Performance and Reliability - IDEMA
Performance and Reliability - IDEMA
Performance and Reliability - IDEMA
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
St<strong>and</strong>ardized Testing for SSD’s <strong>and</strong> HDDs<br />
<strong>Performance</strong> <strong>and</strong> <strong>Reliability</strong><br />
<strong>IDEMA</strong> Technical Symposium<br />
May 2008<br />
Ed Dawson<br />
Flexstar Technology<br />
edawson@flexstar.com<br />
www.flexstar.com
Agenda<br />
<strong>Performance</strong> & <strong>Reliability</strong> Testing<br />
• Why Test<br />
• What is St<strong>and</strong>ardized Testing<br />
• <strong>Reliability</strong> & <strong>Performance</strong> Testing<br />
• Device Specific Tests<br />
• Storage Test Evolution
• Engineering Verification<br />
– EVT DVT etc<br />
– Development debug<br />
• OEM Qualification<br />
– Systems qualification<br />
– <strong>Performance</strong> evaluation<br />
– Four corner testing<br />
– ORT<br />
• Device Manufacturing<br />
– Meet specifications<br />
– <strong>Reliability</strong> / early life failures<br />
– Fast failure analysis<br />
• Field<br />
– Failure analysis<br />
– RMA confirmation<br />
Why Test?
• Engineering Verification<br />
– EVT DVT etc<br />
– Development debug<br />
• OEM Qualification<br />
– Systems qualification<br />
– <strong>Performance</strong> evaluation<br />
– Four corner testing<br />
– ORT<br />
• Device Manufacturing<br />
– Meet specifications<br />
– <strong>Reliability</strong> / early life failures<br />
– Fast failure analysis<br />
• Field<br />
– Failure analysis<br />
– RMA confirmation<br />
Why St<strong>and</strong>ardized Tests?
1<br />
St<strong>and</strong>ardized Testing<br />
Key Features of St<strong>and</strong>ardized Testing<br />
• Accuracy<br />
• Repeatability<br />
– Ability to reliably recreate same test<br />
conditions globally<br />
• Tight control over test conditions<br />
– Environmental<br />
– Test tool environment<br />
• Scripting Based Tests<br />
– Suits individual product requirements<br />
– Easily sharable<br />
• Ability to debug in place<br />
– Interrogate / Trace failure conditions<br />
– Remote testing/debugging<br />
• Provides Data Collection & Correlation<br />
– Consistency across the industry<br />
– Underst<strong>and</strong>able Reports<br />
70,000.00<br />
60,000.00<br />
50,000.00<br />
40,000.00<br />
Rate (KB/s)<br />
30,000.00<br />
20,000.00<br />
10,000.00<br />
0.00<br />
33<br />
65<br />
97<br />
129<br />
161<br />
193<br />
225<br />
257<br />
Serial Number VNRD2AC4CGVRHS<br />
289<br />
321<br />
353<br />
385<br />
417<br />
449<br />
481<br />
513<br />
545<br />
577<br />
609<br />
641<br />
673<br />
Test Points<br />
705<br />
737<br />
769<br />
801<br />
833<br />
865<br />
897<br />
929<br />
961<br />
993<br />
<strong>Performance</strong>
Stop the questioning & finger pointing<br />
by st<strong>and</strong>ardizing<br />
The benefits of st<strong>and</strong>ardized, independent test systems<br />
– St<strong>and</strong>ard test hardware<br />
– Unbiased <strong>and</strong> fair<br />
– Repeatable test solutions<br />
– Common test scripts<br />
– Industry st<strong>and</strong>ard interfaces<br />
– Correlation between<br />
manufactures <strong>and</strong> EOM’s<br />
– Speeds failure analysis<br />
– Investment protection
St<strong>and</strong>ards <strong>and</strong> <strong>Performance</strong> Testing<br />
Storage Test Requirements<br />
• Seek, transfer, & throughput performance<br />
• Interface verification<br />
• Specification compliance<br />
• Head-disk interface integrity<br />
• Media scans<br />
• Voltage margins<br />
• Long term reliability<br />
• Environmental & Edge case
<strong>Performance</strong> Testing<br />
General <strong>Performance</strong> Testing<br />
– Seek & Transfer<br />
• Sequential<br />
• R<strong>and</strong>om<br />
• w/ & w/o compare<br />
• etc, etc…<br />
– Multiple streams<br />
– Variable data stream width/ block<br />
size, knee testing
<strong>Performance</strong> Testing<br />
HDD Specific <strong>Performance</strong> Testing<br />
– Radial <strong>Performance</strong><br />
– Seek mode performance<br />
SSD Specific<strong>Performance</strong><br />
– Fragmented Wear Leveling
<strong>Reliability</strong> Testing<br />
Design &<br />
Manufacturing<br />
Targeted test paths<br />
Field &<br />
<strong>Reliability</strong>
<strong>Reliability</strong> Testing<br />
HDD design verification / qualification<br />
( rotational media)<br />
Root Cause<br />
– Design/Components/DRAM<br />
– Component Failures / i.e.<br />
red phosphorous<br />
– Head/Media instability<br />
– Design margin<br />
Test Process<br />
⇒ Pattern reads/writes/verify<br />
⇒ Power cycling<br />
⇒ Extended test at temperature<br />
⇒ Voltage margining<br />
⇒ Temperature Cycling<br />
⇒ Voltage margining & 4 corner
<strong>Reliability</strong> Testing<br />
HDD variable field failure modes<br />
(variable factors in rotational media)<br />
Root Cause<br />
– Contamination<br />
– H<strong>and</strong>ling, head laps<br />
– Lube issues<br />
– Motor bearings<br />
– Growing defect lists<br />
Test Process<br />
⇒ Multiple full media R/W.<br />
⇒ Head/Surface defects<br />
⇒ W/R adjacent tracks<br />
⇒ Dwell at elevated temperature
<strong>Reliability</strong> Testing<br />
SSD design verification / qualification<br />
– Endurance<br />
Root Cause<br />
– <strong>Performance</strong>-mfg. variability<br />
– Bit failures / Data Retention<br />
– Component Failures / i.e.<br />
red phosphorous<br />
– Write splice<br />
– Metadata corruption<br />
– Write performance<br />
– Erase failures<br />
– Design Margin<br />
Test Process<br />
⇒ Multiple writes<br />
⇒ <strong>Performance</strong> verification<br />
⇒ Disturb testing / pattern writes<br />
⇒ Temperature Cycling<br />
⇒ Voltage margining & 4 corner<br />
⇒ Power cycling mid writes<br />
⇒ R<strong>and</strong>om I/O w/ power cycling<br />
⇒ Cold write<br />
⇒ Margined erase<br />
⇒ Four Corners
<strong>Reliability</strong> Testing<br />
SSD testing for variable failures<br />
Root Cause<br />
– Wear leveling performance<br />
– Data Retention<br />
– Reallocation errors<br />
Test Process<br />
⇒ Fragmentation tests<br />
⇒ Temperature & power cycling<br />
⇒ RW tests w/ power cycling<br />
⇒ Write splice, cold writes
Variable Failure Modes<br />
Post Manufacturing Failure Modes<br />
HDD<br />
• Contamination<br />
• H<strong>and</strong>ling<br />
• Head slaps<br />
• Lube Issues<br />
• Motor Bearings<br />
• Environmental<br />
strains<br />
ODD<br />
• Media Failures<br />
• Motor Issues<br />
• Mechanical<br />
• Optical pickup<br />
SSD<br />
• Endurance
AFR Predictions<br />
Predicting a New ERA<br />
• Historically, AFR cannot be correlated to manufacturing<br />
or line integration failure rates.<br />
• AFR will be more predictable in SSD, primarily due to the<br />
eliminating of mechanical failures.<br />
• St<strong>and</strong>ardized front end testing for SSD's will lead to more<br />
a more reliable predicted AFR benefiting storage<br />
integrators.
SSD vs. HDD<br />
What the Customer Expects<br />
(in a world of tests developed by key OEMs)<br />
• Complete functional equivalence <strong>and</strong> independence.<br />
• Interface testing<br />
• <strong>Reliability</strong> demonstration<br />
• <strong>Performance</strong> demonstrations<br />
• Endurance demonstration<br />
• Power on / spin-up / sleep / resume from sleep<br />
compliance.
Summary<br />
What’s the role of organizations like<br />
<strong>IDEMA</strong>?<br />
At a point where test st<strong>and</strong>ards can be developed for<br />
new solid state storage that are reflective of the<br />
technology.
Thank You<br />
Ed Dawson, Flexstar Technology