EMC DSSD Cubic RAID

EMC DSSD CUBIC RAID
Unprecedented flash reliability
with no additional overhead
DSSD D5 provides industryleading protection against data
failures in flash memory
CHALLENGE: FLASH RELIABILITY, OVERHEAD
Redundant array of independent disks (RAID) is a common method of distributing
data to improve performance and reliability in storage systems—from traditional disk
drives to high-performance flash. The layout of the data and its error correction
information distribution determines the RAID level for a system, which designates the
combination of reliability, availability, performance, overhead, and capacity.
RAID 6: LIMITED RELIABILITY AND CAPACITY
RAID 6 or double-parity RAID is the most common schema used in enterprise storage
for improving reliability. It consists of a stripe of devices in which two of the devices
are used for parity and can remedy data failures in up to two devices in the stripe.
Benefits
•
Provides superior protection over
RAID 6 or RAID DP for flash cell
and flash module failure
•
Requires only 17 percent of
storage dedicated to Cubic RAID
•
Combined with the other flash
resilience features of DSSD,
provides industry leading flash
reliability:
o
Flash physics controls for
enhanced flash die endurance
o
Space-time garbage
collection for lower write
amplification and improved
wear leveling
o
o
Defect avoidance to maintain
Figure 1: RAID 6 (blue) parity devices are protecting (white) data devices
Despite its popularity, RAID 6 has some significant reliability and capacity limitations:
•
RAID 6 can only recover from a maximum of two failures in a stripe of devices
•
The number of devices is generally limited (less than 20 devices in most cases)
due to controller limitations and performance and fault tolerance concerns
•
To achieve enterprise acceptable reliability, RAID 6 requires significant capacity to
be used as overhead for parity rather than for storage of data
MULTI-DIMENSIONAL RAID 6: STILL LIMITED
RELIABILITY
In multi-dimensional RAID 6, the total number of devices (capacity) in a system is
increased significantly by stacking rows of RAID 6 stripes on top of each other with
devices dedicated to parity at the end of each row and column. However, the
technologies for remedying faults in rows and columns remain separate, with any
defect history persistently for
intersection of three failures in a row and a column resulting in data loss. While this
flash module lifetime
method allows for more devices in a storage system, it does not achieve the
enterprise-level reliability required by today’s data and I/O-intensive workloads.
Vaulting for recovery of data
after power or other failure
o
FM resilvering to
automatically rebuild data on
replaced FMs with no
administrative intervention
o
Enterprise grade error
correction codes
DATA SHEET
Figure 2: Multi-dimensional RAID
6 cannot remedy three intersecting
failures in a row and a column.
OVERCOMING THE LIMITATIONS OF RAID 6
EMC® DSSD™ Cubic RAID™ features patented RAID technology that breaks through
the reliability barriers of traditional and multi-dimensional RAID 6. It provides
unprecedented enterprise-level flash memory reliability without any extra overhead.
INTELLIGENT PARITY
Cubic RAID’s system-wide view of data overcomes the limitations encountered by
simpler multi-dimensional Raid 6 algorithms. Cubic RAID protects data in many grids,
each made up of 18 rows and 32 columns, with each grid square representing a flash
die cell. Similar to multi-dimensional RAID 6, two cells at the end of each row and
column are dedicated to parity. What makes Cubic RAID different and results in much
greater reliability for underlying flash memory cells are two key innovations.
First, Cubic RAID dedicates the four intersection cells—where the row and column
parity intersect—to hold parity information for, and remedy failures in, the row and
column parity cells themselves. This feature provides an additional layer of protection.
Second, and more importantly, Cubic RAID cohesively interlinks the different RAID
dimensions. This means that where three intersecting failures in a row and a column
would cause other schemas to lose data, Cubic RAID can recover from those failures.
Cubic RAID accomplishes this by fixing rows or columns with two or fewer failures in
an alternating order to reduce the number of overall failed cells, ultimately enabling
Cubic RAID to recover more than two intersecting failures in a row and a column. In
fact, Cubic RAID can recover data even from large and complex failure patterns.
HOW DSSD CUBIC RAID FIXES COMPLEX DATA FAILURES
A grid with a large and complex data failure
pattern
Step 1: Cubic RAID fixes all columns with two or less
failures
Step 2: Cubic RAID fixes all rows with two or less
failures
Step 3: Now more columns are recoverable at two or
less failures
Step 4: Now more rows are recoverable at two or less
failures
Step 5: A final iteration on remaining columns with
two or less failures
All data failures have been remedied by Cubic RAID!
UNLOCK THE POTENTIAL OF FLASH STORAGE
FOR YOUR MISSION-CRITICAL WORKLOADS
EMC DSSD Cubic RAID:
•
Can recover from data failures that cannot be remedied utilizing traditional RAID
6 or multi-dimensional RAID 6
CONTACT US
To learn more about how EMC
products, services, and solutions can
help solve your business and IT
challenges, contact your local
representative or authorized reseller,
visit www.emc.com, or explore and
compare products in the EMC Store.
•
Requires only 17 percent overhead
•
Is always on and is not configurable
•
Provides a high level of protection at EMC DSSD D5™ high levels of performance
•
Provides unprecedented reliability required for today’s and tomorrow’s data and
I/O intensive workloads
EMC2, EMC, the EMC logo, Cubic RAID, D5, and DSSD are registered trademarks or trademarks of EMC
Corporation in the United States and other countries. © Copyright 2016 EMC Corporation. All rights
reserved. Published in the USA. Data Sheet H14869
EMC believes the information in this document is accurate as of its publication date. The information is
subject to change without notice.