Community
cancel
Showing results for 
Search instead for 
Did you mean: 
idata
Community Manager
1,657 Views

P3700 wear info

i neet to monitor wearout info of my drives.

with my X25-E and S3700 i simply check SMART values like Media_Wearout_Indicator, Power_On_Hours and Host_Writes_32MiB

now with P3700 the closest i can find is "Wear Leveling Count" and "Timed Workload - Media Wear"

no info about how many hours the drive is at work and how many bytes have been written.

i'm using latest ssd toolbox and datacenter tool (isdct show -sensor -intelssd)

where did you hide that two indicators?

and also what's the difference between those two wear indicators? which one should i watch for?

0 Kudos
8 Replies
ASouz7
Honored Contributor II
57 Views

ISRV,

We are going to check this for you and get back to you shortly.

idata
Community Manager
57 Views

i run this:

isdct dump -intelssd 0 datatype=nvmelog logid=202

and got this:

- CVFT5045008H400BGN.F4 -

ID: F4

Description: NAND Bytes Written

Normalized: 100

Raw: 0x20B

- CVFT5045008H400BGN.F5 -

ID: F5

Description: Host Bytes Written

Normalized: 100

Raw: 0x6

how to translate this?

what is nand and host bytes written?

and also which one to watch for:

- CVFT5045008H400BGN.AD -

ID: AD

Description: Wear Leveling Count

Normalized: 100

Raw: 0x100020001

- CVFT5045008H400BGN.E2 -

ID: E2

Description: Timed Workload - Media Wear

Normalized: 100

Raw: 0xFFFF

but still no info about how many hours online

jbenavides
Honored Contributor II
57 Views

Hello ISRV,

The SMART Attributes of the Intel® SSD DC P3700 and other Intel® NVMe* SSD's may be different than those of SATA SSD's.

The Normalized value of the Wear Leveling Count decrements from 100 to 0 as the blocks of the drive are written to/erased. This behaves similar to the Media Wearout Indicator of other drives.

In your case, it still shows 100, which means the average erase cycles of the NAND chips is still minimum in comparison to the rated/maximum number of erase/programming cycles.

Timed Workload is present in the SMART attributes of the P3700 and behaves pretty much like it does in other SSD's.

For more information about this, you can check the following documents:

http://www.intel.com/content/dam/www/public/us/en/documents/product-specifications/ssd-dc-p3700-spec... Intel® SSD DC P3700 Series - Product Specification

idata
Community Manager
57 Views

so "Wear Leveling Count" and "Timed Workload - Media Wear" are just the same?

"Normalized" both 100 currently, but "Raw" are different - 0x100020001 and 0xFFFF

ok, but what about how many mb/gb/tb written and how many hours the drive is in use.

how to check those values?

i've discovered "NAND Bytes Written" and "Host Bytes Written", but don't know what exactly they mean.

again, there's normalized and raw values. but how to understand those i don't know.

maybe there's other interesting commands to isdct.exe ?

jbenavides
Honored Contributor II
57 Views

Timed Workload and Wear Leveling Count are not the same. You can review the details and description for each in the Product Specifications document I linked before.

Timed workload - Media Wear can be used by system administrator's to measure the media wear induced on the drive under a specific workload, this may help estimate the expected lifetime of a drive under system-specific conditions.

Another SMART attribute you may find useful to monitor drive wear is "Percentage Used Estimate", this value increases as the drive is used over time. A value of 100 indicates that the estimated endurance of the device has been consumed.

 

Wear Leveling Count can also provide endurance information, based on the rated life cycles.

In regards to "Host Bytes Written" and "NAND Bytes Written", they refer to the amount of bytes writes requested by the host, and the actual amount of bytes writen to NAND memory chips. These values can be different because of the tasks run by the controller and the technologies it uses to handle the data.

The full list of SMART attributes and their description is found in the http://www.intel.com/content/dam/www/public/us/en/documents/product-specifications/ssd-dc-p3700-spec... Intel® SSD DC P3700 Series - Product Specification.

Also, you might want to check the Intel® SSD Data Center Tool User's Guide, it has all the commands and options supported by ISDCT. The guide can be obtained in the page: https://downloadcenter.intel.com/download/23931/Intel-Solid-State-Drive-Data-Center-Tool Download Intel® Solid-State Drive Data Center Tool

idata
Community Manager
57 Views

thanks. looks like all the info i want is in the "Table 18 SMART Attributes (Log Identifier 02h)"

will check those on my drives later

idata
Community Manager
57 Views

have to lift up this thread again.

so "Host Bytes Written" is how many 32bytes blocks have been requested to write by operating system.

and "NAND Bytes Written" is the same 32b blocks that was actually written to ssd?

so when saying 400gb p3700 write endurance is 7300Gb this means 7300Gb of "NAND Bytes Written", right?

i've checked few of my new drives and there i see "nand" is 3+ times higher than "host".

is that normal? because then looks like actual, real life endurance is 3 times lower than it stated in pdf's (http://www.mouser.com/ds/2/612/ssd-dc-p3700-spec-769407.pdf http://www.mouser.com/ds/2/612/ssd-dc-p3700-spec-769407.pdf )

idata
Community Manager
57 Views

just did a quick test.

first i've copied 11.4gb file to tmpfs (memory)

checked my smart info:

nand: 10203 = 326.496gb

host: 3360 = 107.520gb

then i've copied that ram file to ssd and got:

nand: 10572 = 338.304gb

host: 3716 = 118.912gb

so that operation took:

11.392gb of host writes and 11.808 of nand writes.

i have no clue why i have currently nand writes 3 times higher than host writes, maybe they become closer after few months of use.

i still want to know which of those values i have to watch and compare to stated lifetime endurance (7.3Pb for 400gb version)?

how those 7.3Pb was calculated in your tests?