Administrator Guide

Internal disk management
SSDs use multiple algorithms to manage SSD endurance features. These include wear leveling, support for Unmap commands,
and over-provisioning to minimize write amplification.
Wear leveling
Wear leveling is a technique for prolonging the service life of some kinds of erasable computer storage media, such as the flash
memory used in SSDs. It attempts to ensure that all flash cells are written to or exercised as evenly as possible to avoid any hot
spots where some cells are used up faster than other locations. There are several different wear leveling mechanisms used in
flash memory systems, each with different levels of success.
Vendors have different algorithms to achieve optimum wear leveling. Wear leveling management occurs internal to the SSD. The
SSD automatically manages wear leveling, which does not require any user interaction.
Overprovisioning
The write amplification factor of an SSD is defined as the ratio of the amount of data actually written by the SSD to the amount
of host or user data requested to be written. This is used to account for the user data and activities like wear leveling. This
affects wear leveling calculations and is influenced by the characteristics of data written to and read from SSDs. Data that is
written in sequential LBAs that are aligned on 4KB boundaries results in the best write amplification factor. The worst write
amplification factor typically occurs for randomly written LBAs of transfer sizes that are less than 4KB and that originate on
LBAs that are not on 4KB boundaries. Try to align your data on 4KB boundaries.
TRIM and UNMAP commands
A command (known as TRIM in the ATA command set and UNMAP in the SCSI command set) allows an operating system to
inform an SSD of the blocks of data that are no longer considered in use and can be wiped internally.
Data retention
Data retention is another major characteristic of SSDs that all SSD algorithms take into account while running. While powered
up, the data retention of SSD cells are monitored and rewritten if the cell levels decay to an unexpected level. Data retention
when the drive is powered off is affected by Program and Erase (PE) cycles and the temperature of the drive when stored.
Drive Writes per Day
DWD or DWPD refers to Drive Writes Per Day. Disk vendors rate SSD endurance by how many writes can occur over the
lifetime of an SSD. As lower-cost SSDs that support fewer drive writes per day become available, the cost benefit analysis of
which SSDs to use is highly dependent on your applications and I/O workload, as is the ratio of SSDs to conventional drives. In
some environments, a ratio of 10% SSDs to 90% conventional drives, when combined with Dell EMC real-time tiering, can yield
dramatic performance improvements.
Because data is characterized every five seconds and moved to the appropriate storage device, no fixed rule is used to
determine which SSDs are used. For this reason, using SSDs with the same DWPD values is advised.
About SSD read cache
Unlike tiering, where a single copy of specific blocks of data resides in either spinning disks or SSDs, the Read Flash Cache
(RFC) feature uses one SSD read-cache disk group per pool as a read cache for frequently accessed data only. Each read-cache
disk group consists of one or two SSDs with a maximum usable capacity of 4TB. A separate copy of the data is also kept in
spinning disks. Read-cache content is lost when a controller restart or failover occurs. Taken together, these attributes have
several advantages:
● The performance cost of moving data to read-cache is lower than a full migration of data from a lower tier to a higher tier.
● Read-cache does not need to be fault tolerant, potentially lowering system cost.
● Controller read cache is effectively extended by two orders of magnitude, or more.
20
Getting started