Drives don't connect when plugging in the T4 cables

Posted by: @softraid-support

↑

@henry-in-florida 

Did you determine however, that only using the other drives give better performance?  (no slowdowns). But you also discovered that no combination gave you the numbers I get with AJA?

Both are correct. 

Posted by: @softraid-support

↑

@henry-in-florida 

Only a "certify disk" can try to refresh the media. verify does not do anything for flash media, it requires a write/ then a 0 write to clear the cells. A "zero disk" only marks the "directory" on the disk that it is empty, nothing is actually erased. 
I have never used the Magician software, some users have since they made a Mac version. I would have to find a samsung drive and try it to know how it works.

On a "bad" drive the verify made absolutely no difference. On a marginal drive there was a small improvement to the Write Performance and nothing changed on the Read Test. The bad drive (840EVO) fails miserably in the RAID 0 comparison testing and notably, both the bad drive (840 EVO, 850EVO) are in the same enclosure. The 840 has 39% media life remaining as tested in Drive DX, SATA, and SoftRAID reports. The other 850EVO that has slowdowns still has reasonable life and if a way was found to update the software I would keep it.I am using it now as a Parity Drive in the Test Build RAID4 with minimal speed influence. Wanna see? 

I go by the speed primarily but also I go by the error counts shown in Write test(superimposed red line and the squiggly errors that appear when things go awry. The four drive (in split enclosures) RAID0 test without the parity drive is much more stable. If I restart/reconnect the array, the error rates seem better. Over multiple hours of testing the results on the RAID0 2TB seem very stable and pretty error-free. 

Will now put it back on the Hub and retest the RAID 0 there. I'm thinking that except for the agreement with your figures (which I view as maximum desired), I don't see the need to go further than previous steps. OTOH, if you have a magic cure for the speed, I'm open to it! 

Here's a screenshot of the second drive tested singly to show it slows down regardless of whether it's in a RAID or not. and also that upon checking the latest Samsung Driver (FW) serial, it is on the latest firmware, not an update issue, and the Magician Software doesn't allow updates and the Verification as you indicate does nothing on this issue, neither does reformatting, re-initializing or anything else I can think of to do here. 

Here's the info on the RAID0 with 4 apparently working drive in place through the Hub and various other device on the T4 link cable (one cable through the MBP:

Ordering replacement drives (Crucial BX500 SATA SSD). Thank you for your assistance. Sorry, but I think we've proven the case for slowdowns in the storage media, perhaps that should have been better detected by SoftRAID, but were detected by swapping and elimination various possible other components in the mix, at least to my own satisfaction and with lack of support by the drive MFR, likely far out of warranty. 

@SoftRAID Support,

Replaced the culprit drives - the slow-downs: two Samsung drives 840, 850EVO (Crucial Tech BX500 6Gbs SATA) Testing 6 Drive Test Array RAID 0. Re-established the previous connections: The daisy chain is back in place. Will put the original RAID4 scenario back in place and we're ready to restore the data (~1.8TB) from backup!

@SoftRAID Support,

For you or anyone else interested, found that Crucial BX500 (upon testing a pair) make a pretty poor product for a RAID install or for that matter any general purpose. Poor write performance far below quoted speeds 50% below read speed! 

Second, I found by trial and error that RAID performance suffers from use as an encrypted storage format (APFS Encrypted) which when enabled in MacOS slows down write speed particularly by 40% even on RAID0 and further on RAID4 or. RAID5. This is documented in several sources I found. Hence, I may just stick with RAID0. Don't know if you have any opinions however did note that SoftRAID doesn't offer that functionality at all. 

@henry-in-florida 

Glad this is figured out. 

Couple comments:
"verify" does nothing for flash media, only a "certify" can help clear pre-written sectors and refresh the media.
encryption does have a strong write impact on all volumes, RAID 0 and RAID 10 are best for encrypted volumes.
One of the features we have thought about with SoftRAID is poor performance detection. Its a difficult feature, as it would slow performance unacceptably (having constant monitoring), but ideas are being investigated.

@henry-in-florida 

Our SSD's give good performance BTW.

If you certify the drives, you will also see any failing drives by very long certify times. 

Its also true you had those drives a long time, I think SSD's in mission critical scenarios ought to be replaced in the 30-40 thousand hour range. They can be used separately, but for critical applications, best to replace automatically.

Its interesting to watch how different drives behave as their spare sector counts drop below 75%. It should not impact performance by itself, but in your case, the drives behave worse and worse as their spare sectors got used up.

Posted by: @softraid-support

↑

@henry-in-florida 

Glad this is figured out. 

Couple comments:
"verify" does nothing for flash media, only a "certify" can help clear pre-written sectors and refresh the media.
encryption does have a strong write impact on all volumes, RAID 0 and RAID 10 are best for encrypted volumes.
One of the features we have thought about with SoftRAID is poor performance detection. Its a difficult feature, as it would slow performance unacceptably (having constant monitoring), but ideas are being investigated.

@SoftRAID Support

Thanks for comments.

Also found out that mix/match with non-matching SATA drive configurations make a big difference to write speed particularly. You have to drill down pretty deep for all the details and when done, there's still more under the hood the brands don't say. So stick with brand that works is my thought. Will stick with the Samsung 870EVO for certified compatibility.

RAID0 working best as you said for my situation. Waiting on drive replacements and limping along with RAID0 APFS(encrypted) 2TB, 4 drive 500GB volume which is temporarily large enough for the data restore at this time. 

Posted by: @softraid-support

↑

@henry-in-florida 

Our SSD's give good performance BTW.

What would be a 100% match to 870EVO?

@henry-in-florida 

With your concern about mix and match, I don't know. But the drives I used in the test results were Mercury Extreme 1TB drives.
https://eshop.macsales.com/shop/SSD/OWC/Mercury_6G
The Pro gets you two more years warranty, I don't think the performance difference is significant for your use.
they cost a little more than the Samsung, I notice.

Posted by: @henry-in-florida

↑

Posted by: @softraid-support

↑

@henry-in-florida 

Did you determine however, that only using the other drives give better performance?  (no slowdowns). But you also discovered that no combination gave you the numbers I get with AJA?

Both are correct. 

 

Actually the testing you had me do, combining 500GB into a RAID0 for testing, let me down a rabbit hole, since by themselves the BX-500's worked, but in combination 4 mixed drives, they certainly do not. So, no thanks on combining drives, without a deep dive into their qualities and preferably proving them. So, not this time. If OWC wants to give me six for the price of two... Maybe. 

@SoftRAID Support

Another thing I notice on both my RAID (NVe & SATA) checking performance, hours and media life is that in RAID4 Array the speeds for both is lower than RAID0 and still slower when BOTH RAID4+encryption is employed.

When RAID4+encryption is employed the speeds on NVe media, what is the expectation of speed?

@henry-in-florida 

You requested NVme, so that is what I used. An 8 drive NVme enclosure (the Thunderblade 8)

All numbers rounded to nearest 50. This is not definitive, just a first approximation.

RAID 4 on an M1 Studio
APFS
Write 2100      Read 2650
APFS encrypted:
Write   650      Read 2600

RAID 0 on an M1 Studio
APFS
Write   2650     Read 2950
APFS encrypted:
Write  2000     Read 2600

RAID 4 on an M4 Mini
APFS
Write 1400     Read 2700
APFS encrypted:
Write  350     Read 2050

RAID 0 on an M4 Mini
APFS  2550   Read  3000
APFS encrypted:
Write 1550    Read 3000

RAID 5 on an M4 Mini
APFS
Write 1400    Read 2650
APFS encrypted:
Write   300   Read  1850

For reference:
Apple's RAID 0 64K SU on M4 Mini
APFS Encrypted:
Write 1600    Read 2950
(no significant performance difference)

@RAIDSupport

Thanks for the time spent guiding me through this issue! And the reference materials supplied, too. 😀