist of a logical groups of fields that are generally accessed together
a file data transfer engine is used directly or indirectly through the productivity portal for ad-hoc exports
The _________ pattern can be applied to store data consisting of linked entities. This pattern is typically implemented via the use of a storage device based on a graph NoSQL database that enables defining relationships between entities
The _________ pattern is associated with the processing engine, storage device, workflow engine, resource manager and coordination engine mechanisms
The use of graph NoSQL databases enables finding clusters of connected entities among a very large set of entities, investigating if entities are connected together or calculating distances between entities
The ________ pattern is associated with the storage device (graph) and serialization engine mechanisms
Due to the nature of the deployed processing engine, it may not be possible to execute the entire logic as a single processing run. Even if it were possible to do so, the testing, debugging and maintenance of the logic may become difficult
Semi-structured data conforming to a nested schema often requires storage in a way such that the schema structure is maintained and sub-sections of a particular data item (record) can be individually accessed and updated
The ________ pattern can be applied in circumstances where data represents a document-like structure that is self-describing and access to individual elements of data is required
The ________ pattern is associated with the storage device (document) mechanisms
The _______ pattern is normally applied together with the Automatic Data Replication and Reconstruction pattern so that shards are not lost in the case of a hardware failure and so that the database remains available
this pattern is enabled via a NoSQL database that automatically creates shards based on a configurable field in the dataset ans stores the shards across different machines in a cluster
As the dataset is distributed across multiple shards, the query completion time may be affected if the query requires collating data from more than one shard
The _______ pattern is associated with the storage device mechanism
To make sure that data is not lost and clients can still have access if there are hardware failures, the __________ pattern can be applied, which requires the use of either a distributed file system or a NoSQL database
Streaming Access Storage
The _______ compound pattern represents a part of a Big Data solution environment capable of storing high-volume and high-variety data and making it available for stream access
A large proportion of data processing tasks in Big Data involves acquiring and processing data in batches. When processing data in batches, sequential access to data is critical to timely processing. Therefore a storage device does not need to provide random access to the data, but rather ___________
The _______ pattern can be applied in a scenario whereby data needs to be retrieved in a streaming or sequential manner
The application of this design pattern requires the use of a storage device that provides non-random write and read capabilities and is generally implemented via a distributed file system
Storing large datasets as a single file does not lend itself to the distributed processing technologies deployed within the Big Data solution environment
Distributed processing technologies work on the principle of divide-and-conquer, requiring a dataset to be available as parts across the cluster
This can be achieved through the application of the __________ pattern, which requires the use of a distributed file system storage device
The _________ pattern is associated with the storage device and processing engine mechanisms
The ___________ compound pattern represents a part of a Big Data solution environment capable of handling the range of distinct requirements of large-scale Big Data dataset processing
Large-Scale Batch Processing (core)
Large-Scale Graph Processing (core)
High Velocity Realtime Processing (core)
Processing Abstraction (optional)
Automated Processing Metadata Insertion (optional)
Complex Logic Decomposition (optional)
Cloud-based Big Data Processing (optional)
The application of the __________ pattern enforces processing of the entire dataset as a single processing run, which requires that the batch of data is amassed first in a storage device and only then processed using a batch processing, such as MapReduce
Although computed results are not immediately available, the application of this pattern enables a simple data processing solution, providing maximum throughput
The _________ pattern is associated with the processing engine (batch), data transfer engine (relational/file), storage device (disk-based), resource manager and coordination engine mechanisms
the set of operations that need to be performed on the data is specified as a flowchart that is then automatically executed by the workflow engine at set intervals
Computing results for certain data processing jobs involves executing _________, such as finding the customer with the maximum spend amount based on transaction data for a large number of customers
However, the nature of such visualizations is different and more analysis-specific
The architecture of the entire Big Data platform that enables the execution of multiple Big Data solutions
File-based Sink
Streaming Egress
The majority of enterprise IT systems use relational databases as their storage backends
However, the method of incorporating data analysis results from a Big Data solution into such systems by first exporting results as a delimited file and then importing into the relational database takes time, is error-prone and is not a scalable solution
the ________ pattern can be applied for directly exporting processed data to a relational database, which requires the use of a relational data transfer engine
A workflow engine can be used to automate the whole process and to perform data export at regular intervals
The _________ pattern is associated with the data transfer engine (relational), storage device, processing engine, productivity portal and workflow engine mechanisms
On some occasions, data analysis results from a Big Data solution need to be incorporated into enterprise IT systems that use proprietary storage technologies, such as an embedded database or ________ storage, rather than using a relational database and provide a ________ import method
Like the Relational Sink pattern, manual export from the Big Data platform and import into such systems is not a viable solution
The _________ pattern is associated with the data transfer engine (file), storage device, processing engine, productivity portal and workflow engine mechanisms
Performing data processing activities manually is time-consuming and an inefficient use of development resources
To enable the automatic execution of data processing activities, the _________ pattern can be applied by implementing a workflow engine
This pattern can also be applied together with the Complex Logic Decomposition pattern to automate the execution of multiple processing jobs
The ________ pattern is associated with the workflow engine, data transfer engine, storage device, processing engine, query engine and productivity portal mechanisms
