Asymmetric Caching: Improved Network Deduplication for Mobile Devices

Transcription

1 Asymmetric Caching: Improved Network Deduplication for Mobile Devices ShrutiSanadhya, 1 RaghupathySivakumar, 1 Kyu-Han Kim, 2 Paul Congdon, 2 SriramLakshmanan, 1 JatinderP Singh 3 1 Georgia Institute of Technology, Atlanta, GA, USA 2 HP Labs, Palo Alto, CA, USA 3 Xerox PARC, Palo Alto, CA, USA 1

2 Introduction Network traffic has a lot of redundancy 20% HTTP content accessed on smartphonesis redundant 1 Network deduplication(dedup) leverages this redundancy to conserve network bandwidth Regular Cache Mobile Cache C2 Mobile Sender Dedup Source SGSN Dedup Destination Receiver Rabin Fingerprinting C1 C 2 C3 Hashing H1 H3 Compress C1 C3 C1 C 2 C3 Inflate packet C1 C3 1 Qianet al,, Web Caching on Smartphones: Ideal vs. Reality, MobiSys2012 2

3 The Asymmetry Problem What happens when the mobile cache is more populated than the cache at dedup source? Regular Cache H3 C3 H4 C4 H5 C5 H6 C6 Mobile H7 C7 Cache H8 C8 C2 Mobile Cache Sender Dedup Source Dedup Destination Receiver How can all the past cached information at the mobile be successfully leveraged for dedup by any given dedup source? 3

4 Motivational Scenarios Multi-homed devices Cache WiFi Access Point Cache Mobile Device Cache 3G Base Station (BS) 4

5 Motivational Scenarios Multi-homed devices Resource pooling Cache Mobile Device BS BS RNC RNC BS: Base Station RNC: Radio Network Controller SGSN: Serving GPRS Support Node Cache Cache SGSN SGSN 5

6 Multi-homed devices Resource pooling Memory scalability Cache Motivational Scenarios Cache RNC Cache BS SGSN Cache BS: Base Station RNC: Radio Network Controller SGSN: Serving GPRS Support Node 6

7 Scope and Goals Scope Laptops/smartphones using 3G/WiFi Conserving cellular bandwidth Downstream and unencrypted traffic Goals Overall efficiency: Using downstream andupstream more efficiently Application agnostic: Applicable to any application Limited overheads: Deployable computational and memory complexities 7

8 Asymmetric Caching -Overview Regular Cache Feedback Cache H4 H3 C3 H4 C4 H5 C5 C2 Mobile Cache Sender Dedup Source Feedback Dedup Destination Receiver Mobile cache is more populated than dedup source On receiving downstream traffic, the mobile selectively advertises portions of its cache to dedup source Dedup source also maintains a feedback cache Both regular and feedback cache is used for dedup 8

9 When is feedback sent? Feedback is sent reactively Feedback is sent only when there is downstream traffic Feedback sent is specific to the ongoing traffic Downstream traffic Dedup Source Dedup Destination Feedback 9

10 Where from is feedback selected? Hashes at dedup destination can be organized as per: Order of arrival H1 H3 H4 H5 H6 H7 H8 H9 H10 Same flow (SrcIP, DestIP, SrcPort, DestPort ) H1 H6 H3 H4 H5 H7 H8 H9 H10 Same object (HTML, JPEG or CSS) H1 H6 H7 H3 H4 H9 H10 H8 H5 Objects help in effectively matching new and old content Application agnostic estimate of objects are flowlets 10

11 How are flowlets extracted? Sequence of bytes in a flow is a time-series Flowlets are piecewise stationary segments of a flow Check for flowlet boundary at start of each packet Consider byte series B [0:m] (1 st packet), B [m+1:n] (2 nd packet) and B [0:n] as autoregressive processes of order p: B [0:m] B i = Σ 1<=j<=p a i B i-j + σε, εis white noise d [0:m:n] = gain(b [0:n] ) gain(b [0:m] ) gain(b [m+1:n] ) B 0, B 1,.., B m, B m+1,, B n B [0:n] Gain in the noise power when B [0:n] is in one flowlet instead of different flowlets: B [0:m] and B [m+1:n] If d [0:m:n] > d thresh, then flowlet boundary exists at m B [m+1:n] 11

12 How is feedback selected? Find best matching past flowlet 12

13 How is feedback selected? Find best matching past flowlet H1 13

14 How is feedback selected? Find best matching past flowlet H1 H1 F1, F3 F1, F2, F3 14

15 How is feedback selected? Find best matching past flowlet H1 F1:H1,, H4, H5, H6, H7, H8, H9, H10, H11, H12,. H1 F1, F3 F1, F2, F3 F2:, H5, H10,. F3: H5, H8, H11, H12,.. Last hash matched 15

16 How is feedback selected? Find best matching past flowlet H1 F1:H1,, H4, H5, H6, H7, H8, H9, H10, H11, H12,. H1 F1, F3 F1, F2, F3 F2:, H5, H10,. F3: H5, H8, H11, H12,.. Last hash matched 16

17 How is feedback selected? Find best matching past flowlet H1 F1:H1,, H4, H5, H6, H7, H8, H9, H10, H11, H12,. H1 F1, F3 F1, F2, F3 F2:, H5, H10,. F3: H5, H8, H11, H12,.. Flowlet 1 (F1) is best matched Last hash matched 17

18 How is feedback selected? Find best matching past flowlet H1 F1:H1,, H4, H5, H6, H7, H8, H9, H10, H11, H12,. H1 F1, F3 F1, F2, F3 F2:, H5, H10,. F3: H5, H8, H11, H12,.. Flowlet 1 (F1) is best matched Last hash matched Find start of next feedback in the best matching flowlet 18

19 How is feedback selected? Find best matching past flowlet H1 F1:H1,, H4, H5, H6, H7, H8, H9, H10, H11, H12,. H1 F1, F3 F1, F2, F3 F2:, H5, H10,. F3: H5, H8, H11, H12,.. Flowlet 1 (F1) is best matched Last hash matched Find start of next feedback in the best matching flowlet Best matching past flowlet H1,, H4, H5, H6,H7, H8, H9,H10, H11, H12, H13, 19

20 How is feedback selected? Find best matching past flowlet H1 F1:H1,, H4, H5, H6, H7, H8, H9, H10, H11, H12,. H1 F1, F3 F1, F2, F3 F2:, H5, H10,. F3: H5, H8, H11, H12,.. Flowlet 1 (F1) is best matched Last hash matched Find start of next feedback in the best matching flowlet Best matching past flowlet H1,, H4, H5, H6,H7, H8, H9,H10, H11, H12, H13, Last hash matched 20

21 How is feedback selected? Find best matching past flowlet H1 F1:H1,, H4, H5, H6, H7, H8, H9, H10, H11, H12,. H1 F1, F3 F1, F2, F3 F2:, H5, H10,. F3: H5, H8, H11, H12,.. Flowlet 1 (F1) is best matched Last hash matched Find start of next feedback in the best matching flowlet Best matching past flowlet H1,, H4, H5, H6,H7, H8, H9,H10, H11, H12, H13, Last hash matched Last hash advertised 21

22 How is feedback selected? Find best matching past flowlet H1 H1 Flowlet 1 (F1) is best matched Last hash matched Find start of next feedback in the best matching flowlet Best matching past flowlet F1, F3 F1, F2, F3 δ: temporal offset H1,, H4, H5, H6,H7, H8, H9,H10, H11, H12, H13, Last hash matched F1:H1,, H4, H5, H6, H7, H8, H9, H10, H11, H12,. F2:, H5, H10,. F3: H5, H8, H11, H12,.. Last hash advertised δ 22

23 How is feedback selected? Find best matching past flowlet H1 H1 Flowlet 1 (F1) is best matched Last hash matched Find start of next feedback in the best matching flowlet Best matching past flowlet F1, F3 F1, F2, F3 δ: temporal offset H1,, H4, H5, H6,H7, H8, H9,H10, H11, H12, H13, Last hash matched F1:H1,, H4, H5, H6, H7, H8, H9, H10, H11, H12,. F2:, H5, H10,. F3: H5, H8, H11, H12,.. Last hash advertised δ Start of next feedback 23

24 How is the feedback used? Regular Cache Feedback Cache Dedup Source Dedup source maintains a feedback cache along with regular cache of baseline dedup 24

25 How is the feedback used? Regular Cache H1 Feedback Cache Dedup Source Dedup source maintains a feedback cache along with regular cache of baseline dedup Regular cache is populated by downstream data 25

26 How is the feedback used? Regular Cache H1 Feedback Cache Dedup Source Dedup source maintains a feedback cache along with regular cache of baseline dedup Regular cache is populated by downstream data 26

27 How is the feedback used? Regular Cache H1 Feedback Cache H3 H4 Dedup Source Dedup source maintains a feedback cache along with regular cache of baseline dedup Regular cache is populated by downstream data Feedback hashes are inserted in feedback cache 27

28 How is the feedback used? Regular Cache H1 Feedback Cache H3 H4 Dedup Source Dedup source maintains a feedback cache along with regular cache of baseline dedup Regular cache is populated by downstream data Feedback hashes are inserted in feedback cache Every downstream packet is dedupedusing both regular and feedback cache 28

29 Design Summary When is the feedback sent? Where from is the feedback chosen? How are flowlets extracted? How is the feedback selected? How is the feedback used? Reactively Flowlets at dedup destination Stationarity properties Best matching flowlet and pointers in past flowlet Stored in the feedback cache for dedup 29

30 Trace Based Analysis Data collection 25 laptop and 5 smartphoneusers over 3 months giving 26GB of unsecured downlink data WiFias well as 3G network Packet sniffing through Wireshark and Tcpdump Trace analysis Custom analyzer implemented in Python Mimic mobility by splitting trace into two halves: past and present Past trace populates the initial cache at the dedup destination This is the data remembered from previous networks access 30 random connections from the present create ongoing traffic Dedup is performed using asymmetric caching 30

31 Trace Analysis Results -I Redundancy identified # Redundant bytes found by asymmetric caching Actual # redundant bytes x 100 Average asymmetriccaching Average symmetriccaching Asymmetric caching leverages significant portion of the achievable redundancy 31

32 Trace Analysis Results -II Feedback efficiency # Bytes saved downstream # Bytes sent upstream Split of total hits across the caches at dedup source Asymmetric caching generates efficient and relevant feedback 32

33 Related Work Network layer approaches Spring et al, A protocol-independent technique for eliminating redundant network traffic. SIGCOMM, Aggarwalet al, EndRE: an end-system redundancy elimination service for enterprises. NSDI, Shenet al, REfactor-ingcontent overhearing to improve wireless performance. MobiCom, 2011 Transport layer approaches Zoharet al, The power of prediction: cloud bandwidth and cost reduction, SIGCOMM 2011 Application layer approaches Web browser caches and proxies Content Distribution Networks (CDNs) 33

34 Conclusion and Future Work A dedup strategy that leverages past remembered on mobile devices to perform dedup at any dedup source Application agnostically estimate different objects in a flow byusing stationarity properties of different content Trace analysis of 30 users shows that asymmetric caching: Leverages 89% of achievable redundancy Gives 6x feedback efficiency Prototype implementation on Linux desktop and Android smartphone with deployable overheads Future Work: Upstream dedup, i.e. reduce redundant content sent upstream Extending dedup to end-to-end encrypted traffic Study energy impact of asymmetric caching on mobile devices 34

35 Questions? 35