Page 6 from 11
Carol is a corridor warrior who spends her time moving from meeting to meeting. It is important for
her to be ready to give a presentation at the drop of a hat. She has been frustrated in the past with
slow boot times that meant she had to stall while waiting for her presentation to come up. With her
new HP EliteBook 2760p equipped with a high-performance SSD, she is able to boot the machine in
seconds and access her files much faster than on her previous system.
Enhancing write reliability and performance
Historically, perceptions concerning write reliability and performance may have inhibited the broader
adoption of SSDs. In response, vendors have been introducing technologies to improve these
Since NAND flash cells can only support a limited number of erase cycles before becoming unusable,
it is critical for SSDs to include a wear-leveling mechanism designed to distribute erasures evenly
The need for wear-leveling is exacerbated by the behavior of typical file
systems that have been developed for use with HDDs. Such file systems
may rewrite data structures repeatedly in the same area.
While the even distribution of dynamic (changing) data can extend the lifecycle of an SSD, some
wear-leveling implementations periodically move static data to make less-used cells available for other
Mitigating the effects of write amplification
Write performance in an SSD can be impacted by a phenomenon known as write amplification,
which is a result of the following SSD characteristics:
Data can only be written into an empty NAND flash cell. Thus, if you wish to write to a particular
cell and are unsure of its state, you must first erase the cell.
While writes to an SSD are sector-level (4 KB); however, due to mechanical limitations, erases are
block-level (128 x 4 KB = 512 KB).
Cells that are deleted by the user or the operating system are not erased.
When you first receive an SSD, all its NAND flash cells are in an erased state; thus, writes can
initially be performed sequentially. However, later writes require a read-erase-modify-write cycle, as
shown in Figure 1.
With HDDs, blocks can be over-written individually.
Page 7 from 11
Figure 1.Overview of a read-erase-modify-write cycle in a typical SSD
In this example, a sector that had previously been deleted is to be rewritten with new data. The read-
erase-modify-write cycle is performed as follows:
1.The block (512 KB) containing the deleted sector is read from NAND flash; meanwhile, the new
data is temporarily stored in the SSD’s cache.
2.After the block is written to cache, the original data in NAND flash is erased.
3.Within cache, the new data is added to the block, replacing the deleted data.
4.The updated block is written back to NAND flash.
Thus, for each write operation, you must move or write data multiple times, leading to write
amplification. Since each SSD can only support a finite number of writes, write amplification shortens
the life of the device; in addition, write amplification consumes NAND flash bandwidth, which can
reduce throughput for random writes.
SSD vendors have introduced a range of measures to mitigate the impact of write amplification during
write operations. For example, some vendors have implemented a firmware-based garbage collection
feature that, in effect, de-fragments NAND flash while the SSD is idle, without intervention from the
user or operating system. With sufficient idle time, the unused portions of the SSD can be restored to
as-new condition, thus minimizing the number of erasures required during write operations. However,
there are penalties involved in the use of garbage collection: since it works in the background, there
may be an impact on battery-life; moreover, garbage collection involves additional writes that can
prematurely age NAND flash cells. Moreover, although it operates in the background, the garbage
collection process itself adds write amplification.
Since 2009, SSDs used in HP Business Notebook PCs have support for garbage collection built into
Though not recommended by HP, third-party software products are can be used for garbage
collection on SSDs that do not provide this feature in firmware. However, such products may require
the entire contents of the SSD to be backed up before garbage collection can be performed.
Page 8 from 11
Using the TRIM command
While garbage collection is managed by the drive, the TRIM command is
sent by the device driver.
An SSD is blind; it is unaware which sectors are being used and which are free – this information is
maintained by the file system, to which the SSD has no access.
When a user or the OS deletes a file, the file system merely flags the area as being not in use; no
data is physically deleted. Since the SSD is unable to communicate with the file system, it is unaware
that space has been freed-up, which ultimately leads to write amplification.
However, with the introduction of the ATA TRIM command, the OS is now able to notify the SSD
when sectors have been freed up. Following a delete operation – and, ideally, prior to a subsequent
write – the SSD copies the block containing the deleted data to cache, erases the affected NAND
flash, and only rewrites blocks containing data that is still being used. Thus, the TRIM command can
enhance write performance by minimizing the requirement for read-erase-modify-write cycles.
In addition, note that blocks that have been trimmed using the TRIM command are omitted from
garbage collection, further reducing write amplification.
The TRIM command must be supported by both the SSD and the OS.
The following SSDs, all of which support the TRIM command, are used in 2011 HP Business
Intel Postville (X18-M G2, X25-M G2) and Postville Refresh (320)
Samsung PM810 and PM830
Micron C300 and C400
Table 1 outlines how the above SSDs are enabled at the system-level, via either a particular driver or
a vendor-supplied tool, depending on the particular platform.
Table 1.Enabling TRIM support on particular Windows platforms
Driver or tool Windows XP/Vista Windows 7
Microsoft in-box driver Unsupported Intel/Samsung/Micron SSDs
Intel Matrix Storage Manager (MSM) Unsupported Unsupported
Intel Rapid Storage Technology (RST) 9.6 or later Unsupported Intel/Samsung/Micron SSDs
Intel Solid-State Drive (SSD) Toolbox Intel SSDs N/A
Samsung TEA tool Samsung SSDs N/A
Micron Real SSD manager Micron SSDs N/A
Page 9 from 11
Linux kernel 2.6.28 and later supports the TRIM command.
Additional caveats include:
The TRIM command cannot be used with a RAID volume.
If disk encryption has been implemented, the TRIM command reveals which blocks are in use.
Upgrading an unsupported SSD
Some vendors are providing firmware upgrades that allow older drives to support the TRIM
command. A supported OS is still required.
SSDs are often over-provisioned to accommodate features such as garbage collection, wear-leveling,
and bad-block mapping. For example, a vendor might market an SSD with 128 GB of usable
capacity as an over-provisioned 120 GB drive.
Over 7% of additional over-provisioning can be available via the
difference in the nominal value of a GB (1,000,000,000) and its binary
HP has become a thought leader in the SSD space, ensuring that devices deployed in HP Business
Notebook PCs can achieve high levels of performance and productivity, battery life, and durability.
Part of this value-add is the way in which SSDs are qualified for use in notebooks.
To ensure SSDs used in HP Business Notebook PCs are capable of achieving the high standards
required, HP works with SSD vendors and original device manufacturers (ODMs) to qualify products
at both device- and system-levels. As part of this effort, HP has developed robust testing schema so
that only the best SSD products are used, regardless of vendor.
Initially, vendors are given very specific specifications and guidelines to help them deliver an SSD that
meets HP requirements. For example, mean time between failures must be at least 1.2 million hours,
while performance must satisfy HP standards.
During the development stage of an SSD device, the HP SSD engineering team works with vendors to
perform a range of functional compatibility tests and reliability demonstration tests (RDTs). After all
hardware and firmware issues have been identified and, if possible, resolved, the team releases2the
SSD for system-level testing.
Following the release, ODMs are responsible for working with the HP team and SSD vendors to
ensure the SSD is capable of achieving the specified requirements for use in an HP Business
Notebook PC. ODMs follow the HP test plan to perform device qualification via an HP test tool.
2At this stage, the root cause of any unresolved issue is not related to the SSD’s hardware or firmware. Instead, it may be caused by the systemBIOS, system hardware, or the storage driver.
Page 10 from 11
Testing includes the following areas:
SSD functional testing: Standby/resume, hibernation/resume, reboot, and more
Mechanical testing: Shock, vibration, and more
Electromagnetic interference (EMI)/electrostatic discharge (ESD)
Wireless WAN (WWAN)/radio frequency (RF)
After the SSD has passed the system-level testing, it is approved for production.
SSDs are available in select HP Business Notebook PCs and as an after-market option (AMO).
The following SSDs are available in select HP ProBook and EliteBook models:
The following SSDs are available as after-market options (AMOs) for select HP ProBook and EliteBook