Lecture 01 - Intro

Cloud
Compu)ng
Lecture 1

Intro – By Shmulik Goldstein

Dan Amiga
[email protected]
Dan Amiga – IDC Cloud Compu)ng 2015

Logistics
• All logistics covered by Dan next week

Shmulik Goldsein - About
• Platform Startegy Advisor – Microsoft

• [email protected]

Course Agenda
• Intro
• From 0/1 to Data Centers
• Architecture + AWS
• Architecture + Azure
• Storage
• Cache
• RDBMS
• NoSQL
• Big Data
• Security

Todays Agenda - Intro
• Problems
• History
• What is cloud computing

Problems
• Time
• Cost
• Process
• Performance
• Scale/Parallel
• Storage Size
• Availability
• Connectivity
• Security
• How-to-do-it-right

Users & Performance
• Google style applications
– Massive amount of data
– Massive amount of short time users
– Close to nothing latency
– As available as the internet
• Facebook
– Massive amount of data
– Data relations complexity
– Massive amount of long time users
– Be agile “code wins arguments”, time line, etc.

Time
• Even if you are a small player time is

important
– TTM, Latency, Time to scale, Availability, etc
Development Deployment
Cost 30% 70%
Time 10% 90%

Storage problem

Teaser 1
• What is the fastest way to search an array?

Teaser 2
• Can you break a crypto hash function?

– Assuming no vulnerabilities

Teaser 3
• I Have data from multiple buildings around

the world and would like to ask questions
about that data.

Evolu)on

1950s-1960s… Mainframe
• Big, extremely reliable, secure transaction
oriented business applications.
• Not really measured by MIPS or FLOPS
• CPU + I/O
• Exist today!
– No more punch cards; now have (web)
user interface
– NASA Just powered down it’s last
mainframe (2012, IBM Z9)
• Highly Centralized - Filled the room
Mainframe for developers
• No developer productivity + Time Sharing

• Assembly/Cobol/Fortran

Supercomputers vs. Mainframes
• Supercomputer
• At the frontline of current processing capacity (speed)
• Mostly HPC usage for scientific and engineering
problems (data/number crunching)
• Measured in FLOPS(peta flops) or TEPS (amazon)
• *as of 2011 only 9 computers in the world were
measured in gigaTEPS (but you can do better)
– Mainframes
• Measured in MIPS (instructions usually integers)
• Built for (business) transaction reliability (goods, money,
services, airlines)
• Transaction includes DB I/O

Supercomputers vs Distributed computing
• Supercomputer
• Exotic Architectures
• 100,000s of processors (how do they communicate?)
– Distributed Architectures of Computing
• Usage of computer clusters
– Tightly coupled systems
– Single system image
– Centralized Job management & scheduling system
• Usage of computer grids
– Loosely coupled (Decentralization)
– Diversity and Dynamism
– Distributed Job Management & scheduling

1970s… Minis
– Less then 25K$ vs 1M$ for mainframes
– Made for smaller organizations, departments, relatively
affordable
– UNIX was originally a minicomputer OS, while Windows
NT—the foundation for all current versions of Microsoft
Windows—borrowed design ideas liberally from VMS
and UNIX
- Pascal / C

1980s – middle of 2000… PCs

PC
• 1K USD (?)
• Made for a single person
• Microprocessor
• Programmer Productivity
– Smalltalk, C++
• End of the 20th century we thought this is where the
computation will be
– See Microsoft software and UI directions
– See SETI
• Famous Steve vs Bill
– Unique/Generic HW/SW
– “People who are really serious about software should make
their own hardware” (Alan Kray)
• Finally Socket based programming
1990s… Networking + Servers
• Ethernet, Patch Panels
• Lower cost of LANs
• Software
– Beginning of Java/Corba/RMI
– Microsoft Component Oriented Programming (COM),
DCOM
• Architectures
– Client/Server
– 3-Tier, RPC Based
• Administration explosion

2000s - Internet
• Infrastructure stability and cost go down

• Popularity and reach go up
• Reliable protocol stack
• Software
– ROR/PHP/Python – (Scripting + Interrupters)
– .NET, Java (Component oriented)
– Service Oriented Architectures
• Lack of customization and multi tenancy

Typical web/large scale enterprise app

Tiers vs Layers

Today

Cloud Principals

Pay only for what you use

Ability to scale up and scale down

Cloud Computing
• Store Data
• Run Applications
– Combined with:
• Utility model
• Elastic Nature

What Is Cloud Computing?
Cloud Compu)ng: App and Infrastructure over Internet
Compute as a Service: Applica)ons over the Internet
U)lity Compu)ng: “Pay-‐as-‐You-‐Go” Datacenter Hardware and SoRware
Three New Aspects to Cloud Compu)ng
The Illusion of Infinite Compu)ng Resources Available on Demand
The Elimina)on of an Upfront Commitment by Cloud Users
The Ability to Pay for Use of Compu)ng Resources

on a Short-‐Term Basis as Needed
Why Now?
• It’s possible
• It’s mandatory
• “Web Space Race”
– Build extremely large datacenters (10,000/0’s)
– Driven by growth (more users, more data)
• Operations & Infrastructure expertise
• Broadband got better

Platform Evolution
• Hosted so1ware pla6orm

Client Server Mobile Cloud
• Shared infrastructure
• Virtualized and dynamic
• Increasingly higher level services
• Pay as you go pricing model

Platform Continuum
On-‐Premises Hosted Cloud

Servers Servers Plaòrm
• Bring your own machines, • Ren)ng machines, • Shared, mul)-‐tenant

connec)vity, soRware, connec)vity, soRware infrastructure
etc. • Less control • Virtualized & dynamic
• Complete control • Fewer responsibili)es • Scalable & available
• Complete responsibility • Lower capital costs • Abstracted from the
• Sta)c capabili)es • More flexible infrastructure
• Upfront capital costs for • Pay for fixed capacity, • Higher-‐level services
the infrastructure even if idle • Pay as you go
On-‐Premise

% of
Capital Equipment Budget
spent on IT in 2000?
45%
Dan Amiga – IDC Cloud Compu)ng 2Commerce
015 Department Statistics
% of U)lized
Server Capacity
on Average?
6%
Dan Amiga – IDC Cloud Compu)ng 2015 Economist Survey on IT, 2008
Elasticity – Provisioning for Peak
Real World Server Utilization is 5% to 20%
• Many Services Peak Exceeds Average by a Factor of 2 to 10
• Most Provision for Peak
• Painful to Under-Provision (Lost Customers)
Provisioning for Peak

Without Elas)city,
We Waste Resources
(Shaded Areas)
During Non-‐Peak Times
Elasticity: Risks of Under-Provisioning
Under-‐Provisioning #1
Poten)al Revenue
(Shaded Area) Is
Sacrificed
Under-‐Provisioning #2
Some Users Respond to
Under-‐Provisioning by
Permanently Deser)ng
the Site...
Bad for Revenue!
Fallback: Hosted

WebHos)ng Contract Boilerplate
Agreement Provisions Included – Scope
– Key Tasks and Milestones
– Scope of Services
– Project Deliverables
– Price and Payment
– Time and Cost Es)mates
– Term and Termina)on
– Price and Payment
– Customer Service
– Invoices
– User's Warran)es and Obliga)ons
– Payment
– Ownership of Intellectual Property
– Project Organiza)on and Personnel Requirements
– Warranty and Disclaimer
– Suppor)ng Documenta)on
– Limita)on of Liability
– Expenses
– Indemnifica)on of Host
– Confiden)al Informa)on
– Rela)on of Par)es
– Employee Solicita)on/Hiring
Exhibit B: Service Level Agreement
– Down)me
– Non-‐assignment
– Technical Support
– Arbitra)on
– Aiorneys' Fees
– Severability
– Force Majeure
Exhibit C: Web Hos)ng Acceptable Use
– No Waiver Policy
– En)re Agreement – Acceptable Use
– Repor)ng of Viola)ons of This Acceptable Use Policy
– Revisions to This Acceptable Use Policy
Exhibit A: Statement of Work
– Preamble
– Project Background
Solutions
• Time
• Cost
• Scale
• Storage Size
• Availability
• Connectivity
• Security
• How-to-do-it-right

Cloud Workload Paierns

Workload Paierns Op)mal For Cloud
“On and Off “ “Growing Fast“
Compute
Compute

Inac)vity
Period
Average Usage
Average Usage
Time Time
On & off workloads (e.g. batch job) Successful services needs to grow/scale
Over provisioned capacity is wasted Keeping up w/ growth is big IT challenge
Time to market can be cumbersome Complex lead )me for deployment
“Unpredictable Burs)ng“ “Predictable Burs)ng“

Compute
Compute
Average Usage Average Usage
Time Time
Unexpected/unplanned peak in demand Services with micro seasonality trends

Sudden spike impacts performance Peaks due to periodic increased demand
Can’t over provision for extreme cases IT complexity and wasted capacity
Application Models
Web Hosting High Performance Computing
§ Massive scale infrastructure § Parallel & distributed processing
§ Burst & overflow capacity § Massive modeling & simulation
§ Temporary, ad-hoc sites § Advanced analytics
Application Hosting Information Sharing
§ Hybrid applications § Reference data
§ Composite applications § Common data repositories
§ Automated agents / jobs § Knowledge discovery & mgmt
Media Hosting & Processing Collaborative Processes
§ CGI rendering § Multi-enterprise integration
§ Content transcoding § B2B & e-commerce
§ Media streaming § Supply chain management
Distributed Storage § Health & life sciences
§ External backup and storage § Domain-specific services
Cloud Services
“IaaS”
Infrastructure-as-a-Service
“PaaS”
Platform-as-a-Service
“SaaS”
Software-as-a-Service
host build consume

Cloud Services
Packaged Infrastructure Platform Software

(as a Service) (as a Service) (as a Service)
Software
You manage
Applications Applications Applications Applications
You manage
Data Data Data Data
Runtime Runtime Runtime Runtime
Managed by vendor
Middleware Middleware Middleware Middleware
You manage
Managed by vendor
O/S O/S O/S O/S
Managed by vendor
Virtualization Virtualization Virtualization Virtualization
Servers Servers Servers Servers
Storage Storage Storage Storage
Networking Networking Networking Networking

State of Cloud Computing
• Perceptions > Types
– “The end of software” • Public
– On-demand infrastructure • Private
– Cheaper and better • Internal
• Reality • External
– Hybrid world; not “all-or-nothing” • Hybrid
– Leverage existing IT skills and
investments > Categories
– Seamless user experiences • SaaS
– Evolutionary; not revolutionary • PaaS
• Drivers • IaaS
– Ease-of-use, convenience
– Product effectiveness
– Simplify IT, reduce costs

Public cloud concerns

Economy of scale
• Bandwidth (not so much) OpEx

• Network (not so much) CapEx
• Servers (Yes, but)
– Cirrascale, SGI, Rackable

Amazon Prices March 2012
Ques)ons:
1. Is this cheap or expensive? Why?
2. Will windows be cheaper? Why?
3. Why are the memory numbers are floats?
4. How can I make this cheaper?


Spot Instances – bid!

And, what about these new scenarios?
Performance – FLOPS & Big Data Problems
• Climate Change Prediction Program

– Weather Prediction for a week, takes 24 hours
• 56Gflops
– Climate Prediction for 50 years, takes 30 days
• 4.8 Tflops
• How much on my laptop?

Internet-Scale Application
• 2009 stats:
– +200B pageviews/month – +20B photos, +2B/month growth
– >3.9T feed actions/day – 600,000 photos served / sec
– +300M active users – 25TB log data / day processed
– >1B chat mesgs/day thru Scribe
– 100M search queries/day – 120M queries /sec on memcache
– >6B minutes spent/day (ranked
#2 on Internet)
• Scaling the “relational” data:

– Keeps data normalized,
randomly distributed, accessed
at high volumes
– Uses “shared nothing”
architecture

Internet-Scale Application
• 2007 stats:
– +20 petabytes of data processed / day by +100K MapReduce jobs
– 1 petabyte sort took ~6 hours on ~4K servers replicated onto ~48K disks
– +200 GFS clusters, each at 1-5K nodes, handling +5 petabytes of storage
• ~40 GB/sec aggregate read/

write throughput across the
cluster
• +500 servers for each
search query < 500ms
• Scaling the process:
– MapReduce: parallel
processing framework
– BigTable: structured hash
database
– Google File System:
massively scalable distributed
storage

Lecture 01 - Intro

Uploaded by

Copyright:

Available Formats

Lecture 01 - Intro

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture 01 - Intro

Uploaded by

Copyright:

Available Formats

Cloud

Dan Amiga – IDC Cloud Compu)ng 2015

• All logistics covered by Dan next week

Dan Amiga – IDC Cloud Compu)ng 2015

• Platform Startegy Advisor – Microsoft

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

• Even if you are a small player time is

Cost 30% 70%

Time 10% 90%

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

• What is the fastest way to search an array?

Dan Amiga – IDC Cloud Compu)ng 2015

• Can you break a crypto hash function?

Dan Amiga – IDC Cloud Compu)ng 2015

• I Have data from multiple buildings around

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

• No developer productivity + Time Sharing

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

• Infrastructure stability and cost go down

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Compute as a Service: Applica)ons over the Internet

U)lity Compu)ng: “Pay-­‐as-­‐You-­‐Go” Datacenter Hardware and SoRware

Three New Aspects to Cloud Compu)ng

The Illusion of Inﬁnite Compu)ng Resources Available on Demand

The Elimina)on of an Upfront Commitment by Cloud Users

The Ability to Pay for Use of Compu)ng Resources

Dan Amiga – IDC Cloud Compu)ng 2015

• Hosted so1ware pla6orm

Dan Amiga – IDC Cloud Compu)ng 2015

On-­‐Premises Hosted Cloud

• Bring your own machines, • Ren)ng machines, • Shared, mul)-­‐tenant

Dan Amiga – IDC Cloud Compu)ng 2015

Provisioning for Peak

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

Dan Amiga – IDC Cloud Compu)ng 2015

“Unpredictable Burs)ng“ “Predictable Burs)ng“

Unexpected/unplanned peak in demand Services with micro seasonality trends

host build consume

Dan Amiga – IDC Cloud Compu)ng 2015

Packaged Infrastructure Platform Software

Runtime Runtime Runtime Runtime

Virtualization Virtualization Virtualization Virtualization

Servers Servers Servers Servers

U)lity Compu)ng: “Pay-‐as-‐You-‐Go” Datacenter Hardware and SoRware

On-‐Premises Hosted Cloud

• Bring your own machines, • Ren)ng machines, • Shared, mul)-‐tenant