Adv Topics: MapReduce and Incremental Computation

Data usually gets update on a regular basis. Connolly and Begg (2014) defined that data can be updated incrementally, only small sections of the data, or can be updated completely. An example of data that can be updated incrementally are webpages, computer codes, stale data, data-at-rest, bodies of knowledge, etc. Whereas, some examples of data that can be updated completely are: weather data, space weather data, social media data, data-in-motion, dynamic data, etc. Both sets of data provide their own unique challenges when it comes to data processing. On average, analyzing web data, new to old data can range from 10-1000x (Sakr, 2014). Thus, the focus of this discussion is on incremental data update and how to process data in between two data processing runs.

Incoop is an extension of Hadoop to allow for processing incremental changes on big data, by splitting the main computation to its sub-computation, logging in data updates in a memoization server, while checking the inputs of the input data to each sub-computation (Bhatotia et al., 2011; Sakr, 2014). These sub-computations are usually mappers and reducers (Sakr, 2014). Incremental mappers check against the memoization servers, and if the data has already been processed and unchanged it will not reprocess the data, and a similar process for incremental reducers that check for changed mapper outputs (Bhatotia et al., 2011).

Subsequently, MapReduce is an analytical engine and pattern that takes advantage of distributed systems while keeping the processes and data in one machine (Sadalage & Fowler, 2012). There are a few key principles to using the MapReduce framework and Hadoop efficiently to improve incremental computation:

  • Data partitioning: The MapReduce framework aids in partitioning the data into similar size sets into Hadoop Distributed File System, aka HDFS (Lublinsky, Smith, & Yakubovich, 2013). Thus, MapReduce can support smaller sets of data stored in HDFS. This is part of the scalability of the cluster.
  • Fault tolerance and durability: Given that data can be partitioned to tiny chunks across thousands of computations nodes and run in parallel sometimes these nodes can fail (Sakr, 2014). The MapReduce framework replicates the data in the background and can launch backup jobs if a node fails (Lublinsky et al., 2013; Sakr, 2014). Thus, failure doesn’t disrupt the data processing. However it does increase the number of processors needed (Connolly & Begg, 2014).
  • Parallelization: The partitioned input data are considered as independent sets of data, such that the mapper functions can process the data in a parallel environment (Lublinsky et al., 2013; Sakr, 2014). This principle allows for the sub-functions within the mapper and reducer function to handle smaller data. It allows a data analyst to focus on the main problem rather than low-level parallel coding abstraction, like multithreading, file allocation, memory management, etc. (Sakr, 2014). Serialization does not allow for small incremental updates for large data (Connolly & Begg, 2014).
  • Data reuse: There is no need to read or write of intermediate data, thus preserving the input data to enable the data to be reused because it is unchanged (Lublinsky et al., 2013; Sakr, 2014).
  • Self-Adjusting Computation: Used for incremental computation, which only allows mappers and reducers only work on the smaller size sets of data that are impacted by the change (Sakr, 2014).

Both Bhatotia et al. (2011) and Sakr (2014), suggested an Inc-HDFS which is also an extension of the HDFS, for partitioning data based on content and removal of data duplication. There is the limitation of this approach where the number of files may be grouped in too many content bins or too little content bins and thus may not be evenly be spaced out (Bhatotia et al., 2011). Thus, invoking: too many mapper functions can create an infrastructure overhead, which increases resources and thus cost, or too few mapper functions can create huge workloads for certain types of computational nodes, or too many reducers can provide too many outputs, and too little reducers can provide too little outputs (Lublinsky et al., 2013; Sakr, 2014). A constraint must be added on both ends of the spectrum to allow for evenly distributed data sets (Bhatotia et al., 2011).

Subsequently, the Hadoop out-of-the-box product scheduler doesn’t account for memoization server data, therefore is not built for incremental analysis. Thus, Incoop has a memoization-aware scheduler, that schedules the sub-computations based on the affinity of the task and allows for efficient use of previously (Bhatotia et al., 2011; Sakr, 2014). The scheduler can run tasks on computational nodes that are either faster or locally to where the data is stored (Bhatotia et al., 2011). Using this type of scheduler, the scheduler should place priority on whether it is faster to conduct a data movement and run the task at a faster computational node or reduce data movement, and processes the data locally, while still making effective use of unchanged and processed data.

In the end, a practical application of this technique when it comes to analyzing web data would be to first partition the web data by its content. Scientific content can go in one context partition, corporate financial content into another context partition, etc. under the Inc-HDFS framework. These partitions are capped in size to allow for proper load balance. Incoop will then run the MapReduce function to process the data distributively through using parallel processes. If the data gets updated, like a new corporate update to the SEC 10K data, it will be recorded by the memoization server. This will allow for Incoop to be able to process that incremental change from the corporate financial content partition because it was using the memoization-aware scheduler, and reprocess the data through mapping and reducing function on just this small and partitioned dataset. Therefore, making effective use of unchanged and processed data.

Resources:

  • Sadalage, P. J., Fowler, M. (2012). NoSQL Distilled: A Brief Guide to the Emerging World of Polyglot Persistence, (1st ed.). Vitalbook file.
  • Sakr, S. (2014). Large Scale and Big Data, (1st ed.). Vitalbook file.ok
Advertisements

Adv Topics: MapReduce and Iterative computation

Data-in-motion is the real-time streaming of data from a broad spectrum of technologies, which also encompasses the data transmission between systems, while data that is stored on a database system or cloud system is considered as data-at-rest and data that is being processed and analyzed is considered as data-in-use (Katal, Wazid, & Goudar, 2013; Kishore & Sharma, 2016; Ovum, 2016; Ramachandran & Chang, 2016). Social media data or social network analysis data can be considered as data-in-motion and processing that type of data can be quite problematic. Data-in-motion has to be iteratively processed until there is a certain termination condition is reached and it can be reached between iterations (Sakr, 2014). Data-at-rest is probably considered easier to analyze; however, this type of data can also be problematic. If the data-at-rest is large in size and even if the data does not change or evolve, its large size requires iterative processes to analyze the data.

An example of an iterative process as suggested by Lusblinsky, Smith, and Yakubovich (2014), is solving a linear equation by approximation algorithms on hundreds of equations and variables. The data can be stored in matrix Ax = b, where A is a matrix of coefficients, b is a vector of output values, and x is a vector of variables. If the data is too large, using a simple linear algebraic solution would be impossible, so a quadratic spline solution would consist of the following:

f(x) = ½ xTAx-xT­b

and a superscript “T” represents transposing the vector or matrix. Each iteration of this spline would result in a better vector solution. However, Sakr (2014) stated that MapReduce does not support iterative data processing and analysis directly. Thus workaround is needed to handle iterative programs for situations like data-in-motion or even streaming data.

 Root causes and technical steps to address them

To deal with datasets that require iterative processes to analyze the data, computer coders need to create and arrange multiple MapReduce functions in a loop (Sakr, 2014). This workaround would increase the processing time of the serialized program because data would have to be reloaded and reprocessed, because there is no read or write of intermediate data, which was there for preserving the input data (Lusblinksy et al., 2014; Sakr, 2014). The root cause exists because in its simplest form MapReduce can consist of many mappers and one reducer, which is a performance bottleneck. This simplified model of the MapReduce analytical engine means one cannot reduce across keys, just one key at a time (Sadalage & Fowler, 2012). Thus, the algorithm is not built for iterations. HaLoop is an iteration solution built on top of the Hadoop infrastructure that has a loop control module, task scheduler, which caches invariant data from a previous iteration and uses it in future iterations (Sakr, 2014). This solution can be applied to static and unchanged data.

There are also disadvantages of using MapReduce on these types of data because too many mapper functions can create an infrastructure overhead or too many reducers can provide too many outputs (Lusblinksy et al., 2014; Sakr, 2014). Thus there has to be a basic implementation plan of effective data placement over the Hadoop cluster, to ensure proper load balance of data across the Hadoop servers (Sakr, 2014). Sometimes, there is a need to run two separate MapReduce functions, one to prepare the data by evenly distributing data across the servers and one that iteratively goes through the data (Lusblinksy et al., 2014). CoHadoop allows for data files and related files to be stored on the same server, provide a means of load balancing and fault tolerance by creating a file-level locator property (Sakr, 2014). Sakr describes the locator property as a means to keep track of where the data is stored via a unique identification number for each file in the system. This solution can also be applied to static and unchanged data.

For data-in-motion and streaming data, data has to be iteratively processed until there is a certain termination condition is reached. MapReduce Online has an approach where between the mapper and reducer function data is pipelined to allow for data processing and analysis as soon as the mappers produce their outputs (Sakr, 2014). This can run the MapReduce functions iteratively and provide relatively live reduced data outputs. Sakr, further explains that this approach is where the reducer contacts every mapper upon initiation of the scheduler and the data is temporarily stored on the pipeline, which is an in-memory buffer.

Resources:

  • Katal, A., Wazid, M., & Goudar, R. H. (2013, August). Big data: issues, challenges, tools and good practices. InContemporary Computing (IC3), 2013 Sixth International Conference on (pp. 404-409). IEEE.
  • Kishore, N. & Sharma, S. (2016). Secure data migration from enterprise to cloud storage – analytical survey. BIJIT-BVICAM’s Internal Journal of Information Technology. Retrieved from http://bvicam.ac.in/bijit/downloads/pdf/issue15/09.pdf
  • Lublinsky, B., Smith, K. T., & Yakubovich, A. (2013). Professional Hadoop Solutions. Vitalbook file.
  • Ovum (2016). 2017 Trends to watch: Big Data. Retrieved from http://info.ovum.com/uploads/files/2017_Trends_to_Watch_Big_Data.pdf
  • Ramachandran, M. & Chang, V. (2016). Toward validating cloud service providers using business process modeling and simulation. Retrieved from http://eprints.soton.ac.uk/390478/1/cloud_security_bpmn1%20paper%20_accepted.pdf
  • Sadalage, P. J., Fowler, M. (2012). NoSQL Distilled: A Brief Guide to the Emerging World of Polyglot Persistence, (1st ed.). Vitalbook file.
  • Sakr, S. (2014). Large Scale and Big Data, (1st ed.). Vitalbook file.

Adv Topics: MapReduce and Hadoop

Hadoop allows for data processing through MapReduce and it also allows for data storage (Lublinsky et al., 2014). MapReduce is an analytical engine and pattern that takes advantage of distributed systems while keeping the processes and data in one machine (Sadalage & Fowler, 2012). MapReduce thus contains two functions that work in parallel on distributed systems (Hortonworks, 2013; Sadalage & Fowler, 2012; Sakr, 2014; Sathupadi, 2010):

    1. Mappers functions create and process transactions on the system by mapping and aggregating data by key values. Mappers can read only one data record at a time.
    2. Reducers functions know what that key values are and will take all those values stored in a map to reduce the data to what is relevant. Reducers help summarize the data into a single output. This helps deal with the amount of data moving between multiple computational nodes.

Lublinsky, Smith, and Yakubovich, (2014), stated that an intermediate component of MapReduce is known as the shuffle and sort, where the data from the mapping function outputs are moved and presented to the reducer function.

Thus, MapReduce is a framework that uses parallel sequential algorithms that capitalize on cloud architecture, which became popular under the open source Hadoop project, as its main executable analytic engine (Lublinsky et al., 2014; Sadalage & Fowler, 2012; Sakr, 2014). Essentially, a sequential algorithm is a computer program that runs on a sequence of commands, and a parallel algorithm runs a set of sequential commands over separate computational cores (Brookshear & Brylow, 2014; Sakr, 2014). Thus, a parallel sequential algorithm runs a full sequential program over multiple but separate cores (Sakr, 2014). Another feature of MapReduce is that a reduced output can become another’s map function (Sadalage & Fowler, 2012). Subsequently, the advantages and disadvantages of using MapReduce are (Lusblinksy et al., 2014; Sakr, 2014):

+ aggregation techniques under the mapper function can exploit multiple different techniques

+ no read or write of intermediate data, thus preserving the input data

+ no need to serialize or de-serialize code in either memory or processing

+ it is scalable based on the size of data and resources needed for processing the data

+ isolation of the sequential program from data distribution, scheduling, and fault tolerance

– too many mapper functions can create an infrastructure overhead, which increases resources and thus cost

– too few mapper functions can create huge workloads for certain types of computational nodes

– too many reducers can provide too many outputs, and too little reducers can provide too little outputs

 – it’s a different programming paradigm that most programmers are not familiar with

 – the use of available parallelism will be underutilized for smaller data sets

Given that Hadoop is predominately known for popularizing MapReduce tasks, it is also known for its Hadoop Distributed File System (HDFS) where the data is distributed across multiple systems (Rathbone, 2013). Hadoop’s service is part of the cloud (as Platform as a Service = PaaS).  For PaaS, the end users manage the applications and data, whereas the provider (Hadoop), administers the runtime, middleware, O/S, virtualization, servers, storage, and networking (Lau, 2001). Data is broken up into small blocks, like Legos, such that they are distributed across a distributed database system and across multiple servers and can be processed across all these servers, e.g. Hadoop Cluster (IBM, n.d.).

A common example of a parallel sequential program is dynamical weather forecasting models. In dynamical weather forecasting models, there is a set of defined geodynamic, thermodynamic, and physical sequential algorithms define and evolve the main seven variables of weathers across time. For each time step, the forecasting models run these sequential algorithms over each grid point, which can represent a finite geospatial region. Each of these geospatial regions is split amongst multiple computational scores. This example expands in complexity when data has to travel between different finite geospatial regions through the boundaries, which is an example of data parallelism (Sakr, 2014). MapReduce uses the concept of data parallelism to help map and reduce data. Therefore, weather models could be considered as a loose form of MapReduce algorithm.

Resources:

Adv Topics: The Internet of Things and Web 4.0

The IoT is the explosion of device/sensor data, which is growing the amount of structured data exponentially with tremendous opportunities (Jaffe, 2014; Power, 2015). Both Atzori (2010) and Patel (2013) classified the Web 4.0 as the symbiotic web, where data interactions occur between humans and smart devices, the internet of things (IoT). These smart devices can be wired to the internet or connected via wireless sensors through enhanced communication protocols (Atzori, 2010). Thus, these smart devices would have read and write concurrency with humans, where the largest potential of web 4.0 has these smart devices analyze data online and begin to migrate the online world into the reality (Patel, 2013). Besides interacting with the internet and the real world, the internet of things smart devices would be able to interact with each other (Atzori, 2010). Sakr (2014) stated that this web ecosystem is built off of four key items:

  • Data devices where data is gathered from multiple sources that generate the data
  • Data collectors are devices or people that collect data
  • Data aggregation from the IoT, people, Radio Frequency Identification tags, etc.
  • Data users and data buyers are people that derive value out of the data

Some of the potential benefits of IoT are: assisted living, e-health, enhanced learning, government, retail, financial, automation, industrial manufacturing, logistics, business/process management, and intelligent transport (Sakr, 2014; Atzori, 2010). Atzori (2010) suggests that there are three different definitions or vision on the use of IoT, which is based on the device’s orientation:

  • Things oriented which are designed for status and traceability of objects via RFID or similar technology
  • Internet-oriented which are designed for light internet protocol where the device is addressable and reachable via the internet
  • Semantic-oriented where devices aid in creating reasoning over the data that is generated by these devices by exploiting models

Some of IoT can fall on one, two, or all three definitions or visions for IoT use.

Performance Bottlenecks for IoT

In 2016, IoT has two main issues, if it is left on its own and it is not tied to anything else (Jaffe, 2014; Newman, 2016):

  • The devices cannot deal with the massive amounts of data generated and collected
  • The devices cannot learn from the data it generates and receives

Thus, artificial intelligence (AI) should be able to store and mine all the data that is gathered from a wide range of sensors to give it meaning and value (Canton, 2016; Jaffe, 2014). AI would bring out the potential of IoT through quickly and naturally collect, analyzing, organizing, and feeding valuable data to key stakeholders, transforming the field into the Internet of Learning-Things (IoLT) from the standard IoT (Jaffe, 2014; Newman, 2016). However, this would mean a change in the infrastructure of the web to handle IoLT or IoT. Thus, Atzori (2010) listed some of the potential performance bottlenecks for IoT on a network level:

  • The vast number of internet oriented devices that will be taking up the last few IPv4 addresses, thus there is a need to move to IPv6 to support all the devices that will come online soon. This is just one version of the indexing problem.
  • Things oriented and internet oriented devices could spend a time in sleep mode, which is not typical for current devices using the existing IP networks.
  • IoT devices when connecting to the internet produce smaller packets of data at a higher frequency than current devices.
  • Each of the devices would have to use a common interface and standard protocols as other devices, which can quickly flood the network and increase the complexity of middleware software layer design.
  • IoT is vastly various objects, where each device with their function and has its way of communicating. There is a need to create a level of abstraction to homogenate data transfer and access of data through a standard process.

Proposed solutions would be to use NoSQL (Not only Structured Query Language) databases to help with collection, storage, and analysis of IoT data that is heterogeneous, lacking a common interface with standard protocols and can deal with data of various sizes. This can solve one aspect of the indexing problem of IoT. NoSQL databases are databases that are used to store data in non-relational databases i.e. graphical, document store, column-oriented, key-value, and object-oriented databases (Sadalage & Fowler, 2012; Services, 2015).

  • Document stores use a key/value pair that could store data in JSON, BSON, or XML
  • Graphical databases are use networks diagrams to show the relationship between items in a graphical format
  • Column-oriented databases are perfect for sparse datasets, where data is grouped together in columns rather than rows

Retail is currently using thing oriented RFID for inventory tracking and in-store foot traffic if installed on shopping carts to be used for understanding customer wants (Mitchell, n.d.). Thus, Mitchell (n.d.) suggested that the use of video cameras and mobile device Wi-Fi traffic could help identify if the customer wanted an item or a group of items by seeking hotspots of dwelling time, so that store managers can optimize the store layouts to increase flow and higher revenue. However, these retailers must be considering the added data sources and have the supporting infrastructure to avoid performance bottlenecks to get to reap the rewards of utilizing IoT to generate data-driven decisions.

Resources:

  • Atzori, L., Antonio Iera, A., & Morabito, G. (2010). The Internet of things: A survey. Computer Networks, 54(2). 787–2,805

Adv Topics: The architecture of the Internet

Introduction

Kelly (2007) stated that there are 100 billion clicks per day and 55 trillion links. But, the internet is very pervasive to the human kind. In 2012, 2.27 billion people used the internet. However, globally 1.7 billion people are actively engaged with the internet (Li, 2010; Sakr, 2014). An actively engaged user of the internet would be using social media, where they can watch videos, share status updates, and curated content, with a goal of actively engaging other users (Li, 2010). Cloud-based services rely on the internet to provide services like distributed processing, large scale data storage, support distributed transactions, data manipulation, etc. (Brookshear & Brylow, 2014; Sakr, 2014). Thus, the internet plays a vital role in enabling big data analysis and social engagement, given its humble beginning for storing research projects between multiple research institutions in the 1960s (Brookshear & Brylow, 2014).

Internet Architecture

The internet has evolved into a socio-technical system. This evolution has come about in five distinct stages:

  • Web 1.0: Created by Tim Berners-Lee in the 1980s, where it was originally defined as a way of connecting static read-only information hosted across multiple computational components primarily for companies (Patel, 2013). The internet was a network of computational networks and where communication that is done throughout the internet is governed by the TCP/IP protocol (Brookshear & Brylow, 2014). The internet’s architecture relies on three components: Uniform Resource Identifier (URI), Hyper Text Transfer Protocol (HTTP), and HyperText Markup language (HTML) (Jacobs & Walsh, 2004).
    • A URI is a unique address that is common across all corners of the web and agreed upon convention by the internet community, which is made up of characters and numerical values that allow an end-user of the internet to locate and retrieve information (Brookshear & Brylow, 2014; Jacobs & Walsh, 2004; Patel 2013). This unique addressing convention allows the internet to store massive amounts of information without URI collision, which is when two URI hold the same value, which can confound search engines (Jacobs & Walsh, 2004). An example of a URI could be www.mkhernandez.com.
    • For a computer to locate this URI’s information which is hosted on a web server, a web browser like Goggle chrome, Microsoft edge, or Firefox uses the HTTP protocol to retrieve the information to be displayed by the browser (Brookshear & Brylow, 2014). The web browser would send an HTTP GET request for www.mkhernandez.com via the TCP/IP port 80, and as long as the browser is given access to the information stored in the URI, then the web server sends an HTTP POST or PUT to give the web browser the sought after information (Jacobs & Walsh, 2004). In a sense, HTTP protocols play an important role in information access management.
    • Once the HTTP POST is sent from the web server consisting the information stored in www.mkhernandez.com, the web browser must now convert the information and display it on the computer screen. HTML is a <tag> based notational code that helps a browser read the information sent by the web server and display it in a simple or rich data format (Brookshear & Brylow, 2014; Patel 2013). The <html /> tags are used by the HTTP protocol to identify the content type and encoding style for displaying the information of the web server onto the browser (Jacobs & Walsh, 2004).   The <head /> and <title /> tags are used as metadata about the document. Whereas the <body /> tag contains the information of the URI, and <a /> tags helps you link to other relevant URIs easily.
  • Web 2.0: Changed the state of the internet from a read-only state to a read/write state and had grown communities that hold a common interest (Patel, 2013). This version of the web led to more social interaction, giving people and content importance on the web, due to the introduction of social media tools through the introduction of web applications (Li, 2010; Patel, 2013; Sakr, 2014). Web applications can include event-driven and object-oriented programming that are designed to handle concurrent activities for multiple users and had a graphical user interface (Connolly & Begg, 2014; Sandén, 2011). Key technologies include:
    • Weblogs (Blogs), Video logs (vlogs), and audio logs (podcasts) are all content in various styles that are published chronologically in descending time order, which can be tagged with keywords for categorization and available when people need them (Li, 2010; Patel, 2013). These logs can be used for fact-finding when content is stored chronologically (Connolly& Begg, 2014).
    • Really Simple Syndication (RSS) is a web and data feed format that summarizes data via producing an open standard format, XML file (Patel, 2013; Services, 2015). This is regularly used for data collection (Services, 2015).
    • Wikis are editable and expandable by those who have access to the data, and information can be restored or rolled back (Patel, 2013). Wiki editors ensure data quality and encourage participation from the community in providing meaningful content to the community (Li, 2010).
  • Web 3.0: This is the state the web at 2017. Involves the semantic web that is driven by data integration through the uses of metadata (Patel, 2013). This version of the web supports a worldwide database with static HTML documents, dynamically rendered data, next standard HTML (HTML5), and links between documents with hopes of creating an interconnected and interrelated openly accessible world data such that tagged micro-content can be easily discoverable through search engines (Connolly & Begg, 2014; Patel, 2013). This new version of HTML, HTML5 can handle multimedia and graphical content and introduces new tags like <section />, <article />, <nav />, and <header />, which are great for semantic content (Connolly & Begg, 2014). Also, end-users are beginning to build dynamic web applications for others to interact with (Patel, 2013). Key technologies include:
    • Extensible Markup Language is a tag based metalanguage (Patel, 2013). These tags not limited to the tags defined by other people and can be created at the pace of the author rather than waiting for a standard body to approve a tag structure (UK Web Design Company, n.d.).
    • Resource Description Framework (RDF) is based on URI triples <subject, predicate, object>, which helps describes data properties and classes (Connolley & Begg, 2014; Patel, 2013). RDF is usually represented at the top of the data set with a @prefix (Connolly & Begg, 2014). For instance,

@prefix: <https://mkhernandez.com&gt; s: Author <https://mkhernandez.wordpress.com/about/&gt;. <https://mkhernandez.wordpress.com/about/&gt;. s:Name “Dr. Michael Kevin Hernandez”. <https://mkhernandez.wordpress.com/about/&gt; s:e-mail “dr.michael.k.hernandez@gmail.com.”

  • Web 4.0: It is considered the symbiotic web, where data interactions occur between humans and smart devices, the internet of things (Atzori, 2010; Patel, 2013). These smart devices can be wired to the internet or connected via wireless sensors through enhanced communication protocols (Atzori, 2010). Thus, these smart devices would have read and write concurrency with humans, where the largest potential of web 4.0 has these smart devices analyze data online and begin to migrate the online world into the real world (Patel, 2013). Besides interacting with the internet and the real world, the internet of things smart devices would be able to interact with each other (Atzori, 2010). Sakr (2014) stated that this web ecosystem is built off of four key items:
    • Data devices where data is gathered from multiple sources that generate the data
    • Data collectors are devices or people that collect data
    • Data aggregation from the IoT, people, RFIDs, etc.
    • Data users and data buyers are people that derive value out of the data
  • Web 5.0: Previous iterations of the web do not perceive people’s emotion, but one day it could be able to understand a person’s emotional (Patel, 2013). Kelly (2007) predicted that in 5,000 days the internet would become one machine and all other devices would be a window into this machine. In 2007, Kelly stated that this one machine “the internet” has the processing capability of one human brain, but in 5,000 days it will have the processing capability of all the humanity.

Performance Bottlenecks & Root Causes

There is a trend in terms of “performance bottleneck” to access large-scale Web data as the Web technology evolves. A bottleneck is when the flow of information is slowed down or stopped in its entirety that it can cause a bad end-user experience (TechTarget, 2007, Thomas, 2012). Performance bottlenecks can cause an application to perform poorly towards expectations (Thomas, 2012). As the internet evolved, there are new performance bottlenecks that begin to appear:

  • Web 1.0: When two URI hold the same value, it can confound search engine, hence the move to IPv4 to IPv6 (Jacobs & Walsh, 2004). Transfer protocols rely on network devices: network interface card, firewall, cables, tight security, load balancers, routers, etc., which all affect the flow of information (bandwidth) (Jacobs & Walsh, 2004; Thomas, 2012). Finally, HTTP information is pulled from the web server, thus low capacity computers, broken links, tight security, poor configurations, which can result in HTTP errors (4xx, 5xx), lots of open connections, memory leaks, lengthy queues, extensive data scans, database deadlock, etc. (Thomas, 2012).
  • Web 2.0: Database demands on the web application, because there are more write and read transactions (Sakr, 2014). Plus, all the performance bottlenecks from web 1.0.
  • Web 3.0: Searching for information in the data is tough and time-consuming without a computer processing application (UK Web Design Company, n.d.). Data is tied to the logic and language like HTML without a readily made browser to simply explore the data and therefore may require HTML or other software to process the data (Brewton, Yuan, & Akowuah, 2012; UK Web Design Company, n.d.). Syntax and tags are redundant, which can consume huge amounts of bytes, and slow down processing speeds (Hiroshi, 2007).
  • Web 4.0: IoT creates performance bottlenecks but it also has its issues, if it is left on its own and it is not tied to anything else (Jaffe, 2014; Newman, 2016): (a) the devices cannot deal with the massive amounts of data generated and collected; and (b) the devices cannot learn from the data it generates and collects. Finally, there are also infrastructure potential performance bottlenecks for IoT (Atzori, 2010): (a) the huge number of internet oriented devices that will be taking up the last few IPv4 addresses; (b) things oriented and internet oriented devices could spend a time in sleep mode, which is not typical for current devices using the current IP networks; (c) IoT devices when connecting to the internet produce smaller packets of data at higher frequency than current devices; (d) each of the devices would have to use a common interface and standard protocols as other devices, which can easily flood the network and increase complexity of middleware software layer design; and (e) IoT are vastly heterogeneous objects, where each device with their own function and has its own way of communicating.
  • Web 5.0: Give the assumption of what this version of the web would become, a possible performance bottleneck would be a number of resources consumed to keep the web operating.

 

Overall, Kelly (2007) stated that there are 100 billion clicks per day and 55 trillion links and that there are 2 million emails per second, 1 million IM messages, etc. big data will impact the performance of the web. Big data will be primarily impacting the web server, because of the increasing size of information and potential lack of bandwidth, big data can slow down the throughput performance (Thomas, 2012).

High-level strategies to mitigate

The web should evolve with time to keep up with the demands and needs of society, and it is predicted change significantly. What the web evolves into is yet to be seen, but it will be quite unique. However, with multiple heterogeneous types of devices (IoT) trying to connect to the web, a standardized protocol and interface to the web should be adopted (Atzori, 2010). A move to IPv6 from IPv4 to accommodate the massive number of IoT devices that expected to generate data and connect to the web must happen to accommodate this opportunity to gain more data.

Given, Kelly (2007) centralized view of the web system, the information and data are stored distributively, and better algorithms are needed to connect relevant data and information more efficiently. Data storage is cheap, but at the rate of data creation by IoT and other sources, processing speeds through parallel and distributed processing must increase to take advantage of this explosion of big data into the web. This is the gap created by the fact that data collection is outpacing the speed of data processing and analysis. Given this gap, a data scientist should prioritize which subset of the data is valuable enough to analyze, to solve certain problems. This is a great workaround, however, it ignores the full data set at times, which was generated from a system. It’s not enough to analyze what is deemed to be valuable parts of the data, because another part of the data may reveal more insight (the whole is better than the sum of its parts argument).

NoSQL database types and extensible markup languages have enhanced how information and data are related to each other, but standardization and perhaps an automated 1:1 mapping of the same data being represented in different NoSQL databases may be needed to gain further insights faster. Web application developers should also be more resource conscious, trying to get more computational results for fewer resources.

Resources:

  • Atzori, L., Antonio Iera, A., & Morabito, G. (2010). The Internet of things: A survey. Computer Networks, 54(2). 787–2,805
  • Connolly, T., & Begg, C. (2014). Database Systems: A Practical Approach to Design, Implementation, and Management, (6th ed.). Pearson Learning Solutions. VitalBook file.
  • Sandén, B. I. (2011). Design of Multithreaded Software: The Entity-Life Modeling Approach. Wiley-Blackwell. VitalBook file.
  • Sakr, S. (2014). Large Scale and Big Data, (1st ed.). Vitalbook file.
  • Services, EMC E. (2015) Data Science and Big Data Analytics: Discovering, Analyzing, Visualizing and Presenting Data. John Wiley & Sons P&T. VitalBook file.