2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Emerging trends in copy management
Sunil Kumar EMC Software and services
India Pvt. Ltd
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Agenda
Introduction to copy data Copy management system Emerging use cases
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data
What is copy data ? Copies of primary/production data which represents the data stored on disk at any given point in time. Benefits of copy data: No impact to production environment (host and applications ) as it is
typically created off the host by storage controller. Lower RPO by creating frequent copies (any point in time , hourly
basis) Quick recovery in case of data loss/corruption results in lower RTO
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Types of copy data
-Snapshot -Full copy -CDP(Continuous data protection)
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
copy data - Snapshot Snapshot: Space efficient point in time copy data created by using a bunch of pointers (copying only meta data) to the original data stored on disk in a storage device (LUN / File system)
Typically starts with almost zero size as no physical copy of original data is performed when snapshots are created
Size grows as original data
changes on production LUN/Volume/FS
X Y Z
Original LUN
Snapshot Data
. . .
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
copy data –Full copy
Full copy: Point in time copy data which is created by performing physical copy of all the blocks of original data stored on disk in a storage device (LUN / File system) to another storage device of same size.
May not be available instantly like snapshot (depends on implementation)
Highly available as all blocks of
production data is copied on separate storage device.
Not as space efficient as snapshot as
entire data needs to be copied
X Y Z
Original LUN
Full copy Data
X Y Z
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
copy data - CDP
CDP: Point in time copy data which is created by capturing every single change made to the original data stored on disk in a storage device (LUN / File system) to a separate storage location
Journal every write individually, hence enabling "Any-Point-In-Time" snapshots
Offers near zero data loss(extremely
low RPO) Space consumption depends on
implementation(compression algorithm) used by particular CDP system.
X Y Z
Original LUN
X’ Y’ X’’
Journal volume
HOST
CDP system
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Feature comparison Snapshot Full copy CDP
Recover capability from data corruption error
Yes Yes Yes
Storage Space requirement Low Required only for changed data
High Same size as source
High Maintains local copies and tracks all writes
Requires Original data Yes All unmodified blocks points to original data
No Contains full copy of all blocks from source
No A local copy of original data is maintained.
RPO Offered LOW Space Overhead is low and Creation is much faster, so can be created frequently(mainly hourly)
HIGH Space overhead is high, so typically created on daily basis
Extremely LOW Since every writes are captured, it allows to roll back to any point in time
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data market
As per research done by analyst firm IDC More than 60% of all enterprise disk capacity worldwide is filled with
copy data By 2016, spending on storage for copy data will approach $50 billion
and copy data capacity will exceed 315 million terabytes By 2016, companies will spend 8 times as much on copy data by 2016
as they will on storage for big data and analytics In the next 12 months, [IT departments] expect increased use of data
copies for app development and testing, regulatory compliance, multi-user access and long-term archival
https://idc-community.com/groups/it_agenda/storageanddatamanagement/copydataopportunitygetsrecognizedmeanwhilethemarketcontinuestogrow
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Challenges with copy data Multiple silo systems generating copy data which leads
to Operational inefficiency Network performance Increased expenditure to procure and maintain copy data
storage Copy data primarily perceived as insurance against data loss Without central lifecycle management, many copies end up
sitting around for long periods of time just taking up space for no purpose.
Copy data Production data
Test/Dev
Analytics
BC/DR
backup
production
..
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy management system
Removes infrastructure silos by providing a single pane of glass for all the copy management
operation by eliminating multiple copy generations tools and scripts,
reducing redundant copy data storage cost
Brings in new use cases of copy data besides protection. Improves operational efficiency by automating several
tasks related to copy data use cases.
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy management system Use cases
Oracle
Exchange
SAP
.
Copy Management
System
Oracle clone copy
Restore to production
Oracle snap of clone
Test & Dev
Oracle snap of clone
Long term retention
Oracle snap of clone
Oracle snap of clone
Backup
Analytics
Refreshable copy
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy management system High level architecture
SQL Oracle Exchange
File systems
Volume Managers
Device driver
HBA
Disk Subsystem
Storage Controller
SQL Oracle Exchange
File systems
Volume Managers
Device driver
HBA
Disk Subsystem
Storage controller/ Copy Creator
module
Application/host module
Database & Business logic
Storage Controller
Copy management system
Host Host
Workflow Engine & Scheduler
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Typical copy management system workflow to create a copy
Prepare application and storage for copy data
creation
Put application in backup mode(freeze)
Create point-in-time snapshot of underlying
storage volume
Take application out of backup mode (unfreeze)
Catalog application and Storage specific
information
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data consistency
There are primarily three types of data consistency offered by any copy management system Inconsistent: copy contains what is on disks at the time of copy without bothering about interdependency among different files on multiple volumes/disk. Crash-Consistent: Interdependent files across different LUN/volumes guaranteed to be the same version(i.e. same point in time). Application-Consistent: Involves application in the copy creation process and creates a transaction consistent copy by flushing all of application memory and I/O operations so that the application copy is completely consistent.
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data consistency
Past: Primarily Crash consistent copy (with the use of storage controller provided consistency via consistency groups) without involving application Emerging: Mostly application aware copy by involving application and OS(VSS) backup component in the copy creation workflow to create an application consistent copy which creates transactional consistent copy data.
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data consistency – application consistency options
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data restore/recovery
Past: admin driven and complete application instance (database, volume group, File system etc.) restore as copy data is associated with LUN /File systems and restore from copy data roll backs entire LUN data. Emerging: It makes use of advance application and storage level interfaces to catalog complete application and storage layout and offer item level restore capability from the copy data created by
storage controller. e.g. you can restore individual mails boxes from exchange, Table from a database and individual virtual machine or virtual machine disk or files and directory from a virtual machine host.
Self service restore
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data restore/recovery – item level restore capability
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data retention
Past: Copy data was being primarily used a backup accelerator. For long term retention integration with prominent backup product via script or third party modules was required to store the copy data on long term retention media e.g. tape. Emerging: Advancement in copy data technologies (like increased number of snapshot per device, snapshot shipping and reduced cost of disk based storage) is enabling copy management system to retain, archive and store copy data to long term retention media like de-duplicated disk subsystem or cheaper cloud based storage. This is enabling copy management system itself to act like backup system eliminating the need for a separate backup product.
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data retention
Application data
Copy Management
System
Application copy
Long term Retention in cloud
Store copy data in Deduplicated storage
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Hypervisor aware
Emerging: Increasingly adopting smarter approach while creating copies
in virtualized infrastructure Becoming aware of hypervisor based platforms and
integrating with hypervisors management framework to offer advanced capabilities like: 1. Virtual machine consistent copy of hypervisor file system which
enables copy creation of 100’s of virtual machines in a very short time
2. In case a virtual machine underlying storage is moved from one storage system to another, copy management platform dynamically follows the virtual machine storage to create appropriate copy on the underlying storage system.
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Hypervisor aware
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
SLA driven approach
Past: Earlier copy management system were used just like an orchestration engine to assist in copy generation. SLA specific to applications RPO and RTO were defined and monitored by external application. Emerging: Copy management system are exposing SLA driven approach where an organization can define the SLA to create and manage application copies and its lifecycle. It lets application admin specify appropriate:
RPO for the application Copy retention period i.e. how long the copy needs to be kept Frequency of copy Long term retention/backup frequency
It is also providing interfaces to application admins to monitor whether the SLA are met or violated based on real time and all using a single pane of glass
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
SLA driven copy management system
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data – DevOps
Emerging: Copy management system have started to offer capabilities like: On demand multiple generation repurpose copies
from one gold copy to use for diff. use cases Refresh capability of gold copy to get the latest copy
data. Exposing interfaces (like REST)/API for integration
with other products Integration with Chef, Puppet and other DevOps
tools
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data – DevOps
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Disaster recovery
Emerging: Enhanced to create local as well as remote (DR) copies. Both local and remote copies are created in a single
operation resulting in a single freeze/thaw operation on application. This reduces the impact on production environment.
Getting integrated with proven DR solutions like VMware SRM and others
Vendors are building DR orchestration engine inside the copy management system itself to reduce dependency on third party DR products.
2015 Storage Developer Conference - India. © Insert Your Company Name. All Rights Reserved.
Copy data – Analytics
Emerging: Analytics is normally not performed on the
production environment. Most of the Big data analytics requires a Test/Dev
kind of environment involving multiple copies of data Copy management solutions optimizes the space
consumed per copies by using smarter approach of creating copies out of copies for multiple use cases and managing the life cycle of these copies
Many solutions are using smarter approach to recreate the production environment from Copy data by integrating with the native applications