Single drive RAID-0 scenario

[peterbata]

Hello all. Would such a scenario be possible on the HP Z820 with it's Integrated LSI SAS 2308 Controller? Your thoughts?

Thank you folks. Have yourselves a  safe and wonderful week.

Peter

[DGroves]

before i answer this, i would like you to explain why you think this is possible 

and before you answer, consider that raid 0 is used to increase the effective data xfer speeds so i'm curious as to how a single drive is supposed to suddenly become faster and overcome it's hardware limits

[peterbata]

Good morning @DGroves. From your response it is obvious to me now that I could have / should have provided more details.

I was curious to know if for example I have placed in my 4 internal bays 4x 8TB drives. In VMware ESXi I would like to create 4 datastores (1 for each of those 8TB drives). As opposed to 2x 16TB datastores (2 pairs)

I know that this was possible when I was using a Proliant ML350P Gen8 server. In that case I had 6x 600Gb SAS drives - Bay-1 drive slots 1-6 (configured as RAID-5.

I then added 2x 2TB SAS drives Bay-1 drive slots 7-8. These were configured as RAID-0.

In ESXi I had:

datastore1 = a little under 3.6TB - 6x 600Gb

datastore2 = a little under 2TB - 1x 2TB

datastore3 - a little under 2TB - 1x 2TB

I could then assign each one of those datastores to 4 seperate virtual machines (eg. Nextcloud, Server 2019, Plex media server, etc.)

I suppose that ultimately I was curious to know if this was a limitation of the Z820's integrated SAS controller. 

The ML350P used the P420i/512M FBWC Controller

You will have to forgive me if I am confusing things even further. In any case, I really do appreciate your time and assistance. Peter

[DGroves · Accepted solution]

data xfer depends on several things with the scsi controller and scsi drive being the two most important along with the data path size

you can go no faster than what the scsi host controller will allow

you can go no faster than what the scsi drive itself can xfer to the scsi controller

to overcome the limitation in speed of a single drive,.... raid 0 lays data between two (or more) drives and the data size written can be set to various sizes (stripe size) if the data chunk is smaller than the stripe size it will be written to one drive if it exceeds the stripe size then it will be broken up between the 2 (or more) drives, this is what allows raid 0 to exceed the normal r/w speeds of a single drive

there are several common raid levels, with 0/1/5/10 being available on most scsi controllers lowend controllers may only support raid 0/1 and high end raid 50 you can also have dual scsi host controllers active using "dual port" scsi drives

software based raid usually allows a lot more raid modes than dedicated raid cards and is now what is recommended for todays large capacity drives over dedicated hardware controllers

the Proliant ML350P Gen8 had numerous dedicated scsi controller card options ranging from entry to high end capability

the z820 has a LSI low end single chip "mustang" controller chip that lacks onboard cpu for x/or calculations, no cache and no battery backup as such it can only really do raid 0/1 it's raid 5 capable but this mode is extremely slow

for EsXi i recommend a ESXi supported LSI Raid card or my personal fav, the adaptec 71605 which unlike the LSI cards supports SATA/SAS and raid/JBOD at the same time and has true SCSI passthrough allowing JBOD drives to be removed from the controller and connected to any other SATA controller and most other SAS/SATA controllers

using SATA male to male adapters it's possible to connect the adaptec card to the z820's internal 4 hot swap bays

https://en.wikipedia.org/wiki/Standard_RAID_levels

[peterbata]

Good morning @DGroves and thank you again for that wealth of information. This response is a definite keeper as are all your posts of course. Have a safe and wonderful day. Peter

[peterbata]

Hello again @DGroves When using the Adaptec card I am guessing that the SAS/SATA cables have to be extended from the cage's backplane. If so, what have you used to accomplish that. Thank you

[DGroves]

for the adaptec card i mentioned you need a "FORWARD" type mini sas to sata/sas cable

https://www.ebay.com/itm/143290062511?itmmeta=01HPHKYVWVEJZ3VSBTEY7Y2R08&hash=item215cc0d6af:g:takAA...

to connect the cards cable to the HP internal bays cables you need 4 SATA male to male adapters

https://www.ebay.com/itm/283824303900?itmmeta=01HPHM7C6FJEY1X2R87KS61YEN&hash=item42153f2f1c:g:zg8AA...

[peterbata]

Forgive me for not getting back to you sooner. The bad news id that I tested positive for Covid and the last few days have been a living hell. Thank you again for passing along that information. Always immensely appreciated.

[peterbata]

Hello again @DGroves. I would like to ask you a few questions regarding the Adaptec 71605 card that you referred to in your reply. I have not committed to the card as yet.

I fooled myself (or just simply did not do my due diligence) when I decided to purchase 2x HGST 8TB SAS Drives. See pic for drive details:

HGST 8TB 

I recall you having said that "the z820 has a LSI low end single chip "mustang" controller chip that lacks onboard cpu for x/or calculations, no cache and no battery backup as such it can only really do raid 0/1 it's raid 5 capable but this mode is extremely slow"

This would probably explain why the Z820's RAID configuration utility does not recognize either of the drives. Or, they are just not compatible with the machine or the LSI 2308 Controller at all.

If I do purchase the Adaptec 71605 is there any way of knowing that I will not run into the very same issue. Perhaps, there exists an Adaptec compatability guide that I can refer to. Again, your time and assistance is greatly appreciated. Peter

[SDH]

DGroves has helped me learn about the built-in Windows utility Diskpart. It has been valuable. In the past I've salvaged multi-TB odd server HDDs one at a time using Diskpart from within Windows running it from an elevated CMD (CMD run as admin). I'd consider putting one of those drives in a spare drive bay and check in Computer Management under Disk Management if you can see it. If you can see it there you likely can use Diskpart to clean it of old odd server OS bits and use Diskpart to GPT partition it. Those will take some time to clean and configure if you can see them. I then do a long-type NTFS format with default settings otherwise once they have been GPT partitioned. The long-type format maps out bad sectors. I have a back room old HP xw workstation running W10Pro64 set aside for this purpose (and other boring things).

Peter, he will know much more about what to do on this issue than me.