1 Introduction
1.1 Preamble
This Enterprise Performance Test Specification (PTS-E) 1.1 is a companion specification to the
SNIA Solid State Storage Client Performance Test Specification (PTS-C) 1.1. Both Performance
Test Specifications (PTS) are intended to be used to obtain reliable and comparative
measurement of NAND Flash based solid state storage devices. In both PTS in their present
form, the tests and methodologies are designed to use a synthetic, or known and repeatable,
test stimulus applied to a solid state storage product at the device level. In these PTS, "device
level" refers to measurement of block IO at the physical device level as opposed to file system
IO in the host Operating System.
Both of these PTS are based on test and preparation methodologies developed by the SNIA
SSS TWG for performance test of NAND based solid state storage. NAND Flash based solid
state storage (SSS) performance tends to be highly dependent on the write history of the SSS
device, the type of stimulus applied to the SSS device, as well as the test environment (both
hardware and software) in which the test stimulus is applied and measurements taken. Much of
the preconditioning, test condition set up and parameters take these SSS behaviors into
consideration.
These PTS do not require the use a specific test environment, but test tool requirements,
capabilities and examples are set forth in the specifications. Care should be taken by the test
operator to ensure that the test hardware does not bottleneck the SSS device performance, that
the OS or test software tool has minimal contribution to test measurements, and that the same
hardware and software test combination is used when comparing performance results of
different SSS devices.
This Enterprise PTS-E 1.1 differs from the Client PTS-C 1.1 in the preparation of the Device
Under Test (DUT) for steady state performance measurement and in the amount and type of
test stimulus applied to the DUT. For example, preconditioning LBA ranges may be limited in
the Client PTS-C to less than 100% of the available LBAs while the test stimulus Active Range
may be limited to a reduced number of uniquely touched LBAs (see Client PTS-C Specification).
The use of limited preconditioning and test active ranges are meant to provide test stimulus that
share more characteristics with empirically observed Client workloads.
Readers and industry members are encouraged to participate in the further SNIA SSS TWG
works and can contact the TWG at its website portal at http://www.snia.org/feedback/.
1.2 Purpose
Manufacturers need to set, and customers need to compare, the performance of Solid State
Storage (SSS) devices. This Specification defines a set of device level tests and methodologies
intended to enable comparative testing of SSS devices in Enterprise (see 2.1.5) systems.
Note: While the tests defined in this specification could be applied to SSS devices based on any
technology (RAM, NAND, etc.), the emphasis in this specification, in particular regarding
Preconditioning and Steady State, is oriented towards NAND.
1.3 Background
A successful device level performance test isolates the device being tested from the underlying
test platform (HW, OS, Tools, Applications) so the only limiting variable in the test environment
is the device being tested. To achieve this goal with NAND-based SSS devices, in addition to
typical system/device isolation issues, the test, and test methodologies, must address attributes
unique to NAND-based flash media.
SSS PTS-Enterprise Version 1.1 SNIA Technical Position 17
我的内容管理
展开
