Cross-Domain Cooperation for Small Clients Amy S. Hughes Joe Touch USC/ISI
Overview ■ Problem: DNS overhead ■ Solution: Web-DNS Cooperation ■ Experiment: Squid log analysis ■ Conclusions ■ Future Work 2
DNS Overhead in Web Transactions ■ DNS request is a large part of web transaction ■ DNS request dominates as: – Bandwidth increases – Persistent connections reduce overhead – Latency increases A = Connection Goodput C C B = Connection Establishment < C = DNS Requests A + B + C A + B + C A + B + C = Total Connection Time ■ DNS is multiple RTTs 3
Web Connection Components DNS Client Server Request Server 1 Response SYN 2 SYN/ACK 4 ACK 5 GET 3 Time RESPONSE FIN 1: DNS FIN/ACK 2: Connect 3: First-Response ACK 4: Start-total 5: End-total 4
DNS overhead with low latency 1 1 DNS DNS Connect Connect 0.9 0.9 First-response First-response End-total End-total 0.8 0.8 0.7 0.7 Fractio n of Requests Fraction of R eq uests 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 1 2 3 5 1 2 3 5 0 0 0.001 0.01 0.1 1 10 100 0.001 0.01 0.1 1 10 100 Seconds (log) Seconds (log) Server Server Client Client DNS DNS LAN/remote requests LAN/local requests 5
DNS overhead with high latency 1 1 DNS DNS Connect Connect 0.9 0.9 First-response First-response End-total End-total 0.8 0.8 0.7 0.7 Fractio n of Requests Fractio n of Requests 0.6 0.6 0.5 0.5 0.4 0.4 0.3 0.3 0.2 0.2 0.1 0.1 1 2 3 5 2 3 1 5 0 0 0.001 0.01 0.1 1 10 100 0.001 0.01 0.1 1 10 100 Seconds (log) Seconds (log) Client Server Client Server DNS DNS ISDN/remote requests ISDN/local requests 6
DNS Reuse and Costs ■ Squid logs: 10-15% DNS misses ■ 2MB cache upper bound 7
Cache Anticipation ■ Web Cache – Request stream related to item content – Anticipation possible ■ DNS Cache – No item relation to request stream – No anticipation opportunity 8
Web-DNS Cooperation ■ Opportunity – Web request requires DNS information – Cooperation possible ■ Solution – DNS cache on local client – Web lookahead to anticipate DNS requests 9
DNS Anticipated Cache Size 25000 Simple Caching Anticipation 20000 S ize of Cache ( # o f ent ries) 15000 10000 5000 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Time (hours) DNS Cache Size / Time 10
DNS Misses and Reduction 100 100 Cache Miss Rate Anticipation Benefit Anticipation Miss Rate 90 90 80 80 70 70 Miss Rate C hange Actu al Miss Rate 60 60 50 50 40 40 30 30 20 20 10 10 0 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Time (hours) Time (hours) DNS Miss Rate Miss Rate Change 11
DNS Miss Reduction 20 Anticipation Benefit 15 Miss Rate Ch ange 10 5 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Time (hours) Magnification of MISS Rate Change 12
Prior and Related Work ■ Web Log Analysis ■ Web Anticipation ■ Web Cooperation – Squid – LSAM – Adaptive Web Caching 13
Conclusions ■ DNS caches must be local on client machines to be useful – 90% benefit – esp on ISDN connections ■ DNS-Web cooperation needs more exploration – 15% reduced misses – 3x space increase (<6MB total) 14
Future Work ■ Analyze real-time client traces – Squid logs wrong place in cache hierarchy – Real-time tracing allows examination of time components ■ Define DNS hits and misses – Some DNS misses are partial hits due to mulitple RTTs ■ Implementation of Cross-Domain system – Measure real benefits – Examine DNS aggregation 15
Recommend
More recommend