Petasan for Backup Storage
merlinios
9 Posts
April 4, 2018, 9:06 pmQuote from merlinios on April 4, 2018, 9:06 pmHello ,
Congratulations for the whole project.
Im planning to create a storage cluster for backing up virtual machines and files using Microsoft DPM Server . I want to present through iSCSI to the DPM server storage from Petasan. My concerns have to do with performance because of the heavy writes . As i told this is going to be a backup storage so i want to use cheap hardware and spinable 7200 Sata disks (10-12 6tb per node) . Im planning to use refurbished HP DL server and maybe change the controller card with a capable of doing jbod or use the one it has with raid-o per disk . Minimum server will be 3 with plans to expand as our storage needs grow and also as performance needs grow .
Do you thing that this scenario is going to work in terms of performance ? Do i need journal ssd disks ? Any help on this will be much appreciated
Hello ,
Congratulations for the whole project.
Im planning to create a storage cluster for backing up virtual machines and files using Microsoft DPM Server . I want to present through iSCSI to the DPM server storage from Petasan. My concerns have to do with performance because of the heavy writes . As i told this is going to be a backup storage so i want to use cheap hardware and spinable 7200 Sata disks (10-12 6tb per node) . Im planning to use refurbished HP DL server and maybe change the controller card with a capable of doing jbod or use the one it has with raid-o per disk . Minimum server will be 3 with plans to expand as our storage needs grow and also as performance needs grow .
Do you thing that this scenario is going to work in terms of performance ? Do i need journal ssd disks ? Any help on this will be much appreciated
admin
2,930 Posts
April 5, 2018, 11:27 amQuote from admin on April 5, 2018, 11:27 amIf you can get the raid controller with write back cache it will make a large impact. The ssd journal will double your write speed.
The more (spinning) disks you add, it will increase capacity as well as performance since spinning disks are usually the bottleneck. I would recommend you first try with 3-4 of spinning disks per node without ssd journal, and deduce if you have 12 /24/36 will the performance be good for your needs: the single io write performance will be the same but the total write performance should scale linearly..so if you have multiple write jobs it will be scale. Else you will need write back cache and/or ssd journal.
If you can get the raid controller with write back cache it will make a large impact. The ssd journal will double your write speed.
The more (spinning) disks you add, it will increase capacity as well as performance since spinning disks are usually the bottleneck. I would recommend you first try with 3-4 of spinning disks per node without ssd journal, and deduce if you have 12 /24/36 will the performance be good for your needs: the single io write performance will be the same but the total write performance should scale linearly..so if you have multiple write jobs it will be scale. Else you will need write back cache and/or ssd journal.
merlinios
9 Posts
April 10, 2018, 11:52 amQuote from merlinios on April 10, 2018, 11:52 amThanks a lot for your answer ! I will start as you said with 3-4 spinning disks for testing .
Thanks a lot for your answer ! I will start as you said with 3-4 spinning disks for testing .
Petasan for Backup Storage
merlinios
9 Posts
Quote from merlinios on April 4, 2018, 9:06 pmHello ,
Congratulations for the whole project.
Im planning to create a storage cluster for backing up virtual machines and files using Microsoft DPM Server . I want to present through iSCSI to the DPM server storage from Petasan. My concerns have to do with performance because of the heavy writes . As i told this is going to be a backup storage so i want to use cheap hardware and spinable 7200 Sata disks (10-12 6tb per node) . Im planning to use refurbished HP DL server and maybe change the controller card with a capable of doing jbod or use the one it has with raid-o per disk . Minimum server will be 3 with plans to expand as our storage needs grow and also as performance needs grow .
Do you thing that this scenario is going to work in terms of performance ? Do i need journal ssd disks ? Any help on this will be much appreciated
Hello ,
Congratulations for the whole project.
Im planning to create a storage cluster for backing up virtual machines and files using Microsoft DPM Server . I want to present through iSCSI to the DPM server storage from Petasan. My concerns have to do with performance because of the heavy writes . As i told this is going to be a backup storage so i want to use cheap hardware and spinable 7200 Sata disks (10-12 6tb per node) . Im planning to use refurbished HP DL server and maybe change the controller card with a capable of doing jbod or use the one it has with raid-o per disk . Minimum server will be 3 with plans to expand as our storage needs grow and also as performance needs grow .
Do you thing that this scenario is going to work in terms of performance ? Do i need journal ssd disks ? Any help on this will be much appreciated
admin
2,930 Posts
Quote from admin on April 5, 2018, 11:27 amIf you can get the raid controller with write back cache it will make a large impact. The ssd journal will double your write speed.
The more (spinning) disks you add, it will increase capacity as well as performance since spinning disks are usually the bottleneck. I would recommend you first try with 3-4 of spinning disks per node without ssd journal, and deduce if you have 12 /24/36 will the performance be good for your needs: the single io write performance will be the same but the total write performance should scale linearly..so if you have multiple write jobs it will be scale. Else you will need write back cache and/or ssd journal.
If you can get the raid controller with write back cache it will make a large impact. The ssd journal will double your write speed.
The more (spinning) disks you add, it will increase capacity as well as performance since spinning disks are usually the bottleneck. I would recommend you first try with 3-4 of spinning disks per node without ssd journal, and deduce if you have 12 /24/36 will the performance be good for your needs: the single io write performance will be the same but the total write performance should scale linearly..so if you have multiple write jobs it will be scale. Else you will need write back cache and/or ssd journal.
merlinios
9 Posts
Quote from merlinios on April 10, 2018, 11:52 amThanks a lot for your answer ! I will start as you said with 3-4 spinning disks for testing .
Thanks a lot for your answer ! I will start as you said with 3-4 spinning disks for testing .