1 TIM 50 - Business Information Systems Lecture 18 Instructor: Ram Akella UC Santa Cruz November 24, 2015 Announcements Teaching Evaluations Open November through December Final Business Papers, 12/3 Final Exam, Monday 12/8 Pricing within the Internet Customer pays an ISP Often Flat Rate per month ISP pays a backbone AS Often just flat rate, dependent on access link speed. Sometimes based on total usage Backbone NSPs peer with each other Often for free if they exchange comparable amounts of traffic. Overall… Internet billing today is much more course grained than telephone billing. ISP (Cruzio) Backbone NSP $ Flat Rate $ Flat Rate or simple usage based Backbone NSP Peering Relationship Networks Domain Names IP addresses are inconvenient for people 32 bits hard to remember 32 bits very hard to remember Domain names e.g. ucsc.edu Easier to remember than IP addresses However, we need some way of mapping domain names to IP addresses. Domain Name System (DNS) Berkeley Name Server EEC S Name Server Root Name Server U C SC Name Server SoE Name Server
Transcript
1
TIM 50 - Business Information Systems
Lecture 18
Instructor: Ram Akella
UC Santa Cruz
November 24, 2015
Announcements
¨ Teaching Evaluationsq Open November through December
¨ Final Business Papers, 12/3¨ Final Exam, Monday 12/8
Pricing within the Internet
n Customer pays an ISPq Often Flat Rate per month
n ISP pays a backbone ASq Often just flat rate, dependent on access link speed.q Sometimes based on total usage
n Backbone NSPs peer with each otherq Often for free if they exchange comparable amounts of
traffic.n Overall…
q Internet billing today is much more course grained than telephone billing.
ISP (Cruzio)
Backbone NSP
$Flat Rate
$
Flat Rate orsimple usage based
Backbone NSP
Peering Relationship Networks
Domain Names
IP addresses are inconvenient for peopleq 32 bits hard to rememberq 32 bits very hard to remember
Domain namesq e.g. ucsc.edu
q Easier to remember than IP addresses
q However, we need some way of mapping domain names to IP addresses.
Domain Name System (DNS)
Berkeley
Name Server
EECS
Name Server
Root
Name Server
UCSC
Name Server
SoE
Name Server
2
Hierarchy in Addresses vs. Names
Addresses hierarchical in topologyq Maximize “wild cards” and distribute address
administration
Names hierarchical in administrationq Single administered organizations often
distributed topologically (e.g. ibm.com)
OSI Layers
Physical
Link
Network
Transport
Session
Presentation
Application
Modulation Schemes: QAM, OFDM, etc…
Ethernet, Wi-Fi, SONNET, …
Internet Protocol (IP), …
TCP, UDP
Internet Explorer, Outlook Email,Real Player, …
Some Typical Topologies
Home Network
DSL Modem TelephoneLine
RouterEthernet Switch
(to localOffice)
Small/Medium Business
Routerwith Firewall
T1 LineT1 Modem
EthernetSwitch
Web Site Server
To LocalOffice
ISP Topology
Telephone CompanyLocal Office
Local Loop
TelephoneSwitchLocal Loop
Local Loop
DSL Modem
DSL Modem
DSL Modem DSLA
M
LeasedLine toNAP
ToTelephoneNetwork
ISP Point of Presence
Network Service Provider
NetworkAccessPoint
NetworkAccessPoint
3
Large E-Business
Customers Merchandise Orders
Databases
Application Servers
Web Servers
Load Balancer
Incoming HTTPRequestsPresentation
Logic(AssemblingWeb page)
LogicFlow ofInteraction
Interconnectedwith Gigabit Ethernet or
other technology
Web Caching
n Speed up web page loading by storing previously seen components locally
• Links up to 8 devices in 10-m area using low-power, radio-based communication
• Useful for personal networking (PANs)
• Wi-Fi (802.11)
• Set of standard: 802.11a, 802.11b, 802.11g, 802.11n
• Used for wireless LAN and wireless Internet access
• Use access points: device with radio receiver/transmitter for connecting wireless devices to a wired LAN
The Wireless Revolution
Essentials of Management Information SystemsChapter 6 Telecommunications, the Internet, and
Wireless Technology
Essentials of Management Information SystemsChapter 6 Telecommunications, the Internet, and
Wireless Technology
Cloud computing
Cloud Computing
Ø Cloud Computing: refers to bothq applications delivered as services over internet
n aka Software as a Service (SaaS)q hardware / software in data centers providing
those services -- a cloud
Cloud Computing
Ø 2 flavors:q Public Cloud, available to public
n provides utility computingq Private Cloud
n internal to company
AdvantagesØ SaaS
q Control of Versioningq Users access anywhereq Ease of data sharingq Pay as you go
Ø Additional +’s of Cloud Computingq Deploy new services without building and
provisioning data centers n E.g. Zynga Farmville
q Scale up/down resources as needed
5
Context
Ø 2000’s q Large investments by web giants (e.g. Google,
Amazon) in infrastructuren Giant data centersn Software Infrastructure for such data centers
q MapReduce -- allows computations to be distributed to multiple machines “map”, and then results collected for further processing “reduce.”
q Hadoop – open source version of above
q Above pieces prerequisites to become a cloud provider
Reasons to be a cloud providerØ A big player enjoys economy of scale advantage
Ø Leverage existing investments for new revenue stream (e.g. Amazon)
Ø Defend existing markets (e.g. MS enterprise apps with Azure)
Ø Become a platform (facebook)
Ø Leverage relationships (IBM)
Why is the Cloud becoming big only now?
q shift from large commitment models to contactless short term model
q Mobile interactive applications that need huge data sets
q Parallel batch processing – software like Hadoop makes it easier to do this
q Analytics – less growth in plain transaction processing, more growth in analyzing trends / predictions from large data sets
Ø
Types of Utility ComputingØ Amazon EC2 – to programmer, each instance looks like physical
hardwareq Can control whole layer stack q Other managed services provided (e.g. SimpleDB)
Ø Application Domain specific platformsq Google AppEngine (software dev. platform for web
applications)q Force.com (Salesforce.com) – platform for business apps that
use salesforce.com DBØ MS Azure –
q Provides developers a general purpose software framework .NET
q Compiled to a managed environment (rather than to specific hardware)
Economics
Ø “pay as you go” modelØ add and remove resources at a fine time
scaleq proprietary data centers have to provision
for peakq hard to predict demand of new servicesq poor service quality can alienate customerq large data centers have significant eco. of
scale advantage
Challenges
Ø Availabilityq Can actually be better than in-house data
centersq More robust to DDOS (Distributed denial of
service) attacks by being so large Ø Lock-in
q Data lock-in – online storage services have gone bust
q application programming interfaces not common
6
Challenges
Ø Confidentiality and Auditsq Sarbanes Oxley, HIPPAq Can use encryptionq Audibility can be added as layer
Ø Data Transfer bottlenecksq Slow transfer can offset faster processingq Ship hard drivesq Upload once, use multiple times
Discussion• Imagine you are the CIO of a supermarket chain with
a loyalty card. Your loyalty card collects sales data from your customers. You want to run complex algorithms to do targeted marketing to your customers.q - Would you do this in a public cloud, or internal
data center? • You are launching a new web service to provide
restaurant reviews in a locality. You do not know how much traffic your new service will generate.
• Would you do this in a public cloud, or internal data center?
Pop Quiz
• What are the seven layers of the Internet stack?
• What is congestion control?q Is there a protocol for it? Name?
• What is the difference between MAC and IP addresses in a sentence or two?
