hierarchical video caching in wireless cloud: approaches ...€¦ · realizing advanced video...
TRANSCRIPT
Hierarchical Video Caching inWireless Cloud: Approaches and
Algorithms
IEEE ICC 2012 Workshop onRealizing Advanced Video Optimized Wireless Networks
June 15, 2012
Sujit DeyHastii Ahlehagh
UC San [email protected]
1
Outline
• Motivations• Hierarchical Caching Algorithms• Hierarchical Caching Simulation Results• Video Aware Backhaul Scheduling• Scheduling Simulation Results• Conclusion
2
Motivation for Hierarchical CachingPrevious: Video caching at the RAN to address congestionand improve delay and capacity of video deliveryIssues: 1. High cache miss ratio because of small caches
- Partially addressed by new UPP based caching2. High cache miss ratio due to user mobility
New Developments: Hierarchical caching to address above issues:1. Relatively larger caches at the CN nodes to supplement
RAN caches, while keeping total cache size same2. UPP based caching policies extended to hierarchical caches3. QoE aware Scheduling of RAN backhaul and CN resources
for video fetches from CDN4. Extended simulation framework to include video requests from
multiple cells, and mobility of users between cells
Effects:1. Improved coverage (higher cache hit ratio) than RAN-only caches2. Better address mobility, by increasing likelihood that a video currently being
downloaded is also in CN/RAN cache associated with the new cell during handoff3. Experimental results show significant improvements in cache hit ratio and
capacity, in particular when considering mobility3
What Videos do Users Watch?
• Most Popular Videos (Zipf distribution [1])• National video popularity may not reflect local
video popularity [2], and hence mobile videopreference in an individual cell site, which maydepend on demographics, time of day, etc.
• Strong preference towards some videocategories
SoccerESPN
CNNNews
FoxNews
HuluAuto
SoccerYou
Tube
CNNNews
FoxNews
Football
ESPN
Hulu
SoccerESPN
CNNNews
Hulu
Video Categories
[1] M. Cha et. al., "Analyzing the Video Popularity Characteristics of Large-Scale User Generated Content Systems",IEEE/ACM Transactions on Networking, Vol. 17, No. 5, October 2009
Category Average CumulativeViews – 90 Days
Autos 1501.8
Entertainment 1293.1
Comedy 1267.2
Arts & Animation 1106.3
Animals & Pets 1075.1
Science &Technology 794.6
Sports 745.1
How-To 432.4
Video Games 418.8
Family & Kids 328.1
News & Blogs 302.8
Vlogs 259.1
Travel 157.3
Commercials 124
Reelseo The Online Video Marketing Guide (www.reelseo.com )
[2] Michael Zink, et al.,“Watch Global Cache Local: YouTube Network Traces at a Campus Network - Measurements and Implications.” InProceedings of MMCN 2008, San Jose, CA, USA, Jan 2008
User Preference Profile (UPP): Each user has preference towards specific categories(and sources) of videos watched
4
Cell Site Video Preference [1]
0.34
0.3
0.18
0.060.06 0.060.32
0.34
0.34
0.33
0.32
0.26
0.030.03 0.03
Cell UPP Soccer ESPN (SE)Soccer YouTube (SY)CNN News (CN)Fox News (FN)Football ESPN (FE)Hulu (H)
User 1 User 2
Assumption: Users request video with equal probability
UPP of Active Users in the cell:
= ( )| |=1
Likelihood that a video is requested by the Active Users in the cell?
( ) = ( , )| |=1
Most Likely Requested (MLR):
Least Likely Requested (LLR):{ | = 1. . . . ( ) > }{ . . ( ) = }
Given overall MPV, video popularity distribution inVideo Category, vcj: ( , ) = ( )∑ ( )| |=1( ) = ( ) if belongs to category , else ( ) = 0.
5
[1] H. Ahlehagh, S, Dey, “Video Caching in Radio Access Network: Impact on Delay and Capacity”,In Proc. IEEE Wireless Communications and Networking Conference, Paris, France, April 2012.
Policies for RAN Micro-Caches [1]• MPV
– Proactively cache the most popular videos from MPV list, subject to cache size;– Cache content of each cell site is identical
• LRU– Reactively cache videos that are associated with cache misses– If cache is full, evict the least recently used video
• P-UPP– If active users change, proactively cache videos from new MLR list– Developed techniques to reduce bandwidth needed to download new MLR
videos, without affecting significantly cache hit ratio• R-UPP
– Reactively cache videos associated with cache misses– If cache is full, use LLR list to select candidate(s) for eviction
• Don’t update cache if the requested video causing the miss is in LLR itself
[1] H. Ahlehagh, S, Dey, “Video Caching in Radio Access Network: Impact on Delay and Capacity”,In Proc. IEEE Wireless Communications and Networking Conference, Paris, France, April 2012.
3G 4G Hierarchical Cache
Model wireless network as a tree where the videos traverse down the treeand AUS Information up the hierarchical tree.
7
Hierarchical Caching withinthe Wireless Network
How to distribute the caches most effectively to improve individual cellcoverage and support mobility across cells? Trade off between coverage and mobility
Hierarchical Caching Options1. Inclusive Caching: Suitable for Mobility because
if a user moves from one cell to another, theassociated video that is currently beingdownloaded is guaranteed to be found in the2nd layer cache. 2nd layer cache is limited in size so it may not be able
to replicate all the videos of the RAN caches
2. Exclusive Caching: Improves coverage, as the2nd layer cache stores videos that do not exist inthe 1st layer caches. This leads to overall higherhit ratio. A cache in the 1st layer, e.g. L13 might contain videos
that are more useful for L11 than the rest of thevideos in universe; L11 couldn’t cache the videos dueto the cache size constraints; Not suitable formobility
8
L1nL11
L21
Optimal for Mobility
L21=L11U… U L1n
1
L1nL11
L21
Optimal for Coverage
L21=VR(L11) U…VR(L1n) - {L11U… U L1n}
2
3. Our hybrid approach: Each cache makes its cachingdecision independently based on its AUS, e.g. if AUS ofL21 favors VC2 category, L21 caches more videos of thatcategory. To improve coverage, intersection of all the 1st
layer caches are removed from the 2nd layer cache Suitable for mobility and improves coverageL1nL11
L21
Hierarchical UPP based3
= U … − ∩ ∩
VR(L11): videos that L11 would have cached if enough space was available
Policies for Hierarchical Caching• H-MPV
– Each cache in the hierarchy stores videos according to the Most Popular ranking;A higher layer cache excludes the intersection of all the lower layer caches
• H-LRU– Hierarchical LRU is a straight-forward extension of the single-layer LRU and has
built-in exclusivity; No further optimization is required
Extended RAN-only UPP caching policies to higher layer caches In the hierarchical settings, the AUS of the higher layer node is defined as the union
of all AUSs of the lower layer nodes (child nodes) connected to it• H-P-UPP
– Proactively caches videos based on the UPP of the AUS of each cache.Intersection of all the child caches is excluded from the parent cache
• H-R-UPP– Reactively caches each video associated with a miss; while the video is traversing
towards the leaf in the hierarchy tree, each cache on the way to the leaf decideswhether to cache the video based on its AUS
– use LLR list to select candidate(s) for eviction Don’t update cache if the requested video causing the miss is in LLR itself
9
10
Video request, V
End
Yes
R-UPP
Layer l = 1
Layer l ++
• Calculate UPP for cache(l)• Calculate request probability, PR , based on
UPP of cache(l)• Calculate MLR and LLR sets• Update the cache if the PR (V) - ∑PR (LLR)
> Threshold)
Is V inCache(l)
?
• Download fromCache(l)
• l --
Yes
NoIsl==0?
Layer l --
No
Yes
AUS Changes in an eNodeB(User arrival or departure)
Morecache
layers?
End
Yes
P-UPP
Layer l = 1
Layer l ++
• Calculate UPP for cache(l)• Calculate request probability, PR , based on UPP of cache(l)• Generate MLR and LLR sets• Update the cache if the PR (MLR) - ∑PR (LLR) > Threshold
(Request probability of the video that need to be added fromMLR is greater than sum probability of videos that need tobe evicted for that video to fit in the cache plus a threshold)
Hierarchical R-UPP and P-UPP Policies
Scheduling Cache Misses• Videos resulting in cache misses have to be fetched from CDNs• Backhaul Bandwidth is a limited/shared resourcemay lead to increase in video delay, and/or impact on capacity
• RAN Backhaul Scheduler: coordinates with video clients, and determinesappropriate rates that can be allocated, so as to– Maximize Video Capacity (number of concurrent video requests that
can be served),– While meeting Video QoE (maximum initial video delay, and ensuring
no stalling during playback).• Relationship between data rate and QoE established by use of
Leaky Bucket Parameters (LBP) [4]: N 3-tuples (R, B, F),R: transmission rate, B: buffer size, F: initial fullness– If data rate is R, and client waits for initial delay of
F/R secs, no stalling during playback.
[4] J. Ribas-Corbera, et al., “A Generalized Hypothetical Reference Decoder for H.264/AVC”, IEEE Transactions on Circuits and Systems, vol. 13, no. 7, July 200311
Hierarchical Backhaul Scheduling• Using LBPs, the video client requests the lowest rate that
satisfies its initial delay requirements,• Successful scheduling depends upon the availability of sufficient
backhaul bandwidth; otherwise the request will be blocked.• After any change in the state of the current backhaul
downloads, recalculate the spare backhaul bandwidth capacityand allocate the spare resource, using the LP formulation below,among the users that have been admitted:Maximize: ∑Subject to: ≥ ∀∑ ∈ ≤ = 1, . . ,
bi: bandwidth of the ith flow, ri: the minimum rate required,Fn: set of flows that go through the nth backhaul
12
L2,1
CACHE
L1,1
CACHE
L3,1
CACHE
C2,3
Internet
CDN
L1,2
CACHE
C1,1 C1,2
C3
UkUi…
1 M
bps
Admit the requests according to theirminimum required rate (LBP table)
Redistribute the spare capacity usingLP formulation
(R,B,F) Initial Delay
(193K, 9216, 8146) 42.20 sec
(400K, 9216, 6216) 15.54 sec
(500K, 9216, 6216) 12.42 sec Video’s LBP
Simulation Environment andParameters
• MATLAB based framework to simulate end-to-end video cloud eco-system,including user arrival/departure from a cell, mobility model, UPPs, video requestgeneration, cache policies, backhaul video scheduling
• Mobility is modeled using a Poisson process: users are moved from one cell to aneighboring cell with the mean active cell time of 100seconds
13
Variable Distribution/Parameters Value
Total Number of Videos, Video Requests, VideoCategories, Video Popularity Distribution
20,000, 100,000, 250, Zipf 0.8
Video Frame Size Distribution Proposed in [5]
Video Size min=2, mean=8, max=30 minutes
Video Bit Rate 200kbps (QVGA) to 2Mbps (HD)
Total number of mobile users 5,000UPP Distribution Across VCs Exponential, 2User Arrival/Departure Model and VideoRequest Arrival
Poisson:Mean user inter-arrival time = 100 secondsMean user active time = 2700 secondsMean inter-arrival time per user = 120seconds
Backhaul Delay Thresholds [10,20,30]sec,30sec (LP)Cache Size 50Gbyte, 100Gbyte, 150GbyteMobility Model Poisson: mean active cell time = 100sec
Hierarchical Caching
[5] D. M. B. Masi, et al., “Video Frame Size Distribution Analysis”, The Telecommunications Review 2008, Volume 19, Noblis, Falls Church, VA, September 2008
RAN-only Caching
6PGW-only Caching
Simulation Results
14
50 100 1500
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
Cache Size (Gbyte)(a)
CacheHitRatio
MPV
H-MPVLRU
H-LRU
R-UPP
H-R-UPPP-UPP
H-P-UPP
RAN CN Internet0
50
100
150
200
250
300
350
Backhaul Layer(b)
MeanBackhaulBWrequired
MPV
H-MPV
LRU
H-LRUR-UPP
H-R-UPP
P-UPP
H-P-UPP
NoCache MPV LRU R-UPP P-UPP0
100
200
300
400
500
600
700
Cache Policy(c)
Capacity
RAN only CacheHierarchy Cache
NoCache MPV LRU R-UPP P-UPP0
100
200
300
400
500
600
700
800
Cache Policy(d)
Capacity
RAN only Cache w ith mobilityHierarchy Cache w ith mobility
• Hierarchical caching improves cache hit ratio by up to 25 percentage pointcompared to RAN only caching
• Hierarchical caching can improve network capacity by 21% and 30% using H-P-UPP and H-R-UPP compared to RAN only caching
• UPP based hierarchical caching policies perform significantly better in the caseof mobility; hierarchical UPP performs 47% better than RAN-only R-UPP interms of capacity
Cache Size=150Gbyte
RAN CN Internet0
50
100
150
200
250
300
350
Backhaul Layer(b)
MeanBackhaulBWrequired
MPVH-MPVPGW-MPVLRUH-LRUPGW-LRUR-UPPH-R-UPPPGW-R-UPPP-UPPH-P-UPPPGW-P-UPP
50 100 1500
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Cache Size (Gbyte)(a)
CacheHitRatio
NoCache MPV LRU R-UPP P-UPP0
100
200
300
400
500
600
700
800
Cache Policy(d)
Capacity
RAN only Cache with mobilityHierarchy Cache with mobilityPGW only Cache with Mobility
NoCache MPV LRU R-UPP P-UPP0
100
200
300
400
500
600
700
800
Cache Policy(c)
Capacity
MPVH-MPVPGW MPVLRUH-LRUPGW-LRUR-UPPH-R-UPPPGW-R-UPPP-UPPH-P-UPPPGW-P-UPP
RAN only CacheHierarchy CachePGW only Cache
Simulation Results
15
Cache Size=150Gbyte
• Hierarchical caching improves cache hit ratio by up to 25 percentage pointcompared to RAN-only caching
• Hierarchical caching can improve network capacity by 21% and 30% using H-P-UPP and H-R-UPP compared to RAN only caching
• UPP based hierarchical caching policies perform significantly better in the case ofmobility; hierarchical UPP performs 47% better than RAN-only R-UPP in terms ofcapacity
• PGW-only caching improves cache hit ratio by 8 and 5 percentage point for P-UPP and R-UPP respectively compared to Hierarchical caching. However,improvement in the cache hit ratio comes with the cost of higher RAN and CNrequired backhaul BW compared to RAN-only and Hierarchical caching, andsignificantly lower capacity!
Delay of Scheduled Videos
16
Cache Size=150Gbyte
0 5 10 15 20 25 300.4
0.5
0.6
0.7
0.8
0.9
1
Delay (second)
CDFofDelayofScheduledVideos
MPVLRUR-UPPP-UPP
0 5 10 15 20 25 300.4
0.5
0.6
0.7
0.8
0.9
1
Delay (second)
CDFofDelayofScheduledVideos
H-MPVH-LRUH-R-UPPH-P-UPP
0 5 10 15 20 25 300.4
0.5
0.6
0.7
0.8
0.9
1
Delay (second)
CDFofDelayofScheduledVideos
PGW-MPVPGW-LRUPGW-R-UPPPGW-P-UPP
• RAN-only micro-caching performs the best in terms of probability of successfullyscheduled video requests that can meet certain` initial delay compared withHierarchical and PGW-only caching. For example, of the successfully scheduledrequests, the probability of achieving an initial delay of 5 second or less is about0.79 for P-UPP, 0.76 for R-UPP, 0.68 for H-P-UPP, 0.67 for H-R-UPP , and 0.52 forPGW-P-UPP and PGW-R-UPP.
Conclusion
• Proposed hierarchical caching of video contents in the CNto supplement RAN caches
• Showed that hierarchical caching can significantlyimprove cache hit ratio (for same cache size), andimprove capacity in particular under mobility conditions
• Showed that PWG-only caching can significantly improvecache hit ratio (for same cache size), however, it results inhigher required RAN and CN backhaul BW, andsignificantly lower end-to-end capacity because of thebottlenecks in the CN backhaul.
17