Erasure coding or. replica
msalem
87 Posts
June 28, 2018, 1:41 pmQuote from msalem on June 28, 2018, 1:41 pmHey Admin,
We have setup 15-30 Disks, with High end hardware and 3 way replica.
we have total of 320TB ( 10x8TB Disks OSD + 2TB SSD) .. we got now 291TB of space.
We are wondering if the current config uses Erasure coding or replica, if replica it should be 3/rd the space. ?
Thanks
Hey Admin,
We have setup 15-30 Disks, with High end hardware and 3 way replica.
we have total of 320TB ( 10x8TB Disks OSD + 2TB SSD) .. we got now 291TB of space.
We are wondering if the current config uses Erasure coding or replica, if replica it should be 3/rd the space. ?
Thanks
admin
2,930 Posts
June 28, 2018, 3:39 pmQuote from admin on June 28, 2018, 3:39 pmThe current configuration uses replicas, a 3x replica will give you a net storage of 1/3 of your raw capacity. You can enable inline compression which will enhance this.
We will be supporting EC pools in v2.2, the current kernel does not support it.
EC does carry cpu and iops overhead which could be suitable for cold storage but may not be preferable for latency sensitive applications like vms & databases.
The current configuration uses replicas, a 3x replica will give you a net storage of 1/3 of your raw capacity. You can enable inline compression which will enhance this.
We will be supporting EC pools in v2.2, the current kernel does not support it.
EC does carry cpu and iops overhead which could be suitable for cold storage but may not be preferable for latency sensitive applications like vms & databases.
Last edited on June 28, 2018, 3:44 pm by admin · #2
msalem
87 Posts
June 28, 2018, 3:52 pmQuote from msalem on June 28, 2018, 3:52 pmThanks Admin.
So when is 2.2 release date is ?
and how is it possible that raw 320TB gives me 291TB, it should be around 100TB the most ?
Thanks
Thanks Admin.
So when is 2.2 release date is ?
and how is it possible that raw 320TB gives me 291TB, it should be around 100TB the most ?
Thanks
admin
2,930 Posts
June 28, 2018, 4:30 pmQuote from admin on June 28, 2018, 4:30 pmWe are targeting start of Sep. for 2.2
The dashboard charts showing 291 TB : this is a raw available. so if you write 1 TB, the raw available will be 288
This raw value is a cluster wide value, in v 2.1 you can create different pools with different replica counts for each, the dashboard will show you total (cluster wide) raw used / available.
We are targeting start of Sep. for 2.2
The dashboard charts showing 291 TB : this is a raw available. so if you write 1 TB, the raw available will be 288
This raw value is a cluster wide value, in v 2.1 you can create different pools with different replica counts for each, the dashboard will show you total (cluster wide) raw used / available.
Last edited on June 28, 2018, 4:30 pm by admin · #4
shadowlin
67 Posts
July 30, 2018, 2:05 amQuote from shadowlin on July 30, 2018, 2:05 am
Quote from admin on June 28, 2018, 3:39 pm
The current configuration uses replicas, a 3x replica will give you a net storage of 1/3 of your raw capacity. You can enable inline compression which will enhance this.
We will be supporting EC pools in v2.2, the current kernel does not support it.
EC does carry cpu and iops overhead which could be suitable for cold storage but may not be preferable for latency sensitive applications like vms & databases.
I just did some test on v2.0. It seems I can use targetcli to create iscsi target wtih ec rbd backstore. Dose that mean the current kernel can support ec rbd?
The performance is really bad though.
Quote from admin on June 28, 2018, 3:39 pm
The current configuration uses replicas, a 3x replica will give you a net storage of 1/3 of your raw capacity. You can enable inline compression which will enhance this.
We will be supporting EC pools in v2.2, the current kernel does not support it.
EC does carry cpu and iops overhead which could be suitable for cold storage but may not be preferable for latency sensitive applications like vms & databases.
I just did some test on v2.0. It seems I can use targetcli to create iscsi target wtih ec rbd backstore. Dose that mean the current kernel can support ec rbd?
The performance is really bad though.
Erasure coding or. replica
msalem
87 Posts
Quote from msalem on June 28, 2018, 1:41 pmHey Admin,
We have setup 15-30 Disks, with High end hardware and 3 way replica.
we have total of 320TB ( 10x8TB Disks OSD + 2TB SSD) .. we got now 291TB of space.
We are wondering if the current config uses Erasure coding or replica, if replica it should be 3/rd the space. ?
Thanks
Hey Admin,
We have setup 15-30 Disks, with High end hardware and 3 way replica.
we have total of 320TB ( 10x8TB Disks OSD + 2TB SSD) .. we got now 291TB of space.
We are wondering if the current config uses Erasure coding or replica, if replica it should be 3/rd the space. ?
Thanks
admin
2,930 Posts
Quote from admin on June 28, 2018, 3:39 pmThe current configuration uses replicas, a 3x replica will give you a net storage of 1/3 of your raw capacity. You can enable inline compression which will enhance this.
We will be supporting EC pools in v2.2, the current kernel does not support it.
EC does carry cpu and iops overhead which could be suitable for cold storage but may not be preferable for latency sensitive applications like vms & databases.
The current configuration uses replicas, a 3x replica will give you a net storage of 1/3 of your raw capacity. You can enable inline compression which will enhance this.
We will be supporting EC pools in v2.2, the current kernel does not support it.
EC does carry cpu and iops overhead which could be suitable for cold storage but may not be preferable for latency sensitive applications like vms & databases.
msalem
87 Posts
Quote from msalem on June 28, 2018, 3:52 pmThanks Admin.
So when is 2.2 release date is ?
and how is it possible that raw 320TB gives me 291TB, it should be around 100TB the most ?
Thanks
Thanks Admin.
So when is 2.2 release date is ?
and how is it possible that raw 320TB gives me 291TB, it should be around 100TB the most ?
Thanks
admin
2,930 Posts
Quote from admin on June 28, 2018, 4:30 pmWe are targeting start of Sep. for 2.2
The dashboard charts showing 291 TB : this is a raw available. so if you write 1 TB, the raw available will be 288
This raw value is a cluster wide value, in v 2.1 you can create different pools with different replica counts for each, the dashboard will show you total (cluster wide) raw used / available.
We are targeting start of Sep. for 2.2
The dashboard charts showing 291 TB : this is a raw available. so if you write 1 TB, the raw available will be 288
This raw value is a cluster wide value, in v 2.1 you can create different pools with different replica counts for each, the dashboard will show you total (cluster wide) raw used / available.
shadowlin
67 Posts
Quote from shadowlin on July 30, 2018, 2:05 amQuote from admin on June 28, 2018, 3:39 pmThe current configuration uses replicas, a 3x replica will give you a net storage of 1/3 of your raw capacity. You can enable inline compression which will enhance this.
We will be supporting EC pools in v2.2, the current kernel does not support it.
EC does carry cpu and iops overhead which could be suitable for cold storage but may not be preferable for latency sensitive applications like vms & databases.
I just did some test on v2.0. It seems I can use targetcli to create iscsi target wtih ec rbd backstore. Dose that mean the current kernel can support ec rbd?
The performance is really bad though.
Quote from admin on June 28, 2018, 3:39 pmThe current configuration uses replicas, a 3x replica will give you a net storage of 1/3 of your raw capacity. You can enable inline compression which will enhance this.
We will be supporting EC pools in v2.2, the current kernel does not support it.
EC does carry cpu and iops overhead which could be suitable for cold storage but may not be preferable for latency sensitive applications like vms & databases.
I just did some test on v2.0. It seems I can use targetcli to create iscsi target wtih ec rbd backstore. Dose that mean the current kernel can support ec rbd?
The performance is really bad though.