Hybrid Cloud
Multi-Cloud
Serverless Computing
Data Containers
Artificial Intelligence Platforms
Service mesh
Immutable Infrastructure Focused On Containers
The Internet of Things (IoT)
Cloudlet
Cloud Security
Backup and Disaster Recovery (DR)
1. Latest Trends In Cloud
Computng
How Do They Support Business
Needs?
2. Objecties
• Hybrid Cloud
• MultiCloud
• Serierless Computng
• Data Containers
• Articial Intelligence Platorms
• Seriice mesh
• Immutable Infrastructure Focused On Containers
• The Internet of Things (IoT)
• Cloudlet
• Cloud Security
• Backup and Disaster Recoiery (DR)
3. Sofware Companies
• Eiery company in eiery industry is becoming a computer
company:
i Selfidriiing cars
i Remote Medical Monitoring Deiices
i Remote Home Appliances
i Actie Packaging
i Smart Clothing
Companiesi are haiing to innoiate at a high leiel of
change in robotcs, iirtualizaton, digitzaton,
automaton
4. Science Grows Exponentally
• The iolume of medical knowledge is doubling
eiery six years
• Battery technology is improiing
• New substances are discoieredi Remember?
One new chemical substance allowed Apple to
minimize the hard driie for the original iPod,
leading to billion dollar industry
• High percent of Apple’s reienue comes from
products that did not exist a few years ago
5. Blockchain Redeines Money
• Central banks are out, and distributed ledgers are in
• Ethereum goes one step further than Bitcoin, by embedding the
contract concept of an ofer and acceptance into the real meaning
of money
• Intelligent moneyi Ether is a token whose blockchain is generated
by the Ethereum platorm (open source, public, blockchainibased
distributed computng platorm and operatng system featuring
smart contract functonality that supports a modiied iersion of
Nakamoto consensus iia transactonibased state transitons)
• But we don’t quite know how secure the massiie connectiity and
accelerated informaton may be (as Amazon Echo and Google
Home deiices exposed a new security risk when they came out)
6. Blockchain Security
• Blockchain is a contnuous growing list of records linked with hash
pointers, a tmestamp, and unchangeable transacton data
• Shared Ledger
– Append only distributed system of records shared across business networks
(peepitoipeer technology)
• Smart Contract
– Business terms embedded in transacton database
• Priiacy
– Ensuring appropriate iisibility
– Transactons secure authentcated
– Veriiable transactons
• Consensus
– All partes agree to network ieriied transactons
7. World of Amazon
• Those who haie the tools and knowledge to understand what is
going on are the ones to get aheadi Large cloud computng
proiiders are: Amazon Web Seriices, Microsof Azure Cloud
Platorm, Google Cloud, Rack Space, IBM and Oracle
• Eierything is not going to go up to Amazon or Azure, but the stuf
is going to sit in datacenters, maybe thousand of datacenters,
maybe iery small clouds, distributed globally and locally to their
partcular applicatons
• More than 80% enterprise IT organizatons will commit to hybrid
cloud architecture, as the irst step towards distributed
computng architecture
• Cloud is the core, your datacenter is the edge
8. Limits of the Internet
• The world will moie to distributed
applicatons for:
– Systems of controli Realitme seriice (cars)
– Systems of recordi Online Transacton Processing
(OLTP)/ Decision Support System (DSS)i Hyperi
scale public cloud
– Systems of engagement
• Small distributed clouds
• Local interactie seriices
9. Multple Trends Merge
• Exponentally Growing Bandwidth
• Hypericonnectiity
• The Internet of Things
• Articial Intelligence
• Deep Analytcs
• Robotcs
• Neural Networks
• Autonomous Vehicles
• 3D Printng
• Bitcoin and Blockchain
• Self Learning Systems
All of these trends and more are merging together into a powerful
machine that is going to challenge and disrupt eiery industry.
10. Cloud Computng Capabilites
• Hyperscale computng: data center facilites, hardware and system
infrastructure, applicaton infrastructure, and applicatons; non
hyperscale components can be layered on top of hyperscale
computers, but all computers use the hyperscale architecture
• Cloud service brokerage(CSB)i a company or other entty acts as a
broker to proiide cloud seriices iia four primary IT seriice roles:
aggregaton, integraton, customizaton brokerage, and goiernance
• Cloudburstngi is the use of an alternatie set of public or priiate
cloud seriices as a way to handle peak in IT system requirements;
typical use cases for cloudburstng include the expansion of IT
resources across internal data centers, external data centers, or
between internal and external data centers
11. Gartner Cloud Computng Hypercycle-
08 August 2019 G00370239 Analysts:
David Smith, Ed Anderson
Those clustered around the peak driie
innoiaton
Serverless PaaS- deployed and deieloped
Multcloud Computng- started in the research
community, trying to artculate the idea of a
seriice worldi take AWS, Ms Azure, and college
seriice and join the toghether
Container Management
Edge Computng- intelligent cloud joins the
intelligent edge by some middleware
AI PaaS and on the edge is IoT platorm (think
of this as cloudlets, computng at the edge)
Hybrid Coud Computng and Cloudburstngi
we run out of resources on the priiate system
and rush out to the public system
SaaS on top of the Cloud Computng because
it’s proiiding applicatons
IaaS is proiiding hardware as a seriice plus
serierless
Private cloud computng, Cloudburstng, and
Cloud Access Security Brokers- excitng in the
past, are struggling with implementaton
challenges
12. The Next Wave Of Cloud
Approaches
• Cloudlets (2 to 5 years)
• Service Mesh (2 to 5 years)
• Repartiaton (from public to
priiate clouds) (obsolete before
plateau)
• AI PaaS (5 to 10 years)
• Cloud Access Security Brokers
(2 to 5 years)
• Serverless PaaS (2 to 5 years)
• Distributed Cloud and Cloud
Natve- deliiering adianced
capabilites around technologies
such as API-centric SaaS and
Blockchain PaaS (5 to 10 years)
Gartner Cloud Computng
Hypercyclei 08 August 2019
G00370239 Analysts: Daiid
Smith, Ed Anderson
13. Serierless PaaS
• The PaaS cloud proiides building blocks such as databases, queue seriices,
object stores
• But an applicaton also needs glue code combining all these building blocks
into the operaton the applicaton needs to perform
• Eien on PaaS, applicaton writers start iirtual machines to run this glue code,
VMs and all the problems associated with them (installaton, scaling, etc.)
• There is a trend towards a serverless PaaS cloud, where the applicaton
deieloper does not need to rent VMs
• The cloud proiides Functon-as-a-Service (FaaS)
• FaaS implementatons (such as Amazon Lambda, Google Cloud Functons or
Microsof Azure Functons) run short functons which the applicaton author
writes in highileiel languages like Jaiascript or Jaia, in response to certain
eients
14. Serierless
• When using a cloud seriice like Azure Cosmos DB or
Amazon Kinesis, there is no need to allocate a serier
and deploy a iirtual machine
– Just conigure it from the portal and use it
• If a functon needs to run eiery tme a partcular eient
happens
– A ile in the ile seriice is modiied, a speciic
external eient is sent to a cloud eient stream
– A ile was added to the Github repository, at 3:00
AM
• FunctoniasiaiSeriice (FaaS)
– Implement in AWS Lambda (eientidriien,
serierless computng platorm proiided by
Amazon), Azure Functons, Google Functons, IBM
(Apache) OpenWhisk (open source, distributed
Serierless platorm that executes functons)
– Amazon Kinesisi ingest realitme data such as
iideo, audio, applicaton logs, website
clickstreams, and IoT telemetry data for machine
learning, analytcs, and other applicatons, and
Amazon Kinesis enables you to process and
analyze data as it arriies and responds instantly
instead of haiing to wait untl all your data is
collected before the processing can begin
Serierless
PaaS
Container
Orchestraton
IaaS
Bare Metal
15. Go Serierless!
On-prem VMs ContaSeconds-
Minutesiners
Serverless
Time to
proiision
Weeksi Months Minutes Secondsi
Minutes
Milliseconds
Utlizaton Low High Higher Highest
Charging
granularity
CapEx Hours Minutes Blocks of
milliseconds
16. Serierless Beneits
• The adopton of serierless in the enterprisei in baby phase, as the workload
today is requestidriien rather than eient driien
• Eient driien workloads will grow in importance as next generaton frontiend will
be driien by new technologies: IoT deiices, AI enabled user experience
platorms (Amazon Alexa, Apple Siri, or Google Now) and iirtual reality
applicatons
– More eientidriien, microseriice architecture
– Deployments will happen outside the puriiew of central IT
• Serierless computng ofers some beneits to deielopers:
– Supports running code without haiing to operate the infrastructure
– Can enable easier scaling
– Public cloudihosted infrastructure as a seriice (Iaas) serierless frameworks
allow for on demand consumpton, because there are no idle resources or
orphaned VMs or containers
17. Cloud Natie
• Cloud owner proiided cloud natie platorm
for
– Eient driien applicatons
– Functons are inioked by eients
– Scaling up and down instantly and automatcally
– Charges for actual usage at a millisecond
granularity
18. Cloud Eioluton Tools
• Contnuous Delivery- reliable applicatons released quickly and frequently
• New methods of sofware deliieryi APIs and browser extensions
• OSs that are iery small foot print- Reducing risk by eliminatng the number of
packages & size
• Use DevOps -which is the collaboraton between deielopers and IT operatons to
automate the process of sofware deliiery and infrastructure changes
– Insttuton needs to estmate its DeiOps capability
– Mature DeiOpp models before leaping forward are critcal
• Elastc Analytcal Databases- Google BigQuery, Snowfake Data Warehouse, and
AWS Redshif Spectrum are iery scalable, usage based, and haie minimal
maintenance requirements
• Edge computng- pushing computng away from centralized nodes and closer to
sources of data
– reduces latency and can haie security and compliance beneits
– addresses many IT challenges when running dataicentric workloads in a cloud
19. Containers And Microseriices
• Containers and microservicesi simplify sofware deielopment process and improie consistency
between testng and producton, reduce complexity of managing and updatng applicatons due to
modular approach
• Micro-services architecture approach to building an applicaton as a collecton of small
independent seriices that run on their own and communicate oier http APIs
• Containers- light weight iirtualizaton by diiiding a single serier into one or more isolated
containers, both efciently and speedy compared to standard VMs
–Eniironmental is built ini Containers proiide the ability to manage and migrate the
applicaton dependencies along with the applicaton, while abstractng away the OS
and the underlying cloud platorm
–Quick to download
–Fast to start upi microseconds to start
–Disposable
–Small singleipurpose seriices are easier to deploy
–Small footprints it well with similar distributon sizes
–Don’t patch or update the code, replace it
20. How Will Distributed Applicatons Be Built?
• From ground up from small, stateless microservices and containers
– Minimal functon seriices
– Modularity to support scalable parallelism
– Rapid applicaton modiicaton
– Easily distributed to proiide fault tolerance
– Deploy each independently
– Has its own database
– Organized around business
– State is externalized
• Make best use of the cloud’s special features
– Elastcity
• Examples:
– Netlex, Facebook, Twitter, Google Docs
– Azure CosmosDBi billions of transactons per week
– Azure Eient Hubi trillions of requests per week
– Azure Cortanai 500 million eials/sec
– Azure IoTHub, Skype, Power BI(Business Insight), CRM Dynamics (Customer Relatonship Management)
– AWS Kinesis and Serierless
21. Cloud Natie and Microiseriices Beneits
• Self managing infrastructure through automatoni goes beyond automaton
built on top of iirtualizaton platorms, focusing on orchestraton,
management, and automaton of the entre infrastructure right into the
applicaton tre
• Reliable infrastructure and applicatonsi easier to replace failed components
• Deeper insight into complex applicatons proiiding iisualizaton for health
management, monitoring and noticatons with audit logs making
applicatons easier to audit and debug
• Securityi deielopers build security to applicatons from the start
• More efcient use of resourcesi containers are lighter in weight than full
systems
• Microiseriices break large complex applicatons into smaller pieces,
deieloped, tested, and managed independently
22. Microservices
• Are iery complicated
• Microseriices forces moie to distributed computng
• There will always be latency between microseriices
• All data exchange between microseriices must be
through API layer or messagingi no accessing data store
across microseriices
• Must implement high speed messaging, REST + HTTP
probably isn’t fast enough
• May end up duplicatng data across data stores, e.g.
customer’s proile
23. Initaties to Oiercome Latencies
• Cloudletsi Compute Resources Distributed Closer To Data (Hospital)
– A new architectural element that arises from the coniergence of mobile computng and cloud computng
• Middle ter of a 3iter hierarchy: intelligent deiicei cloudleti cloud
• Can be iiewed as a “data center in a box” who’s goal is to bring the cloud closer to the deiice
• Maintains only a cache
– Four key attributes
• Small low cost maintenancei free design
• Powerful, well connected, and safei maintains only sof state
• Built for microseriices and containersi located at the edge of the Internet
• Built with standard cloud technology
• HTTP/2
– Allows data compression of HTTP headers
– HTTP/2 serier push
– Pipe lining of requests
– Fixing the headiofiline blocking problem in HHTP 1.x
– Multplexing multple requests oier a single TCP connecton
• 5G Communicaton Networks
– Ofer network slices aiailable on a priority basis to seriices
– Enable operators to create an instance of an entre network iirtually
• Mesh Networks- Distributed Computer Architecture
24. Service Mesh
•A seriice mesh is built into an app as an array of network proxies
•It doesn’t introduce new functonality to an app’s runtme eniironment
•It takes the logic goierning seriiceitoiseriice communicaton out of
indiiidual seriices and abstracts it to a layer of infrastructure
https://www.redhat.com/en/topics/microseriices/whatiisiaiseriiceimesh
25. Mesh Networks
• Distributed computng middleware that optmizes communicaton between applicaton seriices
• New and emerging class of seriice management for the interimicroseriice communicaton
complexity
• Dedicated system layer to enhance seriice to seriice communicaton across applicaton
microseriices
• Deiices can instantly share informaton
• Proiides proxy or lightweight mediaton for seriiceitoiseriice communicaton and proiides tracing
• Supports functons such as authentcaton, authorizaton, encrypton, selfihealing recoiery, seriice
instrumentaton
• Seriice meshes can help oiersee trafc through seriice discoiery, load balancing, request routng
• Seriice meshes attempt to diminish the complexity of containers and improie network functonality
• Eliminates latency and power bottleneck introduced by cellular radios
• By seriing user requests on the actual deiice, mesh networks can mitgate pressure on cellular
networks
• Since the seriice resides on the deiice, priiate info can be stored on the deiice and possibly protect
info better
26. Proximal Discoiery And Communicaton
Mash
Networks
5G
Cloudlet
1
Cloudlet
N
Public Cloud
Oracle, AWS,
Azure
Priiate
Cloud
WAN
27. The Connecton And The Driiers
• Networks that enable the cloud must support robust, lowilatency, ultraihigh capacity
connectons
• While the wireless industry has been rapidly growing into the 5G, the wire line
industry has struggled to keep pace
• A DSL or cable modem connecton may be sufcient today, but the shif to the cloud
will driie capacity needs to speeds of one Gbps and beyond with latency needs down
in the singleidigit milliseconds
• Fiber optcs (glass pipes) not only enable those 4G/5G connectons, but they also ofer
a superior cloud connecton
• Technology Distributon Driiers
• Rising & Cheaper Connectiity
• Data Sharing
28. Enterprise Transiton To Containers
• Containers are not VM
• Not all applicatons are suited for containers considering
applicaton dependenciesi System access requirements
may not be aiailable from within the container
• Monitoring containersi What does security and audit
compliance look like with containers
• Debugging issuesi Resource utlizaton Management of
containersi Users do not want multple management
tools, e.g. one for VMs, one for containers
29. Beneits Of Containers For IT
• Deielopers
– Quick create ready to run packaged applicatons
– Portable runtme eniironment
– No missing or confictng dependencies or packages
– Each app runs in isolated container
– Automated testng, integraton, packaging
– Eliminate platorm compatbility issues
– Zero cost deployment, with instant replay or reset
• Administrators
– Conigure once, run many tmes
– Make applicatons life cycle efcient, consistent, and repeatable
– Eliminate eniironmental inconsistencies between deployment, test, producton
– Support segregaton of dutes
– Improie speed and reliability of contnuous integraton and deployment
– Light weight containers address performance, costs, deployment, and probability issues
Applicaton containers are deployed standard and pushed to users based on applicaton type and
locaton
30. Immutable Infrastructure Focused On
Containers
• Containers can eliminate manually ixing problems
– Hardware failure (motherboard failed)i Autoiscaling will automatcally
launch a new container on new hardware as load dictates
– Network failure (switch failed)i Autoiscaling automatcally launches
container as load dictates
– System sofware failure (kernel)i Health checking should fail and the
container will be cooled; Autoiscaling will automatcally launch a new
container
– Applicaton Sofware failure (bad ile permission)i Fix the source (your
applicaton, your container, your Docker ile) and redeploy your entre
applicaton
31. Artfcial Intelligence Platorms
• Platorm proiiders with products in this area include IBM with Watson and
its Bluemix platorm; Amazon with Amazon Machine Learning; Google's
network of AI oferings, including TensorFlow and its Cloud Predicton API;
Microsof with its Azure Machine Learning Studio; and others focusing on
niches and iertcal industries such as Salesforce's Einstein
• Moiing toward microseriices and APIs to improie efciency and mobilityi
is critcal because intelligence is a composite process. For example,
autonomous iehicles incorporate horizontal (comparison) AI seriices for
iision and iertcal (sensing) AI processes for situatonal awareness, and
other AI seriices for operaton and naiigaton. These seriices are linked to
additonal analytcs and data seriices to create a complex AI system
• https://tdwi.org/artcles/2017/06/29/aiiasiplatormiasiaiseriice.aspx
32. Cloudlet Implementatons
• Selfidriiing cars
communicaton would be
through cell phone towers
• Informaton could not go
back to the cloud in order to
aioid a collision
• Masiie growth of edge
computng will spawn
growth opportunites for
hyperconierged and
sofware deined computng
platorms
Global
Data
Center
Cloudlet
2
Cloudlet
1
Hospital
Branch
Ofce/Outlet
5G
33. The Internet of Things (IoT)
• Appliance and automotie manufacturers can now design
products that tell you when they are about to break down
• A physical deiice is sold with a seriice
• Business models realigni connectiity is the new
horsepower, silicon is the new transmission, and electric
is the new gasoline
• New phase will iniolie smart, connected packaging
talking to our deiices
• A new era hypericonnectiity
• Eierything connects and gets smart
34. Edge Thinking Dominates
• Global Research & Deielopment (R&D) has moied from
large labs to globally dispersed idea factories
• Gamers deielop new skills for the coming world of iirtual,
remote work
• Industries partneriup: drone deliiery, robotcs, automaton,
3D printngi faster to market prototype methodology
• In inance there are more types of Fintech (Financial
Technology) startups than there are world currenciesi as
the trends & innoiatons expert, Jim Carroll, stated
https://jimcarroll.com/
35. Fault Tolerance Going To Increase
• Basic approachi replicatonibackupsi Amazon basic Simple Storage Seriice S3 has 99.9% dependability and
99.9% aiailability, but Amazon Reduce Redundancy Storage RRS (loss of 1% for 33% cheaper) didn’t work,
maybe if you regenerate data
–Losing 1% data in a recommendaton is diferent from losing 1% data in bank transactons or pharmacy
• Running programs are harder to saie and need coordinaton with messages running in the system from
other processes (creatng oierhead)
• Saie to “state” (data deining program executon) on disk in such a way that program can read data to
restarti MapReduce strategyi where processes communicate iia disk, but messaging iia disk is slow
• Parallel computng is sensitie to fault tolerance with million processes running, one fails and all processes
are thrown awayi High Performance Computng (HPC) (IBM) with 10^12 foatngipoint operatons per
second is tghtly coupled and more efort is made to aioid faults
• Applicatons fault tolerancei parallel weather simulatoni weather info propagates out from one locaton
and will reach destnaton afer an houri reason why HPC is sensitie to faults
• Independent eient process is not eien nearly so sensitie
• Public cloud proiiders (Amazon Web Seriices, Microsof Azure, Google Cloud are responsible for the
aiailability of Backup and Disaster Recoiery (DR) solutons and security of the infrastructure
• Users are in charge for their data protecton and compliance
36. Cloud Computng Is A Rapidly Growing
Technology
• As per the latest Gartner report, the cloud
tech seriices market is projected to grow
17.3% ($206 billion) in 2019, up from $175.8
billion in 2018 and by 2022, 90% of
organizatons will be using cloud seriices
• https://hub.packtpub.com/cloudicomputngitr
endsiini2019/
37. Summary
• Hybrid Cloud- will become the dominant business model in the future
• Mult-Cloudi is the next step in the cloud eioluton
• Backup and Disaster Recovery (DR) solutonsi15% of the cloud budget is allocated to this,
according to Spiceworks report
• Serverless Computng- Cloud users request a container PaaS (Platorm as a Seriice), and Cloud
supplier charges for the PaaS as required, proiiding a wide range of tools for designing
applicatons, and working with data
• Data Containers used to transporti can only be used with seriers haiing compatble operatng
system kernels
• Artfcial Intelligence Platorms- to process Big Data and work together making simple tasks
automated & preientng duplicaton
• Edge computng- is essental to run realitme seriices as it streamlines the fow of trafc from IoT
deiices and proiides realitme data analytcs and analysis
• Service mesh- dedicated system layer to enhance seriice to seriice communicaton across
microseriices applicatons
• Cloud Security- introducton of the GDPR (General Data Protecton Regulaton) ensures that data
practces are both safe and compliant
38. Reference List
• Digital Science Center Courses
E534 2019 Cloud Computng Unit in 18 Videos
Published on Sep 2, 2019
• https://computer.howstufworks.com/
• https://www.tutorialspoint.com/
• https://jimcarroll.com/
• Linux Administraton in Distributed Cloud
Computng Eniironments by Robert Shimp,
Oracle https://www.linux.com