Cloud, Semantic Web
PA160, spring 2022
Martin Kuba, ÚVT MU
makub@ics.muni.cz
Outline
● Scalability
● Cloud Computing
○ IaaS
○ PaaS
○ SaaS
● Deploying apps to cloud
○ Containers, Docker, Kubernetes
○ Infrastructure as Code, Ansible
○ Continuous Integration / Continuous Delivery
● Semantic Web
○ expressing semantics
○ RDF, OWL, ontologies
○ Semantic Web Services
Load spike problems
● on 9/11 2001 news sites like cnn.com or
bbc.com could not handle the load
● recent debacles of Czech e-government
○ highway e-vignetes (2 x 106
users)
○ on-line census (107
users)
● for scale:
○ Facebook has 1.85 x 109
daily users
○ 70,000 Google searches per second
○ seznam.cz has 2.5 x 106
users monthly, 6 x 105
daily
Scalability
● property of a system to handle a growing
amount of work by adding resources to the
system
● scaling vertically means adding resources
to a single computer (CPUs, memory)
● scaling horizontally means adding more
computers to a distributed software
application
How to scale
● make modular and loosely-coupled systems
● use caches wherever possible
● use CDN - Content Delivery Network - for
static files like images, videos, CSS, JS
○ routes traffic to the nearest or fastest datacenter
● use load balancers between clients and
servers
● for database scaling use replication and/or
sharding (horizontal partitioning of tables)
● Google search used Map-Reduce
● automatically add or remove computers from
a computing cloud (elasticity)
Cloud Computing
Cloud Computing
● use of computing resources (hardware and
software) that are delivered as a service
over a network
● in 1960’s called “utility computing”
● the retailer Amazon was dealing with peak
traffic during Christmas seasons
● in 2006 Amazon released AWS - Amazon
Web Services
○ EC2 - Elastic Compute Cloud
○ S3 - Simple Storage Service
Cloud definition
● cloud computing is a general term for
anything that involves delivering hosted
services over the Internet
● definition by NIST (National Institute of Standards and Technology, U.S.
Department of Commerce): Cloud computing is a model for enabling
ubiquitous, convenient, on-demand network access to a shared pool of
configurable computing resources (e.g., networks, servers, storage,
applications, and services) that can be rapidly provisioned and released
with minimal management effort or service provider interaction.
● five essential characteristics: on-demand self-service, broad network
access, resource pooling, rapid elasticity and measured service
Cloud Service Models
● Software as a Service (SaaS)
● Platform as a Service (PaaS)
● Infrastructure as a Service (IaaS)
SaaS - Software as a Service
● best known to computer users, the only cloud
they directly use, provides device independence
● on-demand access to software, either to
downloadable code executed on client
computers, or through remote API calls to code
executed on servers
● examples:
○ web mail – GMail, Outlook.com, email.cz
○ maps - Google Maps, Google Earth, mapy.cz
○ file services – Dropbox, Google Drive, OneDrive, pCloud
○ video libraries – YouTube, Vimeo, Netflix
○ music libraries - Spotify, Apple Music, Amazon Music
○ office tools – Google Docs, Microsoft Office 365
○ communication tools – zoom.us, Google Meet, WebEx
○ business software – Salesforce, NetSuite
PaaS - Platform as a Service
● platform is a software environment used to
develop and run applications
● not visible to end users, targeted to application
developers and maintainers delivering their SaaS
applications
● the environment includes concrete programming
languages, their specific libraries, and additional
services like SQL, noSQL, newSQL storage
● examples:
○ Amazon Elastic Beanstalk (Go, Java, Node. js, PHP, Python, Ruby)
○ Google App Engine (Go, Java, Node.js, PHP, Python)
○ Microsoft Azure Websites (Java, Node.js, PHP, Python, .NET)
○ Heroku (Go, Java, Node.js, PHP, Python, Ruby, Scala, Clojure)
IaaS - Infrastructure as a Service
● provides a virtual data center
● IaaS provider provides virtual machines
(VMs) with complete operating systems
● many VMs can be hosted on a single
physical machine running hypervisor
software (Xen, KVM, VMware)
● resources hired from an IaaS cloud can be
used directly (e.g. on-demand movie
rendering) or as a layer under a PaaS or
SaaS cloud
IaaS Providers Provide
● disk images with pre-installed popular
operating systems (various versions of
Linux, MS-Windows)
● networking services - virtual local area
networks, virtual private networks, IP
addresses, firewalls, load balancers, domain
name service (DNS)
● storage services - virtual block storage, file
storage, object storage, relational database
storage, noSQL storage, tape archive
storage, content delivery network (CDN)
IaaS Examples
● providers:
○ Amazon Elastic Compute Cloud
○ Google Compute Engine
○ Microsoft Azure
○ IBM Cloud
○ Digital Ocean
● software:
○ OpenStack (used at https://cloud.muni.cz/)
○ VMware vSphere (used at https://vcenter.ics.muni.cz/)
○ OpenNebula (used at https://stratus.fi.muni.cz/)
○ Apache CloudStack
○ Eucalyptus
OpenStack - creating a new VM
OpenStack - creating a new VM (2)
Cloud Service Models Summary
● Software-as-a-Service model provides on-demand
access to software
● Platform-as-a-Service model provides on-demand
software environment for deploying applications
● Infrastructure-as-a-Service model provides on-demand
resources from a virtual data center
Serverless computing
● also Function-as-a-Service - FaaS
● execute code as an event driven function
without managing the underlying server
resources
● FaaS extends PaaS with complete
abstraction from the underlying infrastructure
through a pay-per-execution billing model
● examples:
○ AWS Lambda (introduced 2014)
○ Google Cloud Functions (2016)
○ Microsoft Azure Functions (2016)
● hard to debug, strong vendor lock-in
Docker containers
● real Virtual Machines have some overhead
● Docker is a tool for deployment of software in
so-called containers
● a container is an isolated environment with complete
system libraries, running inside a hosting OS
● a container is in principle a chroot directory with
cgroups and namespaces, with exportable directories
and TCP ports for linking to other containers
● a container image can contain e.g. OS Ubuntu 20.04
with Apache web server, but can run on any Linux,
MacOS X or Windows host
Docker containers (2)
● Docker container images are like .deb or .rpm
packages, but OS-independent
● anybody can create a new container image by
modifying another container image
● running containers can be linked using TCP ports
(e.g. Apache+PHP -> PostgreSQL)
● containers are ephemeral (short-lived), disposable
● containers do not store persistent data
● persistent data can be stored outside of containers
using mounted directories (called volumes)
Containers vs. Virtual Machines
● each VM has its own kernel, containers
share a kernel from the host
● a VM runs many processes (daemons, e.g.
sshd, crond), a container only one process
● update of a container is done by replacing it,
in VM by making permanent changes in it
Docker Container Registries
● DockerHub at https://hub.docker.com/
○ the default registry preconfigured in Docker
○ official containers for famous software (e.g. PostgreSQL,
Apache, Ubuntu, Debian, OpenJDK,…)
○ image names without any prefix, e.g.
■ postgres:14
■ httpd:2.4.52
■ ubuntu:22.04
● GitHub, GitLab
○ each git repository has associated Docker registry
○ image names contain registry, e.g.
■ docker.pkg.github.com/cesnet/perun/perun_rpc:3.9
■ registry.gitlab.ics.muni.cz:443/perun/perun_docker/perun_rpc:3.9
Kubernetes
● container-orchestration system
● platform for automating deployment, scaling,
and operations of application containers
across clusters of hosts
● provisions both (stateless) containers and
their persistent data
● released by Google in 2015
Pets and cattle
● machines can be treated in two ways
● pets - you are emotionally attached to them,
because you have spent a lot of effort setting
them up, you feel terrible when losing them,
you give them names
● cattle - use them, then kill them when not
needed; numbered, not named
● cloud machines are cattle
Infrastructure as Code (IaC)
● managing computer data centers through
machine-readable definition files
● no interactive configuration tools
● necessary when managing hundreds of
machines
● definitions may be in a version control
system (e.g. git)
● tools like Ansible, Puppet, SaltStack, Chef
● declarative approach defines the desired
state of machines
Push and Pull Configuration Strategies
● Push - configurations of machines are
changed from outside after an event
○ used by Ansible
○ scales poorly to huge numbers of machines
● Pull - a machine downloads and applies a
configuration periodically
○ Puppet downloads from a central server
○ ansible-pull downloads from a git repository
○ scales well
○ may be a problem when managing differing
machines
Idempotent actions
● idempotent - can be run repeatedly with the
same result as when run just once
● machine configuration actions should be
idempotent
● if a machine already is in the desired state,
nothing is changed
● e.g. adding a line to a file is not idempotent
● adding a line to a file only if it is not yet
present is idempotent
Ansible
● configuration management tool written by
RedHat in Python
● configuration is written in a YAML-formatted
file called a playbook
● a playbook contains one or more plays
● a play declares
○ a list of managed machines (hosts)
○ the user making the changes (remote_user)
○ variables (vars)
○ executed actions (tasks)
○ handling of events (handlers) (e.g. service restarts)
Ansible - playbook example
Ansible - executed playbook
Ansible features
● highly customisable by extensions
● in addition to physical and virtual machines,
it can also modify Docker containers
● can manage Docker servers
● can manage (create, destroy) virtual
machines in a cloud (OpenStack, Amazon,
Azure, Google, and many others)
● can manage networks and network services
CI/CD - Continuous Integration /
Delivery / Deployment
● CI - the practice of frequently integrating new or
changed code with the existing code repository
● used in combination with automated tests
● detects errors as quickly as possible, reduces
integration problems
● builds triggered by every commit to a repository
● C. Delivery - changes are automatically bug
tested and uploaded to a repository
● C. Deployment - automatically releasing a
developer’s changes from the repository to
production
Famous CI/CD Tools
● Travis CI - since 2011, service for CI
of projects hosted at GitHub,
activated by file .travis.yml
● GitLab CI - since 2016, for own
projects, most flexible, activated by
file .gitlab-ci.yml
● GitHub CI - since 2019, for own
projects, activated by files
.github/wokflows/*.yml
GitLab CI/CD
● CI/CD pipeline consists of stages (build,
test, deploy) which consist of jobs
● jobs are simple linux shell commands that
return non-zero status when they fail
● a pipeline is run by a GitLab Runner
inside a specified Docker container
● a GitLab Runner can run anywhere, on
a dedicated server or on user’s notebook
● built packages (e.g. .jar files) or containers
are uploaded to a registry
Happy CI/CD scenario
1. a developer pushes a commit into a git repo
2. CI runs automated tests
3. CI builds a new version of a container with
the software
4. CD uploads the container to registry
5. CD deploys the container into Kubernetes
cluster
6. new version of the software is running
Hands-on experience:
● example of CI/CD of a web service
https://github.com/martin-kuba/muni-pa160-c
hat-service
● example of Ansible deployment of the web
service
https://github.com/martin-kuba/muni-pa160-a
nsible
Semantic Web
Semantic Web
● idea introduced by Tim Berners Lee
(inventor of WWW) in 2001
● “The Semantic Web is not a separate Web
but an extension of the current one, in which
information is given well-defined meaning,
better enabling computers and people to
work in cooperation”
● web instead of platform for distributed
presentations would be platform for
distributed knowledge
Semantics
● semantics = meaning
● semantic in SW means machine-processable
● semantic continuum (Uschold 2003)
a. implicit semantics in the minds of humans
b. explicit informal semantics (text description in natural
language, e.g. HTML specification)
c. formal semantics for humans (in formal language
processed by humans, e.g. modal logic)
d. formal semantics for machine processing
● the goal is to create robotic decision-making
devices
● a form of Artificial Intelligence
Michael Uschold - Where Are the
Semantics in the Semantic Web?
Syntax vs Semantics
● syntax refers to grammatical structure,
defines allowed strings of symbols
● semantics refers to the meaning of the
vocabulary symbols arranged with that
structure
● Syntax describes the way to construct a
correct sentence. For example, “this water is
triangular” is syntactically correct.
● Semantics relates to the meaning. “this
water is triangular“ does not mean anything,
though the grammar is ok.
Expressing Semantics
● folksonomies
● microdata, JSON-LD
● RDF triples and RDF Schema vocabularies
● OWL-DL ontologies for automated reasoning
Folksonomies
● keyword metadata as tags
● e.g. an image of a dog may be tagged with
tags dog, collie or pet
● (+) low entry barrier, no user training
● (-) no synonym control, flat structure
● tag clouds - tag size proportional to usage
Microdata
● competing Microdata, Microformats, RDFa
● nesting semantics within existing content on web
pages
● Microdata provides JavaScript API
● Microdata use namespace-qualified vocabularies
predefined at data-vocabulary.org or schema.org
● supported by Google search engine
● opposite vision than in 2000:
○ XML with CSS or XSLT - semantic markup with
presentational metadata
○ HTML5 with Microdata - presentational markup with
semantic metadata
Comparison of Microdata and others
JSON-LD
● Microdata are deprecated since 2016
● Microdata DOM API was deprecated since
2018 in Mozilla, removed since Firefox 49
● JSON-LD (JavaScript Object Notation for
Linked Data) replaces them
● W3C Recommendation since 2014
● Google parses JSON-LD and uses them in
searches
Example of JSON-LD
RDF - Resource Description Framework
● statements about web resources
● triples subject-predicate-object
● subject and predicate are URIs
● object can be a URI or a data value
● reification - an RDF statement is assigned a
URI and treated as a resource
● producers and consumers of RDF
statements must agree on the semantics of
the resource identifiers, conveyed by some
controlled vocabulary
RDF Schema
● tool for defining controlled vocabularies
● defines
○ classes of things
○ properties (binary predicates)
○ subsumption relationships (subclasses, subproperties)
○ rdf:type - resource is an instance of a class
● SPARQL (SPARQL Protocol and RDF Query
Language) is an SQL-like language for
querying RDF graphs
● entailment rules allow to entail e.g. that when
a resource is in a particular class, then it is
also in all its superclasses
RDF Schema example
● RDFS can define two classes:
○ Person
○ Student as subclass of Person
● a RDF statement may state that a resource
representing John Doe is of rdf:type Student
● by entailment, John Doe is also a Person
OWL
● Web Ontology Language defined by W3C
● ontology is a term from artificial intelligence
● ontology is “an explicit (written) formal
conceptualization”, used for capturing
knowledge about some domain of interest
● “conceptualization is an abstract simplified view of some selected part of the world, containing
the objects, concepts, and other entities that are presumed of interest for some particular
purpose and the relationships between them”
● OWL 1 released in 2004, OWL 2 in 2009
● two different (incompatible) semantics
○ RDF based - OWL Full
○ DL (Description Logics) based - OWL DL
Types of logic
● propositional (Boolean) logic
○ formulae made of atomic propositions with values
true or false, and logical connectives like negation
(¬A), and (A∧B), or (A∨B) and implication (A→B)
○ sound, complete and decidable in finite time
● predicate logic
○ adds predicates, quantifiers, terms
○ formulae look like ∀x∃y(P(x)→Q(f(y)))
● first order predicate logic
○ quantifiers can range only over elements of sets
○ sound and complete, but not decidable
● description logics
○ logics designed to be as expressive as possible
while retaining decidability
OWL DL
● Description Logics is a decidable fragment of
First Order Predicate Logic (FOPL) plus
decidable extensions
● reasoners - software able to entail (decide)
complete inferable knowledge in finite time
● OWL DL ontology is a set of assertions about:
○ classes
○ individuals
○ properties (binary relations)
■ object properties (between two objects)
■ data properties (between object and data literal)
● can use SWRL (Sem. Web Rule Language)
Prefix(:=<http://goo.gl/kWFWj#>)
Prefix(xsd:=<http://www.w3.org/2001/XMLSchema#>
Ontology(<http://goo.gl/kWFWj>
Declaration( Class( :Person ) )
Declaration( Class( :MarriedPerson ))
Declaration( NamedIndividual( :Martin ) )
Declaration( NamedIndividual(:Lenka ) )
Declaration( ObjectProperty(:hasSpouse ) )
Declaration( DataProperty(:hasEmail ) )
SymmetricObjectProperty( :hasSpouse )
FunctionalObjectProperty( :hasSpouse )
ClassAssertion( :Person :Lenka )
ClassAssertion( :Person :Martin )
DifferentIndividuals( :Martin :Lenka )
ObjectPropertyAssertion( :hasSpouse :Martin :Lenka )
DataPropertyAssertion( :hasEmail :Martin "makub@ics.muni.cz"^^xsd:string )
SubClassOf( :MarriedPerson :Person )
EquivalentClasses( :MarriedPerson ObjectSomeValuesFrom( :hasSpouse :Person ))
Example of OWL DL ontology
OWL DL Tools
● ontology editor with GUI - Protege
○ https://protege.stanford.edu/
● reasoners
○ Pellet
○ HermiT
○ FACT++
○ Stardog
● Java API for OWL - OWL API
○ https://github.com/owlcs/owlapi/
Limits of OWL DL
● DL is based on FOPL ∀x∃y(P(x)→Q(f(y)))
● thus DL cannot express:
○ fuzzy expressions - “It often rains in autumn.”
○ non-monotonicity - “Birds fly, penguin is a bird, but penguin does not fly.”
○ propositional attitudes - “Eve thinks that 2 is not a prime number.”
○ modal logic
■ possibility and necessity - “It is possible that it will rain today.”
■ epistemic modalities - “Eve knows that 2 is a prime number.”
■ temporal logic - “I am always hungry.”
■ deontic logic - “You must do this.”
● Transparent Intensional Logic (TIL)
○ can express anything that can be said
○ has no calculus or reasoning algorithms
Semantic Web Services
● research efforts OWL-S, WSDL-S, WSMO
● semantics can enhance discovery
○ on the semantic continuum move it from b) to d)
○ e.g. search for "getHardDriveQuote" can find also
"getQuoteForHardDrive" (synonym) and "getSCSIDriveQuote"
(subsumed term)
● web service semantics
○ Data semantics - it defines meaning of the data, i.e. inputs and outputs of operations
○ Functional semantics - it defines meaning of the operations, i.e. how they transform input
to output
○ QoS semantics - it provides meaning for quality aspects, like price, availability, level of
trust etc. Service selection may be based on such characteristics.
○ Execution semantics - it provides details like preconditions and effects of service
invocation, conversation patters of service invocation etc
That’s it.
Thank you for your attention