Best RAID and Corresponding Drive Volume Configuration for Hyper-V Host multiple running VMs and required volumes for SQL database and file/app shares?
I have a HP ProLiant DL360 g8 (2 processors, 64GB RAM, 2x quad port Gigabit LAN cards, and a HP P420i RAID Controller that can handle all RAID configurations including Raid 6 because of the Advance License) that I am upgrading to server 2016 enterprise edition from SBS 2011. I have 8 10K 900GB SAS drives that I would like to configu workre in a RAID configuration with various volumes built on top of the RAID array(s). I am taking 6x 300GB 10K SAS drives out of the machine (to be used else where as needed).
The server will host at least 2 hyper-v VMs's - one to host exchange 2016 and the other as a file/app/web server which will require SQL Server Professional to run on the second VM. There might be a third VM to host sharepoint but this might be hosted elsewhere.
Additionally, I have an external raid array in a standalone device that is connected to the server via USB 3.0. I also have 2 1TB SSD drives likethat I would like to attach the server via a dual USB drive cradle connected via USB 3.0. There are also multiple USB 3.0 8TB external hard drives that are connected to be used as drive/data backup using Windows Server Backup (differencing backup running at least 1 but probably twice a day with replication to a remote off-site QNAP as a backup several times a week).
What is the best raid/volume confugaration for this setup. The Raid card does not have the optional battery installed.
OPTION 0: Should I do 4x 2-drive pairs in RAID 1, each for a different purpose (OS/VMs on 1st pair; Exchange DBs/Files on second pair; SQL Databases/WSUS files on third pair; File shares on fourth pair).
- OPTION 0B: Some combination of 2x 2-drive RAID 1 configs and then a 4-drive Raid 5 or Raid 10 set).
OPTION 1: Should the host OS be on a Raid 1 (900GB is a more than 3 times as much as the host OS would ever need) and then should I put the remaining 6x 900GB drives in a Raid 5 or Raid 10 configuration or as some other combination.
OPTION 2: Should I put all 8 drives in a Raid 10 configuration and split off a host OS volume (probably 200 GB), VM storage Volume (probably 600GB - 1TB), and then the rest in several smaller volumes for databases, exchange 2016 files/DBs, and finally a file/app share volume.
- OPTION 1B/2B - Same as above but should I store the VMs on the external 1TB SSD connected via USB 3.0 and if I do this can I setup some sort of backup/drive replication for the VMs.
The primary OS will most likely be my primary DC and DHCP, DNS, etc will be handled by the File/App Server VM.
The External USB RAID drive array is used for mixed storage - files/file shares (either directly or via hard link junctions to shares on the currect integrated volume), database file storage storage, WSUS repository, etc.
I have at least 500GB of files that I would like on the server file share but before this reconfiguration the file share drive was 1.2TB and only had 50GB available (at least 500GBs of this is old and will be archived to the external Raid file share). (Should this be stored in a volume directly on top of the RAID array or on virtual hard drives on that array)?
I need at least 1TB for Exchange databases (on a volume or a virtual disk?)
- Research has led me to think that virtual disks as file storage are more efficient than using a true drive attached to a VM or a SCSI drive attached but please give me your oppinion on this as well).
Approximately 11-14 users on the network. The office is a law firm so mostly file storage of documents/pdfs and video/audio files though very little streaming except to review the video/audio on network. Mostly reading/writting to the document/pdf files except for the billing software runs a sql server database accessed from each end user’s workstation and the office has a multiuser version of quickbooks where the company files are stored on the app/file server.
Please help!!!
Thank you,
Adam
The server will host at least 2 hyper-v VMs's - one to host exchange 2016 and the other as a file/app/web server which will require SQL Server Professional to run on the second VM. There might be a third VM to host sharepoint but this might be hosted elsewhere.
Additionally, I have an external raid array in a standalone device that is connected to the server via USB 3.0. I also have 2 1TB SSD drives likethat I would like to attach the server via a dual USB drive cradle connected via USB 3.0. There are also multiple USB 3.0 8TB external hard drives that are connected to be used as drive/data backup using Windows Server Backup (differencing backup running at least 1 but probably twice a day with replication to a remote off-site QNAP as a backup several times a week).
What is the best raid/volume confugaration for this setup. The Raid card does not have the optional battery installed.
OPTION 0: Should I do 4x 2-drive pairs in RAID 1, each for a different purpose (OS/VMs on 1st pair; Exchange DBs/Files on second pair; SQL Databases/WSUS files on third pair; File shares on fourth pair).
- OPTION 0B: Some combination of 2x 2-drive RAID 1 configs and then a 4-drive Raid 5 or Raid 10 set).
OPTION 1: Should the host OS be on a Raid 1 (900GB is a more than 3 times as much as the host OS would ever need) and then should I put the remaining 6x 900GB drives in a Raid 5 or Raid 10 configuration or as some other combination.
OPTION 2: Should I put all 8 drives in a Raid 10 configuration and split off a host OS volume (probably 200 GB), VM storage Volume (probably 600GB - 1TB), and then the rest in several smaller volumes for databases, exchange 2016 files/DBs, and finally a file/app share volume.
- OPTION 1B/2B - Same as above but should I store the VMs on the external 1TB SSD connected via USB 3.0 and if I do this can I setup some sort of backup/drive replication for the VMs.
The primary OS will most likely be my primary DC and DHCP, DNS, etc will be handled by the File/App Server VM.
The External USB RAID drive array is used for mixed storage - files/file shares (either directly or via hard link junctions to shares on the currect integrated volume), database file storage storage, WSUS repository, etc.
I have at least 500GB of files that I would like on the server file share but before this reconfiguration the file share drive was 1.2TB and only had 50GB available (at least 500GBs of this is old and will be archived to the external Raid file share). (Should this be stored in a volume directly on top of the RAID array or on virtual hard drives on that array)?
I need at least 1TB for Exchange databases (on a volume or a virtual disk?)
- Research has led me to think that virtual disks as file storage are more efficient than using a true drive attached to a VM or a SCSI drive attached but please give me your oppinion on this as well).
Approximately 11-14 users on the network. The office is a law firm so mostly file storage of documents/pdfs and video/audio files though very little streaming except to review the video/audio on network. Mostly reading/writting to the document/pdf files except for the billing software runs a sql server database accessed from each end user’s workstation and the office has a multiuser version of quickbooks where the company files are stored on the app/file server.
Please help!!!
Thank you,
Adam
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Philip Elder
@kevinhsieh The price is pretty good on the 5210 ION but write speeds are definitely questionable. They are great for archival storage but not much else. Micron's PRO series SSDs would be a better choice albeit a bit more expensive.
I am deploying smaller ION drives to servers right now. They run DC, file, and RDS VMs. We will see how they do.
@kevinhsieh We plan on deploying them into our test setup to thrash them. Their write rating is ~250MiB/Second so half what most mainstream enterprise SATA SSDs can do. The price makes them, and their 15.36TB NVMe cousins, very attractive. :)
ASKER
Thank you all for your quick and excellent responses!
Starting from the first response above and moving down:
Kevin, you are correct - the SSDs are Samsung 860 EVO drive - Not enterprise.
My OC (or C drive) is what I was referring to above as 200GB. What about using something like OpenFiler and having the VMs VHDX’s physically sit on another machine (I have both a mega raid SAS 9271-8i and a PERC H310 that I could utilize in 2 other machines - both Optiplex 7020 machines and 6x 300GB 10K SAS drives that I could split between them or run off the MegaRAID card on one machine). I don’t have any “production” experience hosting VM virtual disk on a different machine than that running them but I have seen it mentioned).
As for exchange, the 1TB is an overestimate but at the rate our mailboxes have been growing over the last year, this is planning for 2 years of email growth plus log files/overhead/etc...
As for licenses, I’ve 20 for Windows Server 2016 and 2 Datacenter licenses that could host as many VMs as the hardware will support. Unfortunately, I don’t have any Server 2019 license.
I think I will go with separate VHDX for the VM Data drives as opposed to connecting a physical disk to the VM. More complicated disk architecture doesn’t bother me as long as I use a good naming schema for the VHDX’s.
Are there any speed/performance increases to creating the VMDX “at once” instead of allowing it to grow? Also, what are your thoughts on VHDX speed/performance vs movability in creating a single disk file vs. split disk files?
My external 4 drive RAID array (connected by USB now) is a NexStar HX4R device that could connect via e-SATA. Any thoughts on one verse the other?
So just to confirm, you all agree that my host C drive should my DC with DNS and DHCP (approximately 60GB)? I usually don’t partition a C-drive that is less than 100GB but my rational behind that is from having long running machines that have bloated windows SXS folders that take up a lot of space and can’t be cleaned up.
What about volume format? NTFS vs ReFS? I don’t think that I can use ReFS for anything but database drives because they don’t allow for hard links or bootable drives or page file support.
Doesn’t RAID 10 have a lot of overhead and increases the amount of work the server actually has to do? I know that I read this somewhere but I don’t remember if it had to do with the RAID card and the amount of writing to the disks or if it had to do with overall server resources.
Any other recommendation or thoughts/advise that you want to volunteer beyond what I have asked is greatly appreciated.
Thanks again!
Adam
Starting from the first response above and moving down:
Kevin, you are correct - the SSDs are Samsung 860 EVO drive - Not enterprise.
My OC (or C drive) is what I was referring to above as 200GB. What about using something like OpenFiler and having the VMs VHDX’s physically sit on another machine (I have both a mega raid SAS 9271-8i and a PERC H310 that I could utilize in 2 other machines - both Optiplex 7020 machines and 6x 300GB 10K SAS drives that I could split between them or run off the MegaRAID card on one machine). I don’t have any “production” experience hosting VM virtual disk on a different machine than that running them but I have seen it mentioned).
As for exchange, the 1TB is an overestimate but at the rate our mailboxes have been growing over the last year, this is planning for 2 years of email growth plus log files/overhead/etc...
As for licenses, I’ve 20 for Windows Server 2016 and 2 Datacenter licenses that could host as many VMs as the hardware will support. Unfortunately, I don’t have any Server 2019 license.
I think I will go with separate VHDX for the VM Data drives as opposed to connecting a physical disk to the VM. More complicated disk architecture doesn’t bother me as long as I use a good naming schema for the VHDX’s.
Are there any speed/performance increases to creating the VMDX “at once” instead of allowing it to grow? Also, what are your thoughts on VHDX speed/performance vs movability in creating a single disk file vs. split disk files?
My external 4 drive RAID array (connected by USB now) is a NexStar HX4R device that could connect via e-SATA. Any thoughts on one verse the other?
So just to confirm, you all agree that my host C drive should my DC with DNS and DHCP (approximately 60GB)? I usually don’t partition a C-drive that is less than 100GB but my rational behind that is from having long running machines that have bloated windows SXS folders that take up a lot of space and can’t be cleaned up.
What about volume format? NTFS vs ReFS? I don’t think that I can use ReFS for anything but database drives because they don’t allow for hard links or bootable drives or page file support.
Doesn’t RAID 10 have a lot of overhead and increases the amount of work the server actually has to do? I know that I read this somewhere but I don’t remember if it had to do with the RAID card and the amount of writing to the disks or if it had to do with overall server resources.
Any other recommendation or thoughts/advise that you want to volunteer beyond what I have asked is greatly appreciated.
Thanks again!
Adam
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Why does a firm with <15 users have Windows Server datacenter licenses?
If my VMs have any user data on them at all, they have more than 1 VHDX. There will be an OS VHDX, and then 1 or more data VHDX. A file server will have at least OS, Data, and ShadowCopy volumes. A SQL server would have OS, PROG, DATA, and LOG volumes. Similarly Exchange has OS, PROG, DATA, and LOG volumes. At this scale though, Exchange should be O365.
I always set my VHDX to grow dynamically. We monitor space on host to be sure we don't grow beyond what the host can handle. We also size VHDX reasonably, so we don't have 1 TB VHDX with only 100 GB data on it. If you have only 100 GB data, maybe the VHDX should be 150-200 GB, depending on data growth. They can be easily expanded with no interruption. There hasn't been a performance issue with dynamic VHDX vs. fixed for nearly a decade.
If my VMs have any user data on them at all, they have more than 1 VHDX. There will be an OS VHDX, and then 1 or more data VHDX. A file server will have at least OS, Data, and ShadowCopy volumes. A SQL server would have OS, PROG, DATA, and LOG volumes. Similarly Exchange has OS, PROG, DATA, and LOG volumes. At this scale though, Exchange should be O365.
I always set my VHDX to grow dynamically. We monitor space on host to be sure we don't grow beyond what the host can handle. We also size VHDX reasonably, so we don't have 1 TB VHDX with only 100 GB data on it. If you have only 100 GB data, maybe the VHDX should be 150-200 GB, depending on data growth. They can be easily expanded with no interruption. There hasn't been a performance issue with dynamic VHDX vs. fixed for nearly a decade.
+1 for RAID 10 for everything unless you really need a giant storage for files or something that is also backed up, only then is RAID5/6 an option. Disks are cheap.
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER
I have the data center licenses from another business that failed where I sold the hardware but not the licenses. Thats why I have licenses to spare.
The other computer has lots of hard drive space to spare but not enouoh memory or CPU resources to run the VMs.
The general take away I’m seeing is that I should configure the hyper-V VM with virtual drives in the same manner I would as this was a physical server with different physical drives/volumes. Is that correct?
The other computer has lots of hard drive space to spare but not enouoh memory or CPU resources to run the VMs.
The general take away I’m seeing is that I should configure the hyper-V VM with virtual drives in the same manner I would as this was a physical server with different physical drives/volumes. Is that correct?
ASKER
Aaron, what does “+1 for RAID 10” mean?
+1 means a vote of support for the comment of using RAID 10.
With all what others have covered, IMHO, 64GB Ram, is a bit low depending on how many VMs you are running and how large your DBS are for SQL.
No comment has been added to this question in more than 21 days, so it is now classified as abandoned.
I have recommended this question be closed as follows:
Split:
-- 'kevinhsieh' (https:#a42965770)
-- 'Philip Elder' (https:#a42965773)
-- 'Philip Elder' (https:#a42966046)
-- 'CompProbSolv' (https:#a42965986)
-- 'kevinhsieh' (https:#a42965914)
-- 'CompProbSolv' (https:#a42965796)
If you feel this question should be closed differently, post an objection and the moderators will review all objections and close it as they feel fit. If no one objects, this question will be closed automatically the way described above.
seth2740
Experts-Exchange Cleanup Volunteer
I have recommended this question be closed as follows:
Split:
-- 'kevinhsieh' (https:#a42965770)
-- 'Philip Elder' (https:#a42965773)
-- 'Philip Elder' (https:#a42966046)
-- 'CompProbSolv' (https:#a42965986)
-- 'kevinhsieh' (https:#a42965914)
-- 'CompProbSolv' (https:#a42965796)
If you feel this question should be closed differently, post an objection and the moderators will review all objections and close it as they feel fit. If no one objects, this question will be closed automatically the way described above.
seth2740
Experts-Exchange Cleanup Volunteer