Understanding RAID with Dell SC Series Storage Dell Engineering September 2016 A Dell Technical White Paper
Revisions Date Description February 2016 Initial release September 2016 Added changes for redundancy for SCOS 7, 7.1 THIS WHITE PAPER IS FOR INFORMATIONAL PURPOSES ONLY, AND MAY CONTAIN TYPOGRAPHICAL ERRORS AND TECHNICAL INACCURACIES. THE CONTENT IS PROVIDED AS IS, WITHOUT EXPRESS OR IMPLIED WARRANTIES OF ANY KIND. © 2016 Dell Inc. All rights reserved. Dell and the Dell EMC logo are trademarks of Dell Inc. in the United States and/or other jurisdictions.
Table of contents Revisions.............................................................................................................................................................................2 Executive summary.............................................................................................................................................................5 Acknowledgements ....................................................................................................................
7.1 SSD considerations ..........................................................................................................................................32 7.2 SSDs in Tier 1 dual redundancy .......................................................................................................................32 7.3 Techniques for specific workloads....................................................................................................................32 8 Summary ...............
Executive summary Dell™ SC Series storage is designed to take advantage of redundant components, including RAID, in order to achieve the greatest level of performance and availability. Core to the SC Series Fluid Data™ Architecture is the automatic placement of data across the available disks. The disks are logically grouped by type based on speed and capacity. RAID protection is built on a page architecture that provides data optimization for availability, capacity, and performance.
1 SC Series storage basics 1.1 Tiers SC Series arrays include redundant, hot-swappable components (for example, physical disks, control modules, fans, and power supplies) for a no-single-point-of-failure configuration. Along with redundant hardware, several different RAID levels are supported with each configuration optimized to maximize performance, availability, and capacity of the SC Series architecture.
The RAID redundancy options are: Non-redundant: Uses RAID 0 in all classes, in all tiers. Data is striped but provides no redundancy. If one disk fails, all data is lost. Do not use non-redundant storage for a volume unless the data has been backed up elsewhere. Note: The non-redundant option is not available for the SCv2000 series. 1.3 Single-redundant: Protects against the loss of any one drive.
1 Standard 2 Fast track RAID level are tracks not allocated to Fast Track. Track RAID level applies to HDDs with the Fast Track license; does not apply to SSDs. RAID 5 is not preferred for arrays with 967 GB or larger HDDs. RAID 6 and RAID 10-DM offer significantly higher levels of resiliency with very high capacity disks. 1.4 RAID levels and redundancy A storage type is a pool of storage with a single data page size and a specified redundancy level.
followed for RAID level redundancy are based on capacity and on whether the disks are added to an existing or new storage type. Currently, disks up to 1.9 TB will default to dual redundancy for HDDs and up to 3.9 TB for SSDs. Beyond those capacities, dual redundancy may be required. The requirements and recommendations (defaults) for dual redundancy are shown in Table 3.
2 RAID rebuild, availability, and efficiency comparison In general terms, rebuild rates, availability, and capacity efficiency are dependent on the RAID level. Figure 2 depicts the rebuild rate based on RAID levels for the different SC Series disk options. The results are from rebuilds with a light random, 64KB, 70-percent read workload simulation. 14,000 12,000 GB/day 10,000 8,000 6,000 4,000 2,000 0 RAID 5-9 RI SSD RAID 6-10 SAS 15K RAID 10 NLSAS 7.
RAID 6 and RAID 10-DM provide the greatest protection, while SSDs have the fastest rebuild rates. 15K HDDs rebuild faster than 10K or NLSAS. Larger disks of any type are subject to the rate of rebuild to avoid a dual disk failure. As the number of disks and capacities have increased, RAID 6 or RAID 10-DM have become increasingly important to ensure overall data protection in storage arrays. 2.
RAID levels have different attributes for performance, reliability and rebuilds and are described in Table 4. The first column lists the RAID level while the other columns indicate the best suited workload along with rebuild performance and relative protection. RAID level and redundancy comparison.
3 Spare disks Depending on the RAID level and the total number of disks in each SC Series storage array, one or more spare disks are automatically configured and used in the event of a disk failure. The use of spare disks is highly recommended as an additional level of protection should a disk failure happen. Spare disks will replace the failed disk and allow the RAID set to rebuild.
If a set is degraded: Data is reconstructed on the new disk and after reconstruction, and performance will return to normal.
4 SC Series RAID level use By default, an SC Series array groups all disks in a single disk folder. This disk folder contains all the various speeds and capacities of disks. Redundancy levels can be configured for each tier of disk. Tier 1 disks are the fastest disks in SC Series storage, while other disks are grouped according to their speed in the other two available tiers (Tier 2 and Tier 3). RAID is allocated within a tier across all of the disks.
For illustrative purposes, Figure 5 shows a possible RAID distribution in a twelve-disk enclosure with one global hot spare. Tier 1, which includes the Fast Track feature, has both RAID 10 and RAID 5-5 (RAID 5-9 is also an option). SC Series example of RAID 5-5 and RAID 10 in Tier 1 over 12 disks with one spare. The example in Figure 5 shows RAID distributed across all of the disks, and parity is staggered as much as possible while hot spots are minimized.
When the SC array or disk folder consists only of large disks such as 2 TB HDDs or 4 TB SSDs or larger, the tier is dual redundant. For these situations, RAID 10-DM is used for writes, while RAID 6 is updated by Data Progression. Consider the performance implications for large Tier 1 disk solutions. Application I/O writes only to RAID 10 Data Progression, the automated data movement by the SC Series array, frees up as much RAID 10 space as possible.
Figure 7 indicates the operational advantage during writes that RAID 10 has over RAID 5. In a single redundant tier, all application updates go to RAID 10 with only two I/Os, in contrast to the four I/Os needed for an application writing to RAID 5. Difference in write penalty between RAID 10 and RAID 5 For most SC Series arrays, Tier 1 has RAID 10. However, if the tier requires dual redundancy writes, they occur on RAID 10-DM.
5 5.1 Snapshots and Data Progression Over time, an SC Series array determines the appropriate movement of data based on the frequency of access. A snapshot is a point-in-time-copy (PITC) of a volume that provides fast recovery of data. Snapshots do not copy the data but simply freeze the data as read only. New writes to the volume allocate new space in the highest performing RAID level.
Snapshots prevent writes to data and then Data Progression moves the data to other RAID levels and tiers. Figure 8 is an example of Data Progression for volumes using the recommended storage profile in a typical two-tier environment. Data Progression in a two-tier configuration For volumes using the recommended storage profile in a three-tier configuration, Data Progression moves data to each tier as illustrated in Figure 9. Data progression in a three-tier configuration 5.
Figure 10 is an example of allocation of space by tier and RAID level. Storage types showing space utilization based on tier and RAID level The SC Series arrays provide tools to show where the data will be placed during the next Data Progression cycles. As an example, the Enterprise Manager pressure report shows the direction that data is expected to travel with the next Data Progression (see Figure 11).
Pressure report showing data movement as indicated by green up, orange down. Boxes added for illustration. Note: Since Tier 1 is all SSD, Fast Track (Fast) RAID levels are not allocated. This report shows the expected direction and quantity of data that will be moved during the next Data Progression cycle. In the table, the amount of data is indicated in the Moving Up or Moving Down columns. Data that moves down is first moved to the lower RAID, and then to the lower performing tier.
6 RAID management SC Series arrays are designed to manage RAID levels automatically, however the following operations may be performed if the need arises; 6.
Figure 12 shows Tier 1 storage allocated as single redundancy. A single redundant tier In the Modify tier Redundancy wizard, choices are available to change the redundancy to dual redundancy in each tier. The storage type is configured for all three tiers regardless of the presence of physical disks in those tiers. However, the redundancy may be selected based on tier as needed. With dual redundancy, the RAID levels change from RAID 10 to RAID 10-DM, and from RAID 5 to RAID 6 as shown in Figure 13.
Some occasions require the tier redundancy to be changed from dual redundancy to single redundancy. This may only be achieved for SSDs smaller than 4 TB or HDDs smaller than 2 TB. Factors such as free space, system activity, time of rebalance, and impact on availability should be considered before making this change. Figure 14 shows redundancy options available for each tier, and the change from a dual redundant to a single redundant storage type.
After the rebalance has completed, the storage type will change the RAID 10-DM and RAID 6 to RAID 10 and RAID 5, respectively. Figure 15 shows the results of changing from dual to single redundancy. Dual to single redundancy Note: Changing a dual redundant storage type to a single redundant storage type requires careful consideration and should be discussed with Dell Support prior to implementation. 6.3 Creating a storage type A storage type is a pool of storage with specified redundancy levels.
Figure 16 demonstrates a storage type created on a new disk folder with all Tier 1, 2, and 3 disks. For this storage type example, dual redundancy is chosen to meet a higher availability requirement. New storage type with dual redundant tiers The storage type could be selected by volumes to take advantage of the disks in the SQL-Logs disk folder. Note: Best practices recommend a single storage type per disk folder.
The results of creating a new storage type show that the new disks (in this example, the SQL-Logs disk folder) have both Tier 1 and Tier 3 disks with dual redundancy protection (RAID 10-DM and RAID 6). New disks are assigned to the SQL-Logs disk folder with a dual redundancy storage type The volume will have the choice to use a specific storage type during creation. However, once created, this volume will not be able to be modified to use a different storage type.
Figure 18 shows the new storage type selected for a new volume. New storage type selected during new volume creation.
6.4 Creating a storage profile A storage profile can further associate data to a particular tier as well as one or more RAID levels available in that tier. Occasionally, an application may need to override the automation built into the array and bind a volume to a particular RAID level and/or tier to accommodate performance, capacity or availability needs. Note: Storage profiles cannot be created in the SCv2000 Series arrays. A storage profile is easily created using the SCOS UI.
Each storage profile may be assigned to single or multiple volumes. Figure 20 shows the custom storage profile created in Figure 19 assigned to several volumes.
7 7.1 7.2 Design considerations Once the storage requirements have been determined, a solution may be designed with typically two or three tiers of storage. However, some environments may only use a single tier. Most often in multi-tiered solutions, the top tier is sized for 100 percent of the I/O and 20 to 30 percent of the capacity. However, specific customer needs may require deviation from this starting point. These tiers may consist of SSDs, HDDs, or a combination of the two.
The following conditions are typical reasons that may alter a standard SC Series setup. Site requirements for isolated disks: Multi-tenancy or other business requirements High-write-bandwidth applications: Seismic recording, video recording, video editing, surveillance For these types of applications, create a storage type with a set of disks to allow isolation of the applications, increase or decrease availability, and provide predictive performance.
8 34 Summary SC Series arrays take advantage of disk tiering, RAID levels, and automation to fully optimize disks and data efficiently. RAID levels are allocated to meet the most demanding requirements of capacity, availability, and performance. Understanding the implementation of the SC Series virtualization and automation that allows data to move between types of disks and RAID levels helps in selecting proper protection and performance to match the business requirements.
A Additional resources Dell.com/support is focused on meeting your needs with proven services and support. Dell TechCenter is an IT Community where you can connect with Dell customers and Dell employees for the purpose of sharing knowledge, best practices, and information about Dell products and installations. Dell Storage technical documentation libraries are listed at: http://en.community.dell.com/techcenter/storage/w/wiki/2631.storage-solutions-technical-documents.