Académique Documents
Professionnel Documents
Culture Documents
cloud Keerthana Boloor , Rada Chirkova , Tiia Salo and Yannis Viniotis
Department
of Electrical and Computer Engineering Department of Computer Science North Carolina State University IBM Software Group Research Triangle Park
1 / 17
Agenda
Agenda
Problem description Dynamic request allocation and scheduling scheme Comparison with static allocation and FIFO/Weighted Round Robin scheduling scheme Conclusion
2 / 17
Problem description
Problem description
More web applications are designed to be context aware. Most context aware applications are built on SOA principles. Cloud computing systems - the most preferred platform for deployment. Service Level Agreements (SLA) - terms of service and pricing model. What is this presentation about?
Cloudcom 2010, Indianapolis, Indiana, USA
3 / 17
Problem description
Clients
4 / 17
Problem description
Gateway
Contextdata stores
DATA CENTER
5 / 17
Problem description
Model of an end-server
Class 1
Class 2
Class K
Each context aware application services multiple classes of users Each user class is guaranteed different quality of service based on economic considerations SLA species different service levels and service charges for the different user classes
Cloudcom 2010, Indianapolis, Indiana, USA
6 / 17
Problem description
Profit
100
Conformance(%)
X % - the fraction of requests of a particular user class which need to have a response time less than r seconds $P - The prot charged by the cloud, if the percentile of requests that have response time less than r seconds is greater than or equal to X %
Cloudcom 2010, Indianapolis, Indiana, USA
7 / 17
Problem description
Problem statement
Problem statement Allocate and schedule service requests locally at the end-servers so as to globally:
max
1j K
protj
(1)
where protj is the prot charged for conformance of the requests from users of class j .
8 / 17
Problem description
Problem statement
Problem statement Allocate and schedule service requests locally at the end-servers so as to globally:
max
1j K
protj
(1)
where protj is the prot charged for conformance of the requests from users of class j . This problem is NP-hard!!
8 / 17
Solution
Allocation phase
Allocation phase
Allocation phase
Allocation phase
Allocation phase
Adaptation phase
Adaptation phase
Adaptation phase
00:00
06:00
9 / 17
Solution
Allocation phase
Allocation phase
Allocation phase
Allocation phase
Allocation phase
Adaptation phase
Adaptation phase
Adaptation phase
06:00
Datacenters exchange conformance levels. Allocation phase Rank-based request allocation and gi-FIFO scheduling. Aim at increasing global prot.
Cloudcom 2010, Indianapolis, Indiana, USA
9 / 17
Solution
Prot-score calculation
Prot: pk
Profitscore assigned to each arriving request of class 1 ($)
Required global conformance: ck Current global conformance: cck If cck < ck Prot-score = pk /(ck cck ) Else Prot-score = 0
1500
1000
500
10 / 17
Solution
The arriving request is assigned a rank based on its prot-score and deadline. Does the arriving request meet its deadline? - Machine compatible!!!
11 / 17
Solution
gi-FIFO scheduling
Choose the request of user class with the highest current prot-score
Choose one with maximum waiting time but which results in a response time less than or equal to r
If no such request exists, choose the request with higher waiting time resulting in a response time greater than r
gi-FIFO has been proven to be the most suitable for percentile SLAs for a single server serving multiple classes.
12 / 17
Evaluation
Evaluation
Dynamic scheme vs static schemes
11000 10000 9000 8000 Profit incurred ($) 7000 6000 5000 4000 3000 2000 1000 0 5 10 15 20 25 30 Request rate 35 40 45 50 Dynamic rank based allocation with giFIFO scheduling Static allocation with WRR scheduling Static allocation with FIFO scheduling
13 / 17
Evaluation
11000 10000 9000 8000 Profit incurred ($) 7000 6000 5000 4000 3000 2000 1000 0 0 50 Request rate 100 150 Static allocation with giFIFO scheduling Dynamic rank based allocation with giFIFO scheduling
14 / 17
Evaluation
18000 16000 14000 12000 10000 8000 6000 4000 2000 0 Uniform distribution of classes, stringent SLA Uniform distribution of classes, relaxed SLA Nonuniform distribution of classes, stringent SLA Nonuniform distribution of classes, relaxed SLA
18000 16000 14000 12000 10000 8000 6000 4000 2000 0 Low contextdata load times High contextdata load times Medium Contextdata load times
Profit obtained($)
50
100
150
350
400
450
500
20
40
60
160
180
200
15 / 17
Conclusion
Conclusion
Identied the need for dynamic request scheduling and allocation for context aware applications in a distributed cloud.
Proposed a novel rank-based request allocation and gi-FIFO scheduling scheme for managing percentile SLAs with an aim to maximize prot obtained by the cloud.
16 / 17
Questions??
17 / 17