FairRide: Near-Optimal Fair Cache Sharing
|
|
- Virgil Walters
- 5 years ago
- Views:
Transcription
1 UC BERKELEY FairRide: Near-Optimal Fair Cache Sharing Qifan Pu, Haoyuan Li, Matei Zaharia, Ali Ghodsi, Ion Stoica 1
2 Caches are crucial 2
3 Caches are crucial 2
4 Caches are crucial 2
5 Caches are crucial 2
6 Cache sharing 3
7 Cache sharing Increasingly, caches are shared among multiple users 3
8 Cache sharing Increasingly, caches are shared among multiple users Especially with the advent of cloud 3
9 Cache sharing Increasingly, caches are shared among multiple users Especially with the advent of cloud Backend (storage/network) 3
10 Cache sharing Increasingly, caches are shared among multiple users Especially with the advent of cloud Backend (storage/network) 3
11 Cache sharing Increasingly, caches are shared among multiple users Especially with the advent of cloud Benefits: Provide low latency Reduce backend load Backend (storage/network) 3
12 Cache sharing Increasingly, caches are shared among multiple users Especially with the advent of cloud Benefits: Provide low latency Reduce backend load * Cache Backend (storage/network) 3
13 Problems with cache algorithms Cache * Backend (storage/network) 4
14 Problems with cache algorithms LRU, LFU, LRU-K Cache * Backend (storage/network) 4
15 Problems with cache algorithms LRU, LFU, LRU-K Cache * Backend (storage/network) 4
16 Problems with cache algorithms LRU, LFU, LRU-K Cache * Backend (storage/network) 4
17 Problems with cache algorithms LRU, LFU, LRU-K Cache * Backend (storage/network) 4
18 Problems with cache algorithms LRU, LFU, LRU-K Cache data likely to be accessed in the future Cache * Backend (storage/network) 4
19 Problems with cache algorithms Cache LRU, LFU, LRU-K Cache data likely to be accessed in the future Optimize global efficiency * Backend (storage/network) 4
20 Problems with cache algorithms Cache * Backend (storage/network) LRU, LFU, LRU-K Cache data likely to be accessed in the future Optimize global efficiency Single user gets arbitrarily small cache 4
21 Problems with cache algorithms Cache * Backend (storage/network) LRU, LFU, LRU-K Cache data likely to be accessed in the future Optimize global efficiency Single user gets arbitrarily small cache 4
22 Problems with cache algorithms Cache * Backend (storage/network) LRU, LFU, LRU-K Cache data likely to be accessed in the future Optimize global efficiency Single user gets arbitrarily small cache 4
23 Problems with cache algorithms Cache * Backend (storage/network) LRU, LFU, LRU-K Cache data likely to be accessed in the future Optimize global efficiency Single user gets arbitrarily small cache Prone to strategic behavior 4
24 Problems with cache algorithms Cache * Backend (storage/network) LRU, LFU, LRU-K Cache data likely to be accessed in the future Optimize global efficiency Single user gets arbitrarily small cache Prone to strategic behavior 4
25 Problems with cache algorithms Cache * Backend (storage/network) LRU, LFU, LRU-K Cache data likely to be accessed in the future Optimize global efficiency Single user gets arbitrarily small cache Prone to strategic behavior 4
26 Solution? Statically allocated Cache Cache Cache Backend (storage/network)
27 Solution? Statically allocated Cache Cache Cache Backend (storage/network) Isolation Strategy-proof
28 What we want Statically allocated Globally shared Cache Cache Cache Cache * Backend (storage/network) Isolation Strategy-proof Backend (storage/network) 6
29 What we want Statically allocated Globally shared Cache Cache Cache Cache * Backend (storage/network) Isolation Strategy-proof Backend (storage/network) Higher utilization Share data 6
30 What we want Statically allocated Globally shared Cache Cache Cache Cache * Backend (storage/network) Isolation Strategy-proof Backend (storage/network) Higher utilization Share data 6
31 Our contribution 7
32 Our contribution First analysis of cache allocation policies with well defined resource-sharing properties. 7
33 Our contribution First analysis of cache allocation policies with well defined resource-sharing properties. Impossibility result: no policy achieves all good properties! 7
34 Our contribution First analysis of cache allocation policies with well defined resource-sharing properties. Impossibility result: no policy achieves all good properties! A new policy that is near-optimal and outperforms other policies when users cheat. 7
35 A simple model 8
36 A simple model Users access equal-sized files at constant rates r ij the rate user i accesses file j 8
37 A simple model Users access equal-sized files at constant rates r ij the rate user i accesses file j A allocation policy decides which files to cache p j the % of file j put in cache 8
38 A simple model Users access equal-sized files at constant rates r ij the rate user i accesses file j A allocation policy decides which files to cache p j the % of file j put in cache Users care their hit ratio user i s hit ratio: 8
39 A simple model Users access equal-sized files at constant rates r ij the rate user i accesses file j A allocation policy decides which files to cache p j the % of file j put in cache Users care their hit ratio user i s hit ratio: HR i = total _ hits total _ accesses = å j år ij j p j r ij 8
40 A simple model Users access equal-sized files at constant rates r ij the rate user i accesses file j A allocation policy decides which files to cache p j the % of file j put in cache Users care their hit ratio user i s hit ratio: HR i = total _ hits total _ accesses = å j år ij j p j r ij Results hold with varied file sizes, access partial files, p j is binary, etc. 8
41 Outline What properties do we want? Can we extend max-min to solve the problem? How do we solve it? (FairRide) How well does FairRide work in practice? 9
42 Properties Isolation Guarantee No user should be worse off than static allocation 10
43 Properties Isolation Guarantee No user should be worse off than static allocation Strategy-Proofness No user can improve by cheating 10
44 Properties Isolation Guarantee No user should be worse off than static allocation Strategy-Proofness No user can improve by cheating Pareto Efficiency Can t improve a user without hurting others 10
45 Strategy proofness 11
46 Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses 11
47 Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses Can happen in practice 11
48 Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses Can happen in practice Site1 Site2 MySQL Instance 11
49 Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses Can happen in practice Site1 Site2 Amazon Elasticache * MySQL Instance 11
50 miss ratio (%) Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses Can happen in practice Site1 Site Amazon Elasticache * MySQL Instance 6 0 site1 site time (min) 11
51 miss ratio (%) Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses Can happen in practice Site1 Site Amazon Elasticache * MySQL Instance 6 0 site1 site time (min) 11
52 miss ratio (%) Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses Can happen in practice Site1 Site Amazon Elasticache * MySQL Instance 6 0 site1 site time (min) 11
53 miss ratio (%) Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses Can happen in practice Site1 Site Amazon Elasticache * MySQL Instance 6 0 site1 site time (min) 11
54 miss ratio (%) Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses Can happen in practice Site1 Site Amazon Elasticache * MySQL Instance 6 0 site1 site time (min) 11
55 miss ratio (%) Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses Can happen in practice Site1 Site Amazon Elasticache * MySQL Instance 6 0 site1 site time (min) 11
56 miss ratio (%) Strategy proofness Very easy to cheat, hard to detect e.g., by making spurious accesses Can happen in practice Site1 Site2 18 Amazon Elasticache * MySQL Instance site1 site2 2x time (min) 11
57 Properties Isolation Guarantee No user should be worse off than static allocation Strategy-Proofness No user can improve by cheating Pareto Efficiency Can t improve a user without hurting others 12
58 Outline What properties do we want? Can we extend max-min fairness to solve the problem? How do we solve it? (FairRide) How well does FairRide work in practice? 13
59 What is max-min fairness? 14
60 What is max-min fairness? Maximize the the user with minimum allocation Solution: allocate each 1/n (fair share) 33% 33% 33% 14
61 What is max-min fairness? Maximize the the user with minimum allocation Solution: allocate each 1/n (fair share) 33% 33% 33% Handles if some users want less than fair share 20% 40% 40% 14
62 What is max-min fairness? Maximize the the user with minimum allocation Solution: allocate each 1/n (fair share) 33% 33% 33% Handles if some users want less than fair share 20% 40% 40% Widely successful to other resources: OS: round robin, prop sharing, lottery sched Networking: fair queueing, wfq, wf2q, csfq, drr Datacenter: DRF, Hadoop fair sched, Quincy 14
63 An example 15
64 An example A B C 15
65 An example Alice A B Bob C 15
66 An example Alice 5 req/sec A B Bob 5 req/sec C 15
67 An example Alice 5 req/sec A GB Bob 5 req/sec B C file sizes = 1GB, total cache = 2GB 15
68 An example Alice 5 req/sec A GB Bob 5 req/sec B 100% C file sizes = 1GB, total cache = 2GB 15
69 An example Alice 5 req/sec A B GB Bob 5 req/sec B 100% C file sizes = 1GB, total cache = 2GB 15
70 An example Alice 5 req/sec A A B GB Bob 5 req/sec B 100% C file sizes = 1GB, total cache = 2GB 15
71 An example Alice A B GB 5 req/sec A 50% B 100% Bob 5 req/sec C file sizes = 1GB, total cache = 2GB 15
72 An example Alice 5 req/sec A 50% A Bob B C GB 5 req/sec B 100% C 50% file sizes = 1GB, total cache = 2GB 15
73 An example Alice HR = 83.3% A Bob B C GB 5 req/sec A 5 req/sec 50% B 100% C 50% file sizes = 1GB, total cache = 2GB 15
74 An example Alice HR = 83.3% A Bob HR = 83.3% B C GB 5 req/sec A 5 req/sec 50% B 100% C 50% file sizes = 1GB, total cache = 2GB 15
75 Properties max-min fairness Isolation Guarantee Strategy Proofness Pareto Efficiency 16
76 Properties max-min fairness Isolation Guarantee Strategy Proofness Pareto Efficiency 16
77 Properties Isolation Guarantee max-min fairness? Strategy Proofness Pareto Efficiency 16
78 An example Alice HR = 83.3% A B C GB Bob HR = 83.3% 5 req/sec A 5 req/sec 50% B 100% C 50% file sizes = 1GB, total cache = 2GB 17
79 An example Alice HR = 83.3% A B C GB Bob HR = 83.3% 5 req/sec A 5 req/sec 50% B 100% C 50% file sizes = 1GB, total cache = 2GB 17
80 An example Alice HR = 83.3% Bob HR = 83.3% 5 req/sec A Is it possible for a strategic Bob to 5 req/sec 50% get higher hit rate from the system? B A B C GB 100% C 50% file sizes = 1GB, total cache = 2GB 17
81 An example Alice HR = 83.3% A B C GB Bob HR = 83.3% 5 req/sec A 5 req/sec 50% B 100% C 50% file sizes = 1GB, total cache = 2GB 17
82 An example Alice HR = 83.3% A B C GB Bob HR = 83.3% 5 req/sec A 5 req/sec +10 req/sec 50% B 100% C 50% file sizes = 1GB, total cache = 2GB 17
83 An example Alice HR = 83.3% A B C GB Bob HR = 83.3% 5 req/sec A 5 req/sec +10 req/sec 50% B 100% C 100% 50% file sizes = 1GB, total cache = 2GB 17
84 An example Alice HR = 83.3% A B C GB Bob HR = 83.3% 5 req/sec A 5 req/sec +10 req/sec 50% B 50% 100% C 100% 50% file sizes = 1GB, total cache = 2GB 17
85 An example Alice HR = 83.3% B C GB Bob HR = 83.3% 5 req/sec A 5 req/sec +10 req/sec B 100% 50% 100% C 100% 50% file sizes = 1GB, total cache = 2GB 17
86 An example Alice HR = 83.3% B C GB Bob HR = 83.3% 5 req/sec A 5 req/sec +10 req/sec B 100% 50% 100% C 100% 50% file sizes = 1GB, total cache = 2GB 17
87 An example Alice HR = 83.3% B C GB Bob HR = 83.3% 100% 5 req/sec A 5 req/sec +10 req/sec B 100% 50% 100% C 100% 50% file sizes = 1GB, total cache = 2GB 17
88 An example Alice HR = 83.3% B C GB Bob HR = 83.3% 100% 5 req/sec A 5 req/sec +10 req/sec B 100% 50% 100% C 100% 50% file sizes = 1GB, total cache = 2GB 17
89 An example Alice HR = 83.3% 66.7% B C GB Bob HR = 83.3% 100% 5 req/sec A 5 req/sec +10 req/sec B 100% 50% 100% C 100% 50% file sizes = 1GB, total cache = 2GB 17
90 An example Alice HR = 83.3% 66.7% B C GB Bob HR = 83.3% 100% 5 req/sec A 5 req/sec +10 req/sec B 100% 50% 100% By gaming the system, a user can C 100% 50% increase performance by hurting others! file sizes = 1GB, total cache = 2GB 17
91 Properties max-min fairness Isolation Guarantee Strategy Proofness Pareto Efficiency 18
92 Properties Isolation Guarantee Strategy Proofness Pareto Efficiency max-min fairness 18
93 Properties Isolation Guarantee Strategy Proofness Pareto Efficiency max-min fairness static allocation 18
94 Properties Isolation Guarantee Strategy Proofness Pareto Efficiency max-min fairness static allocation priority allocation 18
95 Properties Isolation Guarantee Strategy Proofness Pareto Efficiency max-min fairness static allocation priority allocation max-min rate 18
96 Theorem No allocation policy can satisfy all three properties! 19
97 Theorem No allocation policy can satisfy all three properties! Best we can do: two of three. 19
98 What makes cache sharing different? 20
99 What makes cache sharing different? Unlike CPU or network links: 20
100 What makes cache sharing different? Unlike CPU or network links: The cost of switching is high Cache misses go to slow storage Prevents efficient time multiplexing 20
101 What makes cache sharing different? Unlike CPU or network links: The cost of switching is high Cache misses go to slow storage Prevents efficient time multiplexing Can be shared in space Shared data can be accessed non-exclusively A CPU cycle used by only one thread A network link sends one packet at a time 20
102 Outline What properties do we want? Can we extend max-min to solve the problem? How do we solve it? (FairRide) How well does FairRide work in practice? 21
103 Properties Isolation Guarantee Strategy Proofness Pareto Efficiency max-min fairness static allocation priority allocation max-min rate 103
104 Properties Isolation Guarantee Strategy Proofness Pareto Efficiency max-min fairness static allocation priority allocation max-min rate FairRide 104
105 Properties Isolation Guarantee Strategy Proofness Pareto Efficiency max-min fairness static allocation priority allocation max-min rate FairRide 105
106 Properties Isolation Guarantee Strategy Proofness Pareto Efficiency max-min fairness static allocation priority allocation max-min rate FairRide Near-optimal 106
107 FairRide 107
108 FairRide Starts with max-min fairness Allocate 1/n to each user Split cost of shared files equally among shared users 108
109 FairRide Starts with max-min fairness Allocate 1/n to each user Split cost of shared files equally among shared users Only difference: blocking users who don t pay from accessing 109
110 FairRide Starts with max-min fairness Allocate 1/n to each user Split cost of shared files equally among shared users Only difference: blocking users who don t pay from accessing Probabilistic blocking: with some probability 110
111 FairRide Starts with max-min fairness Allocate 1/n to each user Split cost of shared files equally among shared users Only difference: blocking users who don t pay from accessing Probabilistic blocking: with some probability Implemented with delaying 111
112 FairRide: Blocking Alice A B C GB Bob HR = 83.3% 100% 5 req/sec +10 req/sec B 100% C 100% 24
113 FairRide: Blocking Alice A B C GB Bob HR = 83.3% 100% 5 req/sec +10 req/sec B 100% C 100% 24
114 FairRide: Blocking Alice B C GB Bob HR = 83.3% 100% 5 req/sec +10 req/sec Allow 5 Block 5 A B 100% C 100% 24
115 FairRide: Blocking Alice B C GB Bob HR = 83.3% 100% 66.7% 5 req/sec +10 req/sec Allow 5 Block 5 A B 100% C 100% 24
116 FairRide: Blocking Alice B C GB Bob HR = 83.3% 100% 66.7% 5 req/sec +10 req/sec Allow 5 Block 5 A B 100% C 100% 24
117 FairRide: Blocking Alice A C GB B 100% Cheating always gives worst performance. Bob HR = 83.3% 100% 66.7% 5 req/sec +10 req/sec Allow 5 Block 5 Dis-incentive strategic behaviors. 100% B C 24
118 Probabilistic blocking 25
119 Probabilistic blocking FairRide blocks a user with p(nj) = 1/(nj+1) probability nj is number of other users caching file j e.g., p(1)=50%, p(4)=20% 25
120 Probabilistic blocking FairRide blocks a user with p(nj) = 1/(nj+1) probability nj is number of other users caching file j e.g., p(1)=50%, p(4)=20% The best you can do in a general case 25
121 Probabilistic blocking FairRide blocks a user with p(nj) = 1/(nj+1) probability nj is number of other users caching file j e.g., p(1)=50%, p(4)=20% The best you can do in a general case Less blocking does not prevent cheating 25
122 Properties Isolation Guarantee Strategy Proofness Pareto Efficiency max-min fairness static allocation priority allocation max-min rate FairRide Near-optimal 122
123 vs. Strategyproofness Paretoefficiency 27
124 vs. Strategyproofness Paretoefficiency More efficient when user cheats Minimal impact on efficiency when no user cheats 27
125 vs. Strategyproofness Paretoefficiency More efficient when user cheats Minimal impact on efficiency when no user cheats Cost of cheating vs. cost of blocking/delaying 27
126 vs. Strategyproofness Paretoefficiency More efficient when user cheats Minimal impact on efficiency when no user cheats Cost of cheating vs. cost of blocking/delaying The latter is small insurance for the former 27
127 vs. Strategyproofness Paretoefficiency More efficient when user cheats Minimal impact on efficiency when no user cheats Cost of cheating vs. cost of blocking/delaying The latter is small insurance for the former Strategy-proofness makes the system stable 27
128 Outline What properties do we want? Can we extend max-min to solve the problem? How do we solve it? (FairRide) How well does FairRide work in practice? 28
129 Evaluation Implemented in Alluxio (formerly Tachyon) FairRide: delay a request as if blocked Compared with max-min fairness. Benchmarked with TPC-H, YCSB, Facebook workloads. 29
130 Evaluation Does FairRide prevent cheating? What is the cost of FairRide? How does it perform end-to-end? 30
131 Avg. response (ms) miss ratio (%) Cheating under Max-min fairness user 1 user Time (s) 31
132 Avg. response (ms) miss ratio (%) Cheating under Max-min fairness user 1 user Time (s) 31
133 Avg. response (ms) miss ratio (%) Cheating under Max-min fairness user 2 cheats user 1 user Time (s) 31
134 Avg. response (ms) miss ratio (%) Cheating under Max-min fairness user 2 cheats user 1 user Time (s) 31
135 Avg. response (ms) miss ratio (%) Cheating under Max-min fairness user 2 cheats user 1 cheats user 1 user Time (s) 31
136 Avg. response (ms) miss ratio (%) Cheating under Max-min fairness user 2 cheats user 1 cheats user 1 user Time (s) 31
137 Avg. response (ms) miss ratio (%) Cheating under Max-min fairness user 2 cheats user 1 cheats user 1 user Time (s) Cheating can greatly hurt user performance. 31
138 Avg. response (ms) miss ratio (%) Cheating under FairRide Time (s) user 1 user 2 32
139 Avg. response (ms) miss ratio (%) Cheating under FairRide Time (s) user 1 user 2 32
140 Avg. response (ms) miss ratio (%) Cheating under FairRide user 2 cheats Time (s) user 1 user 2 32
141 Avg. response (ms) miss ratio (%) Cheating under FairRide user 2 cheats Time (s) user 1 user 2 32
142 Avg. response (ms) miss ratio (%) Cheating under FairRide user 2 cheats Time (s) user 1 cheats user 1 user 2 32
143 Avg. response (ms) miss ratio (%) Cheating under FairRide user 2 cheats Time (s) user 1 cheats user 1 user 2 FairRide dis-incentives users from cheating. 32
144 Avg. response (ms) Avg. miss ratio (%) Many users strategic users other users No. of strategic users (out of 20) 33
145 Avg. response (ms) Avg. miss ratio (%) Many users FairRide: 351ms 333 strategic users other users No. of strategic users (out of 20) 33
146 Avg. response (ms) Avg. miss ratio (%) Many users FairRide: 351ms 333 strategic users other users No. of strategic users (out of 20) 33
147 Avg. response (ms) Avg. miss ratio (%) Many users FairRide: 351ms 333 strategic users other users No. of strategic users (out of 20) FairRide has minimal loss. 33
148 Redcution in Median Job Time (%) Facebook experiments max-min FairRide Bin (#Tasks) 34
149 Redcution in Median Job Time (%) Facebook experiments max-min FairRide Bin (#Tasks) FairRide outperforms max-min fairness by 29% 34
150 Conclusion No policy can satisfy all desirable properties: Isolation guarantee Strategy proofness Pareto efficiency 35
151 Conclusion No policy can satisfy all desirable properties: Isolation guarantee Strategy proofness Pareto efficiency FairRide:isolation guarantee and strategyproofness through probabilistic blocking. Outperforms static allocation and other sharing policies when users cheat. Achieves this with least overhead 35
Cache Management for In Memory. Jun ZHANG Oct 15, 2018
Cache Management for In Memory Analytics Jun ZHANG Oct 15, 2018 1 Outline 1. Introduction 2. LRC: Dependency aware caching 3. OpuS: Fair cache sharing in multi tenant cloud 4. SP Cache: Load balancing
More informationKey aspects of cloud computing. Towards fuller utilization. Two main sources of resource demand. Cluster Scheduling
Key aspects of cloud computing Cluster Scheduling 1. Illusion of infinite computing resources available on demand, eliminating need for up-front provisioning. The elimination of an up-front commitment
More informationAli Ghodsi, Matei Zaharia, Benjamin Hindman, Andy Konwinski, Scott Shenker, Ion Stoica. University of California, Berkeley nsdi 11
Dominant Resource Fairness: Fair Allocation of Multiple Resource Types Ali Ghodsi, Matei Zaharia, Benjamin Hindman, Andy Konwinski, Scott Shenker, Ion Stoica University of California, Berkeley nsdi 11
More informationToday s Papers. Composability is Essential. The Future is Parallel Software. EECS 262a Advanced Topics in Computer Systems Lecture 13
EECS 262a Advanced Topics in Computer Systems Lecture 13 Resource allocation: Lithe/DRF October 16 th, 2012 Today s Papers Composing Parallel Software Efficiently with Lithe Heidi Pan, Benjamin Hindman,
More informationKey aspects of cloud computing. Towards fuller utilization. Two main sources of resource demand. Cluster Scheduling
Key aspects of cloud computing Cluster Scheduling 1. Illusion of infinite computing resources available on demand, eliminating need for up-front provisioning. The elimination of an up-front commitment
More informationMesos: Mul)programing for Datacenters
Mesos: Mul)programing for Datacenters Ion Stoica Joint work with: Benjamin Hindman, Andy Konwinski, Matei Zaharia, Ali Ghodsi, Anthony D. Joseph, Randy Katz, ScoC Shenker, UC BERKELEY Mo)va)on Rapid innovaeon
More informationSparrow. Distributed Low-Latency Spark Scheduling. Kay Ousterhout, Patrick Wendell, Matei Zaharia, Ion Stoica
Sparrow Distributed Low-Latency Spark Scheduling Kay Ousterhout, Patrick Wendell, Matei Zaharia, Ion Stoica Outline The Spark scheduling bottleneck Sparrow s fully distributed, fault-tolerant technique
More informationMesos: A Pla+orm for Fine- Grained Resource Sharing in the Data Center
Mesos: A Pla+orm for Fine- Grained Resource Sharing in the Data Center Ion Stoica, UC Berkeley Joint work with: A. Ghodsi, B. Hindman, A. Joseph, R. Katz, A. Konwinski, S. Shenker, and M. Zaharia Challenge
More informationMemshare: a Dynamic Multi-tenant Key-value Cache
Memshare: a Dynamic Multi-tenant Key-value Cache ASAF CIDON*, DANIEL RUSHTON, STEPHEN M. RUMBLE, RYAN STUTSMAN *STANFORD UNIVERSITY, UNIVERSITY OF UTAH, GOOGLE INC. 1 Cache is 100X Faster Than Database
More informationLRC: Dependency-Aware Cache Management for Data Analytics Clusters. Yinghao Yu, Wei Wang, Jun Zhang, and Khaled B. Letaief IEEE INFOCOM 2017
LRC: Dependency-Aware Cache Management for Data Analytics Clusters Yinghao Yu, Wei Wang, Jun Zhang, and Khaled B. Letaief IEEE INFOCOM 2017 Outline Cache Management for Data Analytics Clusters Inefficiency
More informationCSE 451: Operating Systems. Section 10 Project 3 wrap-up, final exam review
CSE 451: Operating Systems Section 10 Project 3 wrap-up, final exam review Final exam review Goal of this section: key concepts you should understand Not just a summary of lectures Slides coverage and
More informationFair Queueing. Presented by Brighten Godfrey. Slides thanks to Ion Stoica (UC Berkeley) with slight adaptation by Brighten Godfrey
Fair Queueing Presented by Brighten Godfrey Slides thanks to Ion Stoica (UC Berkeley) with slight adaptation by Brighten Godfrey Traditional queueing Traditional Internet - Congestion control mechanisms
More informationSecond Midterm Exam March 21, 2017 CS162 Operating Systems
University of California, Berkeley College of Engineering Computer Science Division EECS Spring 2017 Ion Stoica Second Midterm Exam March 21, 2017 CS162 Operating Systems Your Name: SID AND 162 Login:
More informationReal-Time Internet of Things
Real-Time Internet of Things Chenyang Lu Cyber-Physical Systems Laboratory h7p://www.cse.wustl.edu/~lu/ Internet of Things Ø Convergence of q Miniaturized devices: integrate processor, sensors and radios.
More informationIX: A Protected Dataplane Operating System for High Throughput and Low Latency
IX: A Protected Dataplane Operating System for High Throughput and Low Latency Belay, A. et al. Proc. of the 11th USENIX Symp. on OSDI, pp. 49-65, 2014. Reviewed by Chun-Yu and Xinghao Li Summary In this
More informationOperating Systems. Scheduling
Operating Systems Scheduling Process States Blocking operation Running Exit Terminated (initiate I/O, down on semaphore, etc.) Waiting Preempted Picked by scheduler Event arrived (I/O complete, semaphore
More informationBalancing Fairness and Efficiency in Tiered Storage Systems with Bottleneck-Aware Allocation
Balancing Fairness and Efficiency in Tiered Storage Systems with Bottleneck-Aware Allocation Hui Wang, Peter Varman Rice University FAST 14, Feb 2014 Tiered Storage Tiered storage: HDs and SSDs q Advantages:
More informationXCP: explicit Control Protocol
XCP: explicit Control Protocol Dina Katabi MIT Lab for Computer Science dk@mit.edu www.ana.lcs.mit.edu/dina Sharing the Internet Infrastructure Is fundamental Much research in Congestion Control, QoS,
More informationCS140 Operating Systems Midterm Review. Feb. 5 th, 2009 Derrick Isaacson
CS140 Operating Systems Midterm Review Feb. 5 th, 2009 Derrick Isaacson Midterm Quiz Tues. Feb. 10 th In class (4:15-5:30 Skilling) Open book, open notes (closed laptop) Bring printouts You won t have
More informationMultidimensional Scheduling (Polytope Scheduling Problem) Competitive Algorithms from Competitive Equilibria
Multidimensional Scheduling (Polytope Scheduling Problem) Competitive Algorithms from Competitive Equilibria Sungjin Im University of California, Merced (UC Merced) Janardhan Kulkarni (MSR) Kamesh Munagala
More informationAlgorithms, Games, and Networks March 28, Lecture 18
Algorithms, Games, and Networks March 28, 2013 Lecturer: Ariel Procaccia Lecture 18 Scribe: Hanzhang Hu 1 Strategyproof Cake Cutting All cake cutting algorithms we have discussed in previous lectures are
More informationAbstract. Testing Parameters. Introduction. Hardware Platform. Native System
Abstract In this paper, we address the latency issue in RT- XEN virtual machines that are available in Xen 4.5. Despite the advantages of applying virtualization to systems, the default credit scheduler
More informationA Platform for Fine-Grained Resource Sharing in the Data Center
Mesos A Platform for Fine-Grained Resource Sharing in the Data Center Benjamin Hindman, Andy Konwinski, Matei Zaharia, Ali Ghodsi, Anthony Joseph, Randy Katz, Scott Shenker, Ion Stoica University of California,
More informationCS Advanced Operating Systems Structures and Implementation Lecture 18. Dominant Resource Fairness Two-Level Scheduling.
Goals for Today CS194-24 Advanced Operating Systems Structures and Implementation Lecture 18 Dominant Resource Fairness Two-Level Scheduling April 7 th, 2014 Prof. John Kubiatowicz http://inst.eecs.berkeley.edu/~cs194-24
More informationBeyond Beyond Dominant Resource Fairness : Indivisible Resource Allocation In Clusters
Beyond Beyond Dominant Resource Fairness : Indivisible Resource Allocation In Clusters Abstract Christos-Alexandros Psomas alexpsomi@gmail.com Jarett Schwartz jarett@cs.berkeley.edu Resource allocation
More informationUNIFY DATA AT MEMORY SPEED. Haoyuan (HY) Li, Alluxio Inc. VAULT Conference 2017
UNIFY DATA AT MEMORY SPEED Haoyuan (HY) Li, CEO @ Alluxio Inc. VAULT Conference 2017 March 2017 HISTORY Started at UC Berkeley AMPLab In Summer 2012 Originally named as Tachyon Rebranded to Alluxio in
More informationMulti-Resource Fair Queueing for Packet Processing
Multi-Resource Fair Queueing for Packet Processing Ali Ghodsi Vyas Sekar Matei Zaharia Ion Stoica Electrical Engineering and Computer Sciences University of California at Berkeley Technical Report No.
More informationCongestion Control In the Network
Congestion Control In the Network Brighten Godfrey cs598pbg September 9 2010 Slides courtesy Ion Stoica with adaptation by Brighten Today Fair queueing XCP Announcements Problem: no isolation between flows
More informationHierarchical Scheduling for Diverse Datacenter Workloads
Hierarchical Scheduling for Diverse Datacenter Workloads Arka A. Bhattacharya 1, David Culler 1, Eric Friedman 2, Ali Ghodsi 1, Scott Shenker 1, and Ion Stoica 1 1 University of California, Berkeley 2
More informationReview. Preview. Three Level Scheduler. Scheduler. Process behavior. Effective CPU Scheduler is essential. Process Scheduling
Review Preview Mutual Exclusion Solutions with Busy Waiting Test and Set Lock Priority Inversion problem with busy waiting Mutual Exclusion with Sleep and Wakeup The Producer-Consumer Problem Race Condition
More informationOperating Systems. Lecture Process Scheduling. Golestan University. Hossein Momeni
Operating Systems Lecture 2.2 - Process Scheduling Golestan University Hossein Momeni momeni@iust.ac.ir Scheduling What is scheduling? Goals Mechanisms Scheduling on batch systems Scheduling on interactive
More informationCS Computer Systems. Lecture 6: Process Scheduling
CS 5600 Computer Systems Lecture 6: Process Scheduling Scheduling Basics Simple Schedulers Priority Schedulers Fair Share Schedulers Multi-CPU Scheduling Case Study: The Linux Kernel 2 Setting the Stage
More informationScheduling, part 2. Don Porter CSE 506
Scheduling, part 2 Don Porter CSE 506 Logical Diagram Binary Memory Formats Allocators Threads Today s Lecture Switching System to CPU Calls RCU scheduling File System Networking Sync User Kernel Memory
More informationCS268: Beyond TCP Congestion Control
TCP Problems CS68: Beyond TCP Congestion Control Ion Stoica February 9, 004 When TCP congestion control was originally designed in 1988: - Key applications: FTP, E-mail - Maximum link bandwidth: 10Mb/s
More informationScheduling II. Today. Next Time. ! Proportional-share scheduling! Multilevel-feedback queue! Multiprocessor scheduling. !
Scheduling II Today! Proportional-share scheduling! Multilevel-feedback queue! Multiprocessor scheduling Next Time! Memory management Scheduling with multiple goals! What if you want both good turnaround
More informationFirst-In-First-Out (FIFO) Algorithm
First-In-First-Out (FIFO) Algorithm Reference string: 7,0,1,2,0,3,0,4,2,3,0,3,0,3,2,1,2,0,1,7,0,1 3 frames (3 pages can be in memory at a time per process) 15 page faults Can vary by reference string:
More informationCOMP/ELEC 429/556 Introduction to Computer Networks
COMP/ELEC 429/556 Introduction to Computer Networks Weighted Fair Queuing Some slides used with permissions from Edward W. Knightly, T. S. Eugene Ng, Ion Stoica, Hui Zhang T. S. Eugene Ng eugeneng at cs.rice.edu
More informationSOFTWARE-DEFINED MEMORY HIERARCHIES: SCALABILITY AND QOS IN THOUSAND-CORE SYSTEMS
SOFTWARE-DEFINED MEMORY HIERARCHIES: SCALABILITY AND QOS IN THOUSAND-CORE SYSTEMS DANIEL SANCHEZ MIT CSAIL IAP MEETING MAY 21, 2013 Research Agenda Lack of technology progress Moore s Law still alive Power
More informationExample: CPU-bound process that would run for 100 quanta continuously 1, 2, 4, 8, 16, 32, 64 (only 37 required for last run) Needs only 7 swaps
Interactive Scheduling Algorithms Continued o Priority Scheduling Introduction Round-robin assumes all processes are equal often not the case Assign a priority to each process, and always choose the process
More informationLocality of Reference
Locality of Reference 1 In view of the previous discussion of secondary storage, it makes sense to design programs so that data is read from and written to disk in relatively large chunks but there is
More informationEnlightening the I/O Path: A Holistic Approach for Application Performance
Enlightening the I/O Path: A Holistic Approach for Application Performance Sangwook Kim 13, Hwanju Kim 2, Joonwon Lee 3, and Jinkyu Jeong 3 Apposha 1 Dell EMC 2 Sungkyunkwan University 3 Data-Intensive
More information!! What is virtual memory and when is it useful? !! What is demand paging? !! When should pages in memory be replaced?
Chapter 10: Virtual Memory Questions? CSCI [4 6] 730 Operating Systems Virtual Memory!! What is virtual memory and when is it useful?!! What is demand paging?!! When should pages in memory be replaced?!!
More informationEmpirical Evaluation of Latency-Sensitive Application Performance in the Cloud
Empirical Evaluation of Latency-Sensitive Application Performance in the Cloud Sean Barker and Prashant Shenoy University of Massachusetts Amherst Department of Computer Science Cloud Computing! Cloud
More informationBalancing DRAM Locality and Parallelism in Shared Memory CMP Systems
Balancing DRAM Locality and Parallelism in Shared Memory CMP Systems Min Kyu Jeong, Doe Hyun Yoon^, Dam Sunwoo*, Michael Sullivan, Ikhwan Lee, and Mattan Erez The University of Texas at Austin Hewlett-Packard
More informationSpring 2016 :: CSE 502 Computer Architecture. Caches. Nima Honarmand
Caches Nima Honarmand Motivation 10000 Performance 1000 100 10 Processor Memory 1 1985 1990 1995 2000 2005 2010 Want memory to appear: As fast as CPU As large as required by all of the running applications
More informationEpisode 5. Scheduling and Traffic Management
Episode 5. Scheduling and Traffic Management Part 2 Baochun Li Department of Electrical and Computer Engineering University of Toronto Keshav Chapter 9.1, 9.2, 9.3, 9.4, 9.5.1, 13.3.4 ECE 1771: Quality
More informationPage 1. Multilevel Memories (Improving performance using a little cash )
Page 1 Multilevel Memories (Improving performance using a little cash ) 1 Page 2 CPU-Memory Bottleneck CPU Memory Performance of high-speed computers is usually limited by memory bandwidth & latency Latency
More informationDatacenter Simulation Methodologies Case Studies
This work is supported by NSF grants CCF-1149252, CCF-1337215, and STARnet, a Semiconductor Research Corporation Program, sponsored by MARCO and DARPA. Datacenter Simulation Methodologies Case Studies
More informationMicroFuge: A Middleware Approach to Providing Performance Isolation in Cloud Storage Systems
1 MicroFuge: A Middleware Approach to Providing Performance Isolation in Cloud Storage Systems Akshay Singh, Xu Cui, Benjamin Cassell, Bernard Wong and Khuzaima Daudjee July 3, 2014 2 Storage Resources
More informationPage Replacement Algorithms
Page Replacement Algorithms MIN, OPT (optimal) RANDOM evict random page FIFO (first-in, first-out) give every page equal residency LRU (least-recently used) MRU (most-recently used) 1 9.1 Silberschatz,
More informationBeyond Beyond Dominant Resource Fairness : Indivisible Resource Allocation In Clusters
Beyond Beyond Dominant Resource Fairness : Indivisible Resource Allocation In Clusters Abstract Christos-Alexandros Psomas alexpsomi@gmail.com Jarett Schwartz jarett@cs.berkeley.edu Resource Allocation
More informationMesos: A Platform for Fine-Grained Resource Sharing in the Data Center
Mesos: A Platform for Fine-Grained Resource Sharing in the Data Center Benjamin Hindman Andrew Konwinski Matei Zaharia Ali Ghodsi Anthony D. Joseph Randy H. Katz Scott Shenker Ion Stoica Electrical Engineering
More informationVarys. Efficient Coflow Scheduling. Mosharaf Chowdhury, Yuan Zhong, Ion Stoica. UC Berkeley
Varys Efficient Coflow Scheduling Mosharaf Chowdhury, Yuan Zhong, Ion Stoica UC Berkeley Communication is Crucial Performance Facebook analytics jobs spend 33% of their runtime in communication 1 As in-memory
More informationChoosy: Max-Min Fair Sharing for Datacenter Jobs with Constraints
Choosy: Max-Min Fair Sharing for Datacenter Jobs with Constraints Ali Ghodsi Matei Zaharia Scott Shenker Ion Stoica UC Berkeley {alig,matei,shenker,istoica}@cs.berkeley.edu Abstract Max-Min Fairness is
More informationEpisode 5. Scheduling and Traffic Management
Episode 5. Scheduling and Traffic Management Part 2 Baochun Li Department of Electrical and Computer Engineering University of Toronto Outline What is scheduling? Why do we need it? Requirements of a scheduling
More informationFair Multi-Resource Allocation with External Resource for Mobile Edge Computing
Fair Multi-Resource Allocation with External Resource for Mobile Edge Computing Erfan Meskar and Ben Liang Department of Electrical and Computer Engineering, University of Toronto emeskar, liang@ece.utoronto.ca
More informationRxNetty vs Tomcat Performance Results
RxNetty vs Tomcat Performance Results Brendan Gregg; Performance and Reliability Engineering Nitesh Kant, Ben Christensen; Edge Engineering updated: Apr 2015 Results based on The Hello Netflix benchmark
More informationTales of the Tail Hardware, OS, and Application-level Sources of Tail Latency
Tales of the Tail Hardware, OS, and Application-level Sources of Tail Latency Jialin Li, Naveen Kr. Sharma, Dan R. K. Ports and Steven D. Gribble February 2, 2015 1 Introduction What is Tail Latency? What
More informationOASIS: Self-tuning Storage for Applications
OASIS: Self-tuning Storage for Applications Kostas Magoutis, Prasenjit Sarkar, Gauri Shah 14 th NASA Goddard- 23 rd IEEE Mass Storage Systems Technologies, College Park, MD, May 17, 2006 Outline Motivation
More informationCSE502: Computer Architecture CSE 502: Computer Architecture
CSE 502: Computer Architecture Memory Hierarchy & Caches Motivation 10000 Performance 1000 100 10 Processor Memory 1 1985 1990 1995 2000 2005 2010 Want memory to appear: As fast as CPU As large as required
More informationPricing Intra-Datacenter Networks with
Pricing Intra-Datacenter Networks with Over-Committed Bandwidth Guarantee Jian Guo 1, Fangming Liu 1, Tao Wang 1, and John C.S. Lui 2 1 Cloud Datacenter & Green Computing/Communications Research Group
More informationDiscretized Streams. An Efficient and Fault-Tolerant Model for Stream Processing on Large Clusters
Discretized Streams An Efficient and Fault-Tolerant Model for Stream Processing on Large Clusters Matei Zaharia, Tathagata Das, Haoyuan Li, Scott Shenker, Ion Stoica UC BERKELEY Motivation Many important
More informationManaging Storage: Above the Hardware
Managing Storage: Above the Hardware 1 Where we are Last time: hardware HDDs and SSDs Today: how the DBMS uses the hardware to provide fast access to data 2 How DBMS manages storage "Bottom" two layers
More informationVM Migration, Containers (Lecture 12, cs262a)
VM Migration, Containers (Lecture 12, cs262a) Ali Ghodsi and Ion Stoica, UC Berkeley February 28, 2018 (Based in part on http://web.eecs.umich.edu/~mosharaf/slides/eecs582/w16/021516-junchenglivemigration.pptx)
More informationCOMPUTER ARCHITECTURE. Virtualization and Memory Hierarchy
COMPUTER ARCHITECTURE Virtualization and Memory Hierarchy 2 Contents Virtual memory. Policies and strategies. Page tables. Virtual machines. Requirements of virtual machines and ISA support. Virtual machines:
More informationEpisode 5. Scheduling and Traffic Management
Episode 5. Scheduling and Traffic Management Part 3 Baochun Li Department of Electrical and Computer Engineering University of Toronto Outline What is scheduling? Why do we need it? Requirements of a scheduling
More informationCongestion Control for High Bandwidth-delay Product Networks. Dina Katabi, Mark Handley, Charlie Rohrs
Congestion Control for High Bandwidth-delay Product Networks Dina Katabi, Mark Handley, Charlie Rohrs Outline Introduction What s wrong with TCP? Idea of Efficiency vs. Fairness XCP, what is it? Is it
More informationSE Memory Consumption
Page 1 of 5 SE Memory Consumption view online Calculating the utilization of memory within a Service Engine is useful to estimate the number of concurrent connections or the amount of memory that may be
More informationVirtual Machines Disco and Xen (Lecture 10, cs262a) Ion Stoica & Ali Ghodsi UC Berkeley February 26, 2018
Virtual Machines Disco and Xen (Lecture 10, cs262a) Ion Stoica & Ali Ghodsi UC Berkeley February 26, 2018 Today s Papers Disco: Running Commodity Operating Systems on Scalable Multiprocessors, Edouard
More informationdavidklee.net gplus.to/kleegeek linked.com/a/davidaklee
@kleegeek davidklee.net gplus.to/kleegeek linked.com/a/davidaklee Specialties / Focus Areas / Passions: Performance Tuning & Troubleshooting Virtualization Cloud Enablement Infrastructure Architecture
More informationRe- op&mizing Data Parallel Compu&ng
Re- op&mizing Data Parallel Compu&ng Sameer Agarwal Srikanth Kandula, Nicolas Bruno, Ming- Chuan Wu, Ion Stoica, Jingren Zhou UC Berkeley A Data Parallel Job can be a collec/on of maps, A Data Parallel
More informationQuality-Assured Cloud Bandwidth Auto-Scaling for Video-on-Demand Applications
Quality-Assured Cloud Bandwidth Auto-Scaling for Video-on-Demand Applications Di Niu, Hong Xu, Baochun Li University of Toronto Shuqiao Zhao UUSee, Inc., Beijing, China 1 Applications in the Cloud WWW
More informationEECS750: Advanced Operating Systems. 2/24/2014 Heechul Yun
EECS750: Advanced Operating Systems 2/24/2014 Heechul Yun 1 Administrative Project Feedback of your proposal will be sent by Wednesday Midterm report due on Apr. 2 3 pages: include intro, related work,
More informationCS 425 / ECE 428 Distributed Systems Fall 2015
CS 425 / ECE 428 Distributed Systems Fall 2015 Indranil Gupta (Indy) Measurement Studies Lecture 23 Nov 10, 2015 Reading: See links on website All Slides IG 1 Motivation We design algorithms, implement
More informationMace, J. et al., "2DFQ: Two-Dimensional Fair Queuing for Multi-Tenant Cloud Services," Proc. of ACM SIGCOMM '16, 46(4): , Aug
Mace, J. et al., "2DFQ: Two-Dimensional Fair Queuing for Multi-Tenant Cloud Services," Proc. of ACM SIGCOMM '16, 46(4):144-159, Aug. 2016. Introduction Server has limited capacity Requests by clients are
More informationCIS 601 Graduate Seminar. Dr. Sunnie S. Chung Dhruv Patel ( ) Kalpesh Sharma ( )
Guide: CIS 601 Graduate Seminar Presented By: Dr. Sunnie S. Chung Dhruv Patel (2652790) Kalpesh Sharma (2660576) Introduction Background Parallel Data Warehouse (PDW) Hive MongoDB Client-side Shared SQL
More informationAnnouncements. Reading. Project #1 due in 1 week at 5:00 pm Scheduling Chapter 6 (6 th ed) or Chapter 5 (8 th ed) CMSC 412 S14 (lect 5)
Announcements Reading Project #1 due in 1 week at 5:00 pm Scheduling Chapter 6 (6 th ed) or Chapter 5 (8 th ed) 1 Relationship between Kernel mod and User Mode User Process Kernel System Calls User Process
More informationCSCI-GA Operating Systems Lecture 3: Processes and Threads -Part 2 Scheduling Hubertus Franke
CSCI-GA.2250-001 Operating Systems Lecture 3: Processes and Threads -Part 2 Scheduling Hubertus Franke frankeh@cs.nyu.edu Processes Vs Threads The unit of dispatching is referred to as a thread or lightweight
More information15-744: Computer Networking. Overview. Queuing Disciplines. TCP & Routers. L-6 TCP & Routers
TCP & Routers 15-744: Computer Networking RED XCP Assigned reading [FJ93] Random Early Detection Gateways for Congestion Avoidance [KHR02] Congestion Control for High Bandwidth-Delay Product Networks L-6
More informationAccelerating Pointer Chasing in 3D-Stacked Memory: Challenges, Mechanisms, Evaluation Kevin Hsieh
Accelerating Pointer Chasing in 3D-Stacked : Challenges, Mechanisms, Evaluation Kevin Hsieh Samira Khan, Nandita Vijaykumar, Kevin K. Chang, Amirali Boroumand, Saugata Ghose, Onur Mutlu Executive Summary
More informationCS644 Advanced Networks
What we know so far CS644 Advanced Networks Lecture 6 Beyond TCP Congestion Control Andreas Terzis TCP Congestion control based on AIMD window adjustment [Jac88] Saved Internet from congestion collapse
More informationAdvanced Computer Architecture
ECE 563 Advanced Computer Architecture Fall 2009 Lecture 3: Memory Hierarchy Review: Caches 563 L03.1 Fall 2010 Since 1980, CPU has outpaced DRAM... Four-issue 2GHz superscalar accessing 100ns DRAM could
More informationI, J A[I][J] / /4 8000/ I, J A(J, I) Chapter 5 Solutions S-3.
5 Solutions Chapter 5 Solutions S-3 5.1 5.1.1 4 5.1.2 I, J 5.1.3 A[I][J] 5.1.4 3596 8 800/4 2 8 8/4 8000/4 5.1.5 I, J 5.1.6 A(J, I) 5.2 5.2.1 Word Address Binary Address Tag Index Hit/Miss 5.2.2 3 0000
More informationTowards Deadline Guaranteed Cloud Storage Services Guoxin Liu, Haiying Shen, and Lei Yu
Towards Deadline Guaranteed Cloud Storage Services Guoxin Liu, Haiying Shen, and Lei Yu Presenter: Guoxin Liu Ph.D. Department of Electrical and Computer Engineering, Clemson University, Clemson, USA Computer
More informationScheduling in the Supermarket
Scheduling in the Supermarket Consider a line of people waiting in front of the checkout in the grocery store. In what order should the cashier process their purchases? Scheduling Criteria CPU utilization
More informationNext-Generation Cloud Platform
Next-Generation Cloud Platform Jangwoo Kim Jun 24, 2013 E-mail: jangwoo@postech.ac.kr High Performance Computing Lab Department of Computer Science & Engineering Pohang University of Science and Technology
More informationAnalysis and Implementation of Global Preemptive Fixed-Priority Scheduling with Dynamic Cache Allocation
Analysis and Implementation of Global Preemptive Fixed-Priority Scheduling with Dynamic Cache Allocation Meng Xu Linh Thi Xuan Phan Hyon-Young Choi Insup Lee Department of Computer and Information Science
More informationSRM-Buffer: An OS Buffer Management SRM-Buffer: An OS Buffer Management Technique toprevent Last Level Cache from Thrashing in Multicores
SRM-Buffer: An OS Buffer Management SRM-Buffer: An OS Buffer Management Technique toprevent Last Level Cache from Thrashing in Multicores Xiaoning Ding The Ohio State University dingxn@cse.ohiostate.edu
More informationMultilevel Memories. Joel Emer Computer Science and Artificial Intelligence Laboratory Massachusetts Institute of Technology
1 Multilevel Memories Computer Science and Artificial Intelligence Laboratory Massachusetts Institute of Technology Based on the material prepared by Krste Asanovic and Arvind CPU-Memory Bottleneck 6.823
More informationPreemptive, Low Latency Datacenter Scheduling via Lightweight Virtualization
Preemptive, Low Latency Datacenter Scheduling via Lightweight Virtualization Wei Chen, Jia Rao*, and Xiaobo Zhou University of Colorado, Colorado Springs * University of Texas at Arlington Data Center
More informationwhat is cloud computing?
what is cloud computing? (Private) Cloud Computing with Mesos at Twi9er Benjamin Hindman @benh scalable virtualized self-service utility managed elastic economic pay-as-you-go what is cloud computing?
More informationCS 268: Computer Networking
CS 268: Computer Networking L-6 Router Congestion Control TCP & Routers RED XCP Assigned reading [FJ93] Random Early Detection Gateways for Congestion Avoidance [KHR02] Congestion Control for High Bandwidth-Delay
More informationCS 326: Operating Systems. CPU Scheduling. Lecture 6
CS 326: Operating Systems CPU Scheduling Lecture 6 Today s Schedule Agenda? Context Switches and Interrupts Basic Scheduling Algorithms Scheduling with I/O Symmetric multiprocessing 2/7/18 CS 326: Operating
More informationCoflow. Recent Advances and What s Next? Mosharaf Chowdhury. University of Michigan
Coflow Recent Advances and What s Next? Mosharaf Chowdhury University of Michigan Rack-Scale Computing Datacenter-Scale Computing Geo-Distributed Computing Coflow Networking Open Source Apache Spark Open
More informationDB2 is a complex system, with a major impact upon your processing environment. There are substantial performance and instrumentation changes in
DB2 is a complex system, with a major impact upon your processing environment. There are substantial performance and instrumentation changes in versions 8 and 9. that must be used to measure, evaluate,
More informationA Comparative Study of High Performance Computing on the Cloud. Lots of authors, including Xin Yuan Presentation by: Carlos Sanchez
A Comparative Study of High Performance Computing on the Cloud Lots of authors, including Xin Yuan Presentation by: Carlos Sanchez What is The Cloud? The cloud is just a bunch of computers connected over
More informationWide area networks: packet switching and congestion
Wide area networks: packet switching and congestion Packet switching ATM and Frame Relay Congestion Circuit and Packet Switching Circuit switching designed for voice Resources dedicated to a particular
More informationRobinHood: Tail Latency-Aware Caching Dynamically Reallocating from Cache-Rich to Cache-Poor
RobinHood: Tail Latency-Aware Caching Dynamically Reallocating from -Rich to -Poor Daniel S. Berger (CMU) Joint work with: Benjamin Berg (CMU), Timothy Zhu (PennState), Siddhartha Sen (Microsoft Research),
More informationInteractive Scheduling
Interactive Scheduling 1 Two Level Scheduling Interactive systems commonly employ two-level scheduling CPU scheduler and Memory Scheduler Memory scheduler was covered in VM We will focus on CPU scheduling
More informationCongestion Control for High Bandwidth-delay Product Networks
Congestion Control for High Bandwidth-delay Product Networks Dina Katabi, Mark Handley, Charlie Rohrs Presented by Chi-Yao Hong Adapted from slides by Dina Katabi CS598pbg Sep. 10, 2009 Trends in the Future
More information