Understanding Network Layers & Protocols: Computer Communication - Prof. John Musacchio, Assignments of Accounting

Download Understanding Network Layers & Protocols: Computer Communication - Prof. John Musacchio and more Assignments Accounting in PDF only on Docsity! ISM 50 - Business Information Systems Lecture 17 Guest Instructor: Brad Smith Instructor: John Musacchio UC Santa Cruz November 27, 2007 Class announcements For Thursday 11/29 Database Assignment Due Folio Article 3 Due Read: Messerschmitt Ch 9 (273-289) Upcoming Student Presentations Jason Barbaria (Business paper) Gloria Yeung (Business paper) What are some examples of communications networks? Public Telephone Network Internet LANs (Local Area Networks) What does a network do? 1) Transport data from one host to another. Network Architecture Network architectures are layered Each layer uses the services of the layers below To offer more advanced services to layer above Allows layers to be designed independently We will talk about 3 layers next… Link Network Physical Link Layer Make a Frame link out of a bit link Instead of endless sequence of 1s and 0s, we want distinct “packages” of data that are separate from each other S t t d 2 F ith d tay we wan o sen rames w a a 01010101010111010 and 101010101011010 Concatenate them and send them as a sequence? How can the receiver tell where the new frame begins? Solution: insert a special sequence at the start of frame: for example: 01111110 Link Layer (cont’d) Also does error detection/correction Insert extra information the helps the receiver to determine if the data has been corrupted. Example: parity bit Sender adds a 1 or zero to end of data so number of ones is always odd 10011 or 10000 If receiver counts an even number of ones, than it knows the data was corrupted. More Link Layer.. -- Ethernet Want to allow multiple hosts to share a link Host A Host B Host C Host D How do they avoid talking at the same time? Don’t transmit if you hear another host transmitting If th is c llisi n st p it nd m m unt ere a o o , o wa a ra o a o of time, and try again This is a Medium Access Control (MAC) Protocol Ethernet Switch Host A Host B Host C Host D Switch If switch knows where the destination is Switch forwards an incoming frame to destination only. Otherwise, it broadcasts it to everyone. Thus, parallel conversations possible. Network Layer Host B Link 2 128.114.60.202 Host CLink 1 Link 3 Host A A wants to send some data to C Host D Suppose A knows C’s address A sends a packet towards C A marks his packet with C’s Address (an IP Address) Post Office Analogy Bob in New York Truck to Alice Alice Smith 1156 High St Santa Cruz 95064 High St Truck to 41st Ave Plane to London Truck to Santa RosaNY Post Office Plane to SFO SFO Post Office Truck to Santa Cruz Santa Cruz Post Office Look at beginning of zip code. Make forwarding decision Look at address Make forwarding decision Routing in the Internet Host A Host C Host B Many feasible paths from source to destination . Routing Routing Updating the routing table Objective: each packet gets closer to destination Packet forwarding Transmitting each packet on the appropriate output link Based on routing table Routing Algorithms Wild Card Routers talk to each other to build their routing tables “I am accepting 114.211.1.1 Traffic to 114.211.1.X” Host A Host C HUB 114.211.1.2 Host B 114.211.1.3 114.211.1.4 Routing Concerns Long routes Circular routes Hijacking routes Route flapping IP Addresses vs Mac Addresses Hierarchical The beginning bits tell you which network the host is Not Hierarchical Beginning bits tell nothing useful on Ex: UCSC addresses start with 128.114.X.X The last bits tell you which host of the network Changeable Not ChangeableChanges with location of Host 4 bytes 6 bytes 2 1 llOnly 4.2 billion 8 Tri ion Link and Network Layer Interaction Ethernet Frame Payload Router 128.114.60.202 MAC Header IP Header IP Payload Host A Host C Ethernet Router Router Ethernet Hub Router Hub MAC address 00-A4-B7-34-57-23 Strip MAC header off frame. F d IP k t b d Host B orwar pac e ase on Routing table. Issues In Networking Statistical Multiplexing Link shared in such a way that connections are not assigned fixed fraction of Link. A and B unlikely to offer peak rate at the same ti me. max( A+ B) < max(A) + max(B) A Source A D ti ti A A+B es na on B Source B Destination B Statistical Multiplexing Because resources aren’t reserved. It’s ibl ff d l d hi hposs e o ere oa too g . Packets are put into a queue. If offered load remains too high, queue will fill up and overflow. Source A Destination A Destination BSource B Transport Protocols The Internet is unreliable It will make a “best effort” to get your packet to its destination Packets can be lost because of Congestion Link errors Application Routing problems Session Presentation Link Network Transport Physical UDP Audio d Audio d dStream ofco er eco er packets For some applications packet retransmissions are not worthwhile Why? For those applications, we use UDP UDP i t t t l th ts a ranspor pro oco a Does not do retransmissions Does not do congestion control Congestion Control When networks are congested, certain sessions (Source-destination pairs) should reduce offered rates. Today all TCP sessions slow down when they detect packet losses. UDP sessions do not slow down. What are some alternative strategies? Have those whose applications aren’t as sensitive slow down more? How would we know which are less sensitive Pricing within the Internet Backbone NSP $ Flat Rate or simple usage based Peering Relationship ISP (Cruzio) $ Flat Rate Backbone NSP 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. O llvera … Internet billing today is much more course grained than telephone billing. Hierarchy in Addresses vs. Names Addresses hierarchical in topology Maximize “wild cards” and distribute address administration Names hierarchical in administration Single administered organizations often distributed topologically (e.g. ibm.com) Transport Protocols The Internet is unreliable It will make a “best effort” to get your packet to its destination Packets can be lost because of Congestion Link errors Application Routing problems Session Presentation Link Network Transport Physical OSI Layers Presentation Application Internet Explorer, Outlook Email, Real Player, … Transport Session TCP, UDP Link Network Ethernet, Wi-Fi, SONNET, … Internet Protocol (IP), … Physical Modulation Schemes: QAM, OFDM, etc… ISP Topology Telephone Company Local Office Local Loop Telephone SwitchLocal Loop To Telephone Network Local Loop DSL Modem D S Leased ISP Point of Presence DSL Modem DSL Modem SLA M Line to NAP | Network Service Provider etwo Ss Point etwork AG Point Large E-Business Load Balancer I i HTTP b ncom ng RequestsPresentation Logic (Assembling We ServersWeb page) Logic Interconnected Application ServersFlow of Interaction with Gigabit Ethernet or other technology Databases Customers Merchandise Orders Akamai Case Internet Bottlenecks First Mile (Server Capacity) – 70% of website performance problems according to one study B kb Pl tif l b t h t ithi ac one – en u , u some s or age w n metropolitan areas Peering – Exchange of traffic between NSPs Last Mile to home 56 K modems are slow Shared LAN limitations Solutions Expand Bandwidth Being done Mirroring web cites Put exact copy of same web page to multiple servers Tricky to duplicate content C hiac ng Problem: Stale Content Problem: Hard to count “click throughs” Content Distribution Networks… 2000 Financials $196 Million Loss (Before special charges) $90 million revenue %20 gross margin, after deducting server depreciation payments to network partners D ata center space But, most expenses of shouldn’t grow at same rate as number of customers, so margin should improve $201.5 million SG&A (selling general and administrative) (l l l f )arge y sa es orce cost Again, this might not grow at same rate as the number of customers. $40 illi R&Dm on Competition Hosting firms (substitute) Exodus Other CDNs Sandpiper Adero Mirror Image, , Content Alliances Ak i’ tit b d d t th t h ama s compe ors an e oge er o s are networks 2001 Market Changes Bad Dot-coms bust Customers leave “churn rate goes to 22% per quarter” G doo Hosting firms go bust (exodus) Some CDN competitors go bust. Competing CDN alliances mired in problems Technology Dynamic CDN technology: ESI (edge sides includes) Develop as open standard why? Akamai not big and credible enough to force d f t t d d k ta e- ac o s an ar on mar e Marketing Difference in selling old vs new products: Old product Geared toward speeding up websites Revenues of their clients depended on speed Easier to get sale New Product Simplify company IT function Cost vs. revenue center Harder sell. More data driven… Consequently new product needs more professional sales force Channels? Distribution Partners (IBM) credibility Di S l F rect a es orce too Recent Performance Consolidated Statements of Operations Data: Revenues Total cost and operating expenses Net income (loss) Net income (loss) attributable to common stockholders 2004 2003 2002 2001 2000 (In thousands, except share data) $ 210,015 161,259 $ 144.976 $ 163.214 $ 89.766 161,048 172,370 327.580 2,577,108 989,359 34,364 (29.281 (204.437) (2.435512) (885.785) 34,364 (29,281 (204.437) (2.435.512) (885,785 )