Using Global Behavior Modeling to improve QoS in Cloud Data Storage - - PowerPoint PPT Presentation

▶
using global behavior modeling to improve qos in cloud
SMART_READER_LITE
LIVE PREVIEW

Using Global Behavior Modeling to improve QoS in Cloud Data Storage - - PowerPoint PPT Presentation

Jul 08, 2023 144 likes •411 views

2 nd IEEE International Conference on Cloud Computing Technology and Science Using Global Behavior Modeling to improve QoS in Cloud Data Storage Services Jess Montes , Bogdan Nicolae, Gabriel Antoniu, Alberto Snchez, Mara S. Prez

slide-1
SLIDE 1

Using Global Behavior Modeling to improve QoS in Cloud Data Storage Services

Jesús Montes, Bogdan Nicolae, Gabriel Antoniu, Alberto Sánchez, María S. Pérez

December 2010 Indianapolis (IN), USA 2nd IEEE International Conference on Cloud Computing Technology and Science

slide-2
SLIDE 2

Jesus Montes - jmontes@cesvima.fi.upm.es 2

Introduction

 Cloud Computing

 Computation as utility

 New posibilities for the large public, such as complex

data processing

 Dryad  MapReduce

slide-3
SLIDE 3

Jesus Montes - jmontes@cesvima.fi.upm.es 3

MapReduce

 Platform for large-scale, massively parallel data

processing

 Application that can be deployed in the Cloud  Critical component: Underlying storage service

 Performance: High aggregated throughput under heavy

access concurrency

 Quality of Service: Stable throughput for each individual

access

slide-4
SLIDE 4

Jesus Montes - jmontes@cesvima.fi.upm.es 4

BlobSeer

 Data management for large, unstructured data

 Very large data (TB) – BLOBs: Binary large objects  Highly concurrent, fine-grain access (MB): R/W/A

 Key desing choices:

 Decentralized data and metadata management  Multiversioning exposed to the user  Lock-free concurrent writes (enabled by versioning)

slide-5
SLIDE 5

Jesus Montes - jmontes@cesvima.fi.upm.es 5

BlobSeer architecture

 Providers  Provider manager  Metadata providers  Version manager  Clients

slide-6
SLIDE 6

Jesus Montes - jmontes@cesvima.fi.upm.es 6

Storage QoS in MapReduce

 Resource failures  Complex data access patterns  Complexity of both hardware and software resources

slide-7
SLIDE 7

Jesus Montes - jmontes@cesvima.fi.upm.es 7

Proposal

 Automate the process of identifying and characterizing

events that have an impact on the storage service QoS

 In-depth monitoring  Application-side feedback  Behavior pattern analysis

slide-8
SLIDE 8

Jesus Montes - jmontes@cesvima.fi.upm.es 8

GloBeM

 Global Behavior Modeling

 A Specific methodology to analyze and construct a

model of the global behavior of a large-scale distribited system

 Behavior model characteristics:

 Finite state machine  State characterization based on monitoring parameters  Extended statistical information

slide-9
SLIDE 9

Jesus Montes - jmontes@cesvima.fi.upm.es 9

Behavior modeling cycle

slide-10
SLIDE 10

Jesus Montes - jmontes@cesvima.fi.upm.es 10

Proposal

1.Monitor the storage service 2.Identify behavior patterns 3.Classify behavior patterns according to feedback 4.Predict and prevent undesired behavior patterns

slide-11
SLIDE 11

Jesus Montes - jmontes@cesvima.fi.upm.es 11

Improving BloobSeer's QoS

slide-12
SLIDE 12

Jesus Montes - jmontes@cesvima.fi.upm.es 12

Experimental setup

 Scenario: MapReduce data gathering and analysis

 Read+Write access pattern (10:1 ratio)  I/O+Computation time (1:7 ratio)

 Platform: Grid'5000 resoruces

 x86_64 CPUs, 2GB RAM, 1Gbit/s standard Ethernet  Two clusters (130 and 275 nodes)

 Failure injection framework

 Multi-state resource avaliability characterization (Rood and

Lewis, 2007)

slide-13
SLIDE 13

Jesus Montes - jmontes@cesvima.fi.upm.es 13

Experimental settings

 Setting A (Grid’5000 Lille cluster)

 130 nodes  Total of 11TB accessed (1.3TB written, rest read)  Shared resources for computation and storage

 Setting B (Grid’5000 Orsay cluster)

 275 nodes  Total of 17TB accessed (1.5TB written, rest read)  Separated resources for computation and storage

slide-14
SLIDE 14

Jesus Montes - jmontes@cesvima.fi.upm.es 14

Experimental procedure

1.Running the original BlobSeer instance 2.Performing global behavior modeling 3.Improving BlobSeer 4.Running the improved BlobSeer instance

slide-15
SLIDE 15

Jesus Montes - jmontes@cesvima.fi.upm.es 15

Setting A

 GloBeM identified 4

states

State Average read BW (MB/s) State 1 24.2 State 2 20.1 State 3 31.5 State 4 23.9

slide-16
SLIDE 16

Jesus Montes - jmontes@cesvima.fi.upm.es 16

Setting A

parameter State 1 State 2 State 3 State 4

  • Avg. Read ops.

68.9 121.2 60.0 98.7 Read ops stdev. 10.5 15.8 9.9 16.7

  • Avg. Write ops.

43.2 38.4 45.3 38.5 Write ops stdev. 4.9 4.7 5.2 7.4 Free space stdev. 3.1e7 82.1e7 84.6e7 89.4e7

  • Nr. of providers

107.0 102.7 96.4 97.2

slide-17
SLIDE 17

Jesus Montes - jmontes@cesvima.fi.upm.es 17

Setting B

 GloBeM identified 3

states

State Average read BW (MB/s) State 1 50.7 State 2 35.0 State 3 47.0

slide-18
SLIDE 18

Jesus Montes - jmontes@cesvima.fi.upm.es 18

Setting B

parameter State 1 State 2 State 3

  • Avg. Read ops.

98.6 202.3 125.5 Read ops stdev. 17.7 21.6 21.9

  • Avg. Write ops.

35.2 27.5 33.1 Write ops stdev. 4.5 3.9 4.5 Free space stdev. 17.2e6 13.0e6 15.5e6

  • Nr. of providers

129.2 126.2 122.0

slide-19
SLIDE 19

Jesus Montes - jmontes@cesvima.fi.upm.es 19

Analysis

 When a node is dispatching many read operations, a

node failure causes many read faults

 Read faults force the client to look for another

available replica, reducing the final effective read BW

slide-20
SLIDE 20

Jesus Montes - jmontes@cesvima.fi.upm.es 20

Optimizing BlobSeer

 When, in State 2, selecting a node for writing, reduce

priority of those with excessive read operations

 Reduces the pressure on overwhelmed nodes  Improves load-balance and stabilizes read bandwidth,

increasing QoS

slide-21
SLIDE 21

Jesus Montes - jmontes@cesvima.fi.upm.es 21

Results: Setting A

Initial configuration read BW avg = 24.9 MB/s, stdev = 9.6

Advanced strategy read BW avg = 27.5 MB/s, stdev = 7.3

slide-22
SLIDE 22

Jesus Montes - jmontes@cesvima.fi.upm.es 22

Results: Setting B

Initial configuration read BW avg = 44.7 MB/s, stdev = 10.5

Advanced strategy read BW avg = 44.7 MB/s, stdev = 8.4

slide-23
SLIDE 23

Jesus Montes - jmontes@cesvima.fi.upm.es 23

Conclusions

 Cloud storage backend has to ensure stable

throughput: QoS constraint

 Our proposal combines component monitoring,

application side feedback and global behavior modeling to infer useful knowledge about the storage service

 We have obtained significative improvement in data

access QoS

slide-24
SLIDE 24

Jesus Montes - jmontes@cesvima.fi.upm.es 24

Thank You!

Using Global Behavior Modeling to improve QoS in Cloud Data Storage Services

Jesús Montes, Bogdan Nicolae, Gabriel Antoniu, Alberto Sánchez, María S. Pérez

2nd IEEE International Conference on Cloud Computing Technology and Science

slide-25
SLIDE 25

Jesus Montes - jmontes@cesvima.fi.upm.es 25

Outline

 Introduction  Proposal  Experimental Setup and Results  Conclusions

slide-26
SLIDE 26

Jesus Montes - jmontes@cesvima.fi.upm.es 26

BlobSeer

 A back-end for high-level, sophisticated data

management systems

 Higly scalable distributed file systems  Storage for cloud services  Extremely large distributed databases