The document describes BestPeer++, a peer-to-peer based large-scale data processing platform. BestPeer++ consists of a core and Amazon cloud adapter. The core includes a bootstrap peer that manages peer joins/departures and monitors health. Normal peers extract and load data from business systems into MySQL databases. Benchmarking showed BestPeer++ can efficiently handle workloads and scale query throughput linearly as peers increase.

1. BESTPEER++: A PEER-TO-PEER BASED
LARGE-SCALE DATA PROCESSING
PLATFORM
Submitted by,
PRABHUDEV R
4NI12IS416
1

2. AGENDA
1. Introduction.
2. Overview of the BestPeer++ system.
3. Bootstrap peer.
4. Normal peer.
5. Benchmarking.
6. Advantages.
7. Conclusion.
2

3. 1. INTRODUCTION
 Corporate network shares information with participating
companies of a common interest.
 Companies reduce their operational cost and increase
the revenue.
 Delivers elastic data sharing services.
 Provides economical , flexible and scalable platform.
 Based on pay as you go business model. 3

4. 2. OVERVIEW OF THE BESTPEER++
SYSTEM
BestPeer data management platform
 Adaptive join query processing
 distributed online aggregation techniques to provide efficient
query processing.
BestPeer++, cloud enabled evolution of BestPeer
 Distributed access control
 Multiple types of indexes
 Pay-as-you-go query processing for delivering elastic data
sharing services in the cloud.
 The software components of BestPeer++ are separated
into two parts:-
1. Core.
2. Adapter. 4

5. AMAZON CLOUD ADAPTER
 Elastic hardware infrastructure for BestPeer++ to
operate on by using Amazon Cloud services.
 Launching/terminating dedicated MySQL database
servers and monitoring/ backup/auto-scaling those
servers.
 Finally, the Amazon Cloud Adapter also provides
automatic fail-over service.
5

6. THE BESTPEER++ CORE
 Platform-independent
logic, including query
processing and P2P
overlay.
 Cloud adapter and
consists of two software
components:
1. Bootstrap peer.
2. Normal peer.
6

7. 3. BOOTSTRAP PEER
 The bootstrap peer is run by the BestPeer++ service
provider, and its main functionality is to manage the
BestPeer+ + network
1. Managing Normal Peer Join/Departure.
2. Auto Fail-Over and Auto-Scaling.
7

8. MANAGING NORMAL PEER JOIN/DEPARTURE
 Each normal peer intends to join an existing
corporate network must first connect to the
bootstrap peer.
 The joined peer will receive the corporate network
information including the current participants, global
schema, role definitions, and an issued certificate.
 When a normal peer needs to leave the network, it
also notifies the bootstrap peer first.
8

9. AUTO FAIL-OVER AND AUTO-SCALING
 In addition to managing peer join and peer
departure.
 The bootstrap peer spends most of its running-time
on monitoring the health of normal peers.
 Scheduling fail-over and auto-scaling events.
9

10. 4.NORMAL PEER
Offline data flow
The data are extracted
periodically by a data
loader from the business
production system to the
normal peer instance.
Online data flow
 The query processor
performs user queries
using a fetch and
process strategy.
10

11. SCHEMA MAPPING
 Defines the mapping between the local schema of
each production system and the global shared
schema.
 The mapping consists of metadata mappings and
value mappings and also support instance level
mapping.
11

12. DATA LOADER
 Extracts data from production systems to normal peer
instances according to the result of schema mapping.
 The data loader also creates a snapshot of the newly
inserted data.
 At interval times, re-extracts data from the production system
to create a new snapshot.
 This snapshot is then compared to the previously stored one
to detect data changes.
 Finally, the changes are used to update the MySQL database
hosted in the normal peer. 12

13. DATA INDEXER
 BATON
 The first range, R0, is the subdomain maintained by
the node.
 The second range, R1, is the domain of the sub tree
rooted at the node. 13

14. DISTRIBUTED ACCESS CONTROL
 The basic idea is to use roles as templates to
capture common data access privileges and allow
businesses to override these privileges to meet
their specific needs.
 The information of the users created at one peer is
forwarded to the bootstrap peer and then
broadcasted to other normal peers also.
 The local administrator at this peer can easily
define the role-based access control for any user.
14

15. PAY-AS-YOU-GO QUERY PROCESSING
 BestPeer++ provides two services for the
participants:
1. Storage service
2. Search service
 After data are exported from the local business
system into a BestPeer++ instance, we apply the
schema mapping rules to transform them into the
predefined formats.
15

16. 5.BENCHMARKING
 This section evaluates the performance and throughput
of BestPeer++ on Amazon cloud platform.
1. For the performance benchmark, they compare the
query latency of BestPeer++ with HadoopDB using
five queries selected from typical corporate network
applications workloads.
2. For the throughput benchmark, they create a simple
supply-chain network consisting of suppliers and
retailers and study the query throughput of the
system.
16

17. 6.ADVANTAGES OF BESTPEER++
1. Deliver near linear query throughput as the number of
normal peers grows.
2. BestPeer++ adopts the pay-as-you-go business model
popularized by cloud computing.
3. The role-based access control for the inherent distributed
environment of corporate networks.
4. P2P technology to retrieve data between business
partners.
5. Efficient data sharing within corporate networks. 17

18. 7.CONCLUSION
 The benchmark conducted on Amazon EC2 cloud
platform shows that our system can efficiently
handle typical workloads in a corporate network
and can deliver near linear query throughput as the
number of normal peers grows.
 Therefore, BestPeer++ is a promising solution for
efficient data sharing within corporate networks.
18

19. REFERENCES
1. S. Wu, Q.H. Vu, J. Li, and K.-L. Tan, “Adaptive Multi-Join
Query Processing in PDBMS,” Proc. IEEE Int’l Conf. Data
Eng. (ICDE ’09), pp. 1239-1242, 2009.
2. D. Bermbach and S. Tai, “Eventual Consistency: How Soon
is Eventual? An Evaluation of Amazon s3’s Consistency
Behavior,” in Proc. 6th Workshop Middleware Serv. Oriented
Comput. (MW4SOC ’11), pp. 1:1-1:6, NY, USA, 2011.
3. B. Cooper, A. Silberstein, E. Tam, R. Ramakrishnan, and R.
Sears, “Benchmarking Cloud Serving Systems with YCSB,”
Proc. First ACM Symp. Cloud Computing, pp. 143-154,
2010.
4. Oracle Inc., “Achieving the Cloud Computing Vision,” White
Paper, 2010. 19

20. 20