peer peer and applicat ion level net working cs 218 fall
play

Peer-peer and Applicat ion-level Net working CS 218 Fall 2003 Mult - PowerPoint PPT Presentation

Peer-peer and Applicat ion-level Net working CS 218 Fall 2003 Mult icast Overlays P2P applicat ions Napst er, Gnut ella, Robust Overlay Net works Dist ribut ed Hash Tables (DHT) Chord CAN Much of t his mat erial comes f rom UMASS class


  1. Peer-peer and Applicat ion-level Net working CS 218 Fall 2003 Mult icast Overlays P2P applicat ions Napst er, Gnut ella, Robust Overlay Net works Dist ribut ed Hash Tables (DHT) Chord CAN Much of t his mat erial comes f rom UMASS class slides

  2. Resilient Overlay Net works Overlay net work: � applicat ions, running at various sit es � creat e “logical” links (e.g., TCP or UDP connect ions) pairwise bet ween each ot her � each logical link: mult iple physical links, rout ing def ined by nat ive I nt ernet rout ing � let ’s look at an example:

  3. I nt ernet Rout ing � BGP def ines rout es bet ween st ub net works I nt ernet 2 Berkeley.net UMass.net C&W Mediaone UCLA Noho.net

  4. I nt ernet Rout ing � BGP def ines rout es bet ween st ub net works I nt ernet 2 Berkeley.net UMass.net C&W Mediaone Noho-t o-UMass UCLA Noho.net

  5. I nt ernet Rout ing � BGP def ines rout es bet ween st ub net works I nt ernet 2 Berkeley.net UMass.net C&W Mediaone Noho-t o-Berkeley UCLA Noho.net

  6. I nt ernet Rout ing I nt ernet 2 Berkeley.net UMass.net Congest ion or C&W f ailure: Noho t o Berkely BGP- det ermined rout e may not change (or will Mediaone change slowly) Noho-t o-Berkeley UCLA Noho.net

  7. I nt ernet Rout ing Noho t o UMass t o Berkeley � rout e not visible or available via BGP! � MediaOne can’t rout e t o I nt ernet 2 Berkeley via I nt ernet 2 Berkeley.net UMass.net Congest ion or C&W f ailure: Noho t o Berkely BGP- det ermined rout e may not change (or will Mediaone change slowly) Noho-t o-Berkeley UCLA Noho.net

  8. RON: Resilient Overlay Net works Premise: by building applicat ion overlay net work, can increase perf ormance, reliabilit y of rout ing applicat ion-layer Two-hop (applicat ion-level) rout er noho-t o-Berkeley rout e

  9. RON Experiment s � Measure loss, lat ency, and t hroughput wit h and wit hout RON � 13 host s in t he US and Europe � 3 days of measurement s f rom dat a collect ed in March 2001 � 30-minut e average loss rat es � A 30 minut e out age is very serious! � Not e: Experiment s done wit h “No-I nt ernet 2-f or- commercial-use” policy

  10. An or der-of -magnit ude f ewer f ailures 30-minute average loss rates RON RON Loss Rat e Bet t er No Change Worse 10% 479 57 47 20% 127 4 15 30% 32 0 0 50% 20 0 0 80% 14 0 0 100% 10 0 0 6,825 “path hours” represented here 6,825 “path hours” represented here 12 “path hours” of essentially complete outage 12 “path hours” of essentially complete outage 76 “path hours” of TCP outage 76 “path hours” of TCP outage RON routed around all of these! RON routed around all of these! One indirection hop provides almost all the benefit! One indirection hop provides almost all the benefit!

  11. RON Research I ssues How t o design overlay net works? • Measur ement and self -conf igur at ion • Under st anding per f or mance of under lying net . • Fast f ail-over. • Sophist icat ed met r ics. • applicat ion-sensit ive (e.g., delay ver sus t hr oughput ) pat h • select ion. Ef f ect of RON on underlying net work • I f ever yone does RON, ar e we bet t er of f ? •

Recommend


More recommend