Upload
vunhi
View
279
Download
2
Embed Size (px)
Citation preview
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 11
Data Compression and Deduplication
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 2
Data Redundancy Elimination Landscape
Array Filer
Hosts
(Inline and
Offline)
Network
(Inline)
Storage
(Inline and
Offline)
Data Domain
Netapp’s A-SIS
VMWARE DeDE
IBM DDE for Storage Tank
Solaris ZFS
Cisco MDS + WAAS
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 3
Cisco’s Focus Areas for
Redundancy Elimination
Data Center
Cloud Deployments
Data Center to Data Center Connectivity
Branch Office Connectivity
Disaster Recovery using long distance replication Data Center
Cloud
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 4
Challenges in Deduplication
CPU Intensive Chunking Operations
CPU Intensive Fingerprint Calculation such as SHA-1
Managing Large Indexes in Memory as well as in Persistent Storage
Reducing Disk I/O for Index Lookup
Load Balancing and Scalability in shared SAN environments
Inter Host communication
Host based deduplication for Shared File Systems
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 5
Cisco’s Network Centric Redundancy Elimination
Fully Integrated System for lowering WAN OPEX
Deduplication
Compression
TCP optimization
Encryption
Hardware Assistance for Compression
Scope for Hardware Assists in Deduplication
Coordination less approach for Shared SAN file systems
Application/Host/Array Agnostic
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 6
WAN Optimization
TCP Optimizations
IP WANMDS
Interconnecting Data Centers using MDS 9000
DWDM Optical MAN
SAN Extension
FCIP protocol for SAN extension
Secondary Data CenterPrimary Data Center
MDS
RedundnacyElimination
Hardware Compression
Security
Encryption
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 7
Cisco WAAS for Interconnecting Data Centers
Data Redundancy Elimination includes
Data Deduplication: application-agnostic technique eliminates redundant data from TCP streams providing up to 100:1 reduction
Persistent LZ Compression: session-based compression provides up to an additional 10:1 compression even for messages that have been optimized by deduplication
Dedupe Dedupe
LZ LZ
Synchronized
Compression
History
WAN
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 8
DC-DC Replication Using Cisco WAAS
WAN
Cisco
Catalyst
6500
Cisco
Catalyst
6500
Cisco
WAAS
Cisco
WAASGigE GigE
Storage
arrays
Storage
arrays
DATA CENTER (Primary) DATA CENTER (Backup)
• Ease of deployment
• Integration with existing network topologies
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 9
Storage Replication in WCCP Mode
WAN
Cisco
Catalyst
6500
Cisco
Catalyst
6500
Cisco
WAAS
GigEGigE
WCCPWCCP
Storage
arrays
Storage
arrays
Higher availability and scalability
through N+1 clustering
Cisco
WAAS
DATA CENTER (Primary) DATA CENTER (Backup)
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 10
Storage Replication using Cisco MDS and WAAS
WAN
Cisco
Catalyst
6500
Cisco
Catalyst
6500
Cisco
WAAS
Cisco
WAAS
GigE GigE
Storage
arraysStorage
arrays
Cisco MDS 9000
with IP Service
Modules
Cisco MDS 9000
with IP Service
Modules
FCIP
FCFC
FCIP
• Improve end-to-end network resilience (against port
failure)
• Optimize data replication for legacy FC storage
• Support multiple heterogeneous FC storage arrays
DATA CENTER (Primary) DATA CENTER (Backup)
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 11
Trends in Deduplication
Use of Solid State Disks to overcome Index Lookup and retrieval latency
Price Performance Trade off
Design Choices to overcome poor random write throughput of SSDs
Random write IOPS is still about 350
Adaptive deduplication that combines Inline and Offline methods
Data Agnostic Deduplication
Integrated approach for File and Block Deduplication
© 2010 Cisco Systems, Inc. All rights reserved.LOC 2010 12