"Cloudian’s new solution, Cloudian 2.4, maximizes data protection, availability and security, which are critical for enterprises looking to deploy their own private cloud or move their data to the public cloud. For multi-datacenter deployments, Cloudian now provides Dynamic Consistency Levels, offering unprecedented data protection and ensuring continuous, uninterrupted operations in the case of network or site failures. "
See more detail at Cloudian Press Release.
http://www.cloudian.com/news/press-releases/press-release-25.html
1. Dynamic Consistency
April 2013
Cloudian KK & Inc.
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
1
2. Purpose of this document
Cloudian developed “Dynamic Consistency” and released it in
Cloudian version 2.4.
“Dynamic Consistency” is a unique technology that dynamically
changes data consistency in a high-availability configuration
Provides the following benefits:
- Continue services in case of failures without IT intervention
- Maintain highest possible data consistency
This document covers the following topics:
What is data consistency in a distributed system?
What is the problem with static data consistency?
How does “Dynamic Consistency” solve the problem?
2
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
3. What is “Cloudian”?
Cloudian is packaged software product that provides;
Distributed object storage system
Amazon S3 compatible API
Unlimited scalability to hundreds of petabytes
Data replications in multi-data centers
Tunable data consistency
Cloudian is being used by large organizations to offer cloud storage
services/systems
3
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
4. How Replication works in the distributed system?
One object is replicated to Even in the case of a
multiple servers server failure, client can
get the object
Client Client
Put “object” Get “object”
Server Server Server Server Server Server
1 2 3 1 2 3
Replication
4
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
5. Cloudian “Data Consistency” levels are tunable
Write Consistency
Read Consistency
5
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
6. How “Data Consistency” works in the
distributed system?
After “all” objects are After “some” objects are
replicated, Client receives replicated, Client receives
acknowledgement acknowledgement
Client All example
Client QUORUM(R+W>N) example
(N) Replications=3 (N) Replications=3
(W) Write=3 (W) Write=2
(R) Read=any (R) Read=2
“ack” “ack”
Server Server Server Server Server Server
1 2 3 1 2 3
Client Client
Data is always consistent Data is consistent by
6
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
QUORUM
7. Problem: “All” is disruptive in case of failure &
slow for replication over slow links
“All” or “QUORUM” cannot be completed in the case of failure
“All” or “QUORUM” is slower for clients/applications
Assumptions:
• 3 Replications
• 2 in DC1 Client
• 1 in DC2
No “ack”
Server Server Server Server Server Server
1 2 3 4 5 6
Data Center 1 Data Center 2
7
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
8. Problem: “Local_QUORUM ” does not guarantee
“Data Consistent” in the whole system
“Local _Quorum” cannot provide “data consistency” in the whole
system
Assumptions: Client Client Client may get “object”
• 3 Replications before updated
• 2 in DC1
• 1 in DC2
“ack”
Server Server Server Server Server Server
1 2 3 4 5 6
Data Center 1 Data Center 2
Data is consistent only when object is read from DC1
8
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
9. “Cloudian” provides Dynamic & Flexible
Consistency
Synchronous Asynchronous
Replication Replication
Traditional
Approach
(static &
+ No Data Loss + Not dependent on
inflexible) - Failures if DR site availability of DR site
not reachable - Data Loss in case of
Disaster
Synchronous Each Local Quorum Asynchronous
Cloudian Replication Quorum Quorum Replication
Multi-DC
(static &
flexible)
Cloudian 2.4 Synchronous Each Local Quorum Asynchronous
Multi-DC Replication #1 Quorum Quorum Replication #3
(dynamic & #2
flexible)
9
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
10. Solution: “Dynamic Consistency” changes
consistency level dynamically (1/2)
1. System operates at “All” to keep “Data Consistency”
Assumptions: Client Client
• 3 Replications
• 2 in DC1
• 1 in DC2 Client always get
“ack” the same “object”
Server Server Server Server Server Server
1 2 3 4 5 6
Data Center 1 Data Center 2
Data is consistent in the whole system
10
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
11. Solution: “Dynamic Consistency” changes
consistency level dynamically (2/2)
2. System automatically falls back to “Local_QUORUM” in case
of data center failure
Assumptions:
• 3 Replications
• 2 in DC1 Client Client
• 1 in DC2
Client gets “object” from
“ack” surviving data center
Server Server Server Server Server Server
1 2 3 4 5 6
Data Center 1 Data Center 2
11
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.
12. For more information:
Contact by Email: info@cloudian.com
Homepage: www.cloudian.com
www.cloudian.jp
Cloudian Community Edition:
http://www.cloudian.com/get-started.html
30⽇間無料トライル:
http://www.cloudian.jp/cloud-storage-products/cloudian_eval_agr.php
12
(c) Copyright , Cloudian KK&Inc. 2013, All rights reserved.