How to calculate disk speed/capacity. (iSCSI)
ihyuner
1 Post
February 9, 2022, 2:26 pmQuote from ihyuner on February 9, 2022, 2:26 pmHello
Server 1:
SATA SSD 256GB (OS)
HDD 2TB * 24
NVMe SSD 1TB * 3
eth1 : 1Gbps LAN
eth2 : 40Gbps LAN
CPU : E5-2690v2
RAM : 32GB
Server 2:
SATA SSD 256GB (OS)
HDD 2TB * 24
NVMe SSD 1TB * 3
eth1 : 1Gbps LAN
eth2 : 40Gbps LAN
CPU : E5-2690v2
RAM : 32GB
All disks are individually composed of raid 0.
I have a question when using PetaSan in such a system configuration.
1. Can eth1 be used as a management LAN card and eth2 as a data transmission (storage line) LAN card?
2. HDD and NVMe are installed at a ratio of 8:1 per server, so is that NVMe available only for 4k cache on HDD?
If not, I wonder if it can be used as a cache to raise iops.
3. This third question is the core of this post.
How do you calculate the speed of the disk available in that configuration?
For example, assuming that each hard disk has a read 200MB/s write 150MB/s performance,
If you configure the 24 disks raid0, we will get read 4800MB/s write 3600MB/s (200MB/s*24 & 150MB/s*24).
However, I will not use raid and plan to use pure PetaSAN only.
I wonder how to calculate the read/write speed and the available disk usage of the disk that PetaSAN can release when communicating from PetaSAN to CSI.
I'd really appreciate it if you could answer.
Hello
Server 1:
SATA SSD 256GB (OS)
HDD 2TB * 24
NVMe SSD 1TB * 3
eth1 : 1Gbps LAN
eth2 : 40Gbps LAN
CPU : E5-2690v2
RAM : 32GB
Server 2:
SATA SSD 256GB (OS)
HDD 2TB * 24
NVMe SSD 1TB * 3
eth1 : 1Gbps LAN
eth2 : 40Gbps LAN
CPU : E5-2690v2
RAM : 32GB
All disks are individually composed of raid 0.
I have a question when using PetaSan in such a system configuration.
1. Can eth1 be used as a management LAN card and eth2 as a data transmission (storage line) LAN card?
2. HDD and NVMe are installed at a ratio of 8:1 per server, so is that NVMe available only for 4k cache on HDD?
If not, I wonder if it can be used as a cache to raise iops.
3. This third question is the core of this post.
How do you calculate the speed of the disk available in that configuration?
For example, assuming that each hard disk has a read 200MB/s write 150MB/s performance,
If you configure the 24 disks raid0, we will get read 4800MB/s write 3600MB/s (200MB/s*24 & 150MB/s*24).
However, I will not use raid and plan to use pure PetaSAN only.
I wonder how to calculate the read/write speed and the available disk usage of the disk that PetaSAN can release when communicating from PetaSAN to CSI.
I'd really appreciate it if you could answer.
Last edited on February 9, 2022, 2:28 pm by ihyuner · #1
admin
2,921 Posts
February 10, 2022, 7:23 pmQuote from admin on February 10, 2022, 7:23 pmPetaSAN is a complex distributed system, its speed is not merely computed by adding speed of your drives.
Even the hardisks you mention have 200 MB/s, well this is just if you write large blocks like 4 MB in size, if you write small 4 KB you will likely get 0.5 MB/s so it is a factor of x400 depending on what you write, so it is also dependent on you workload, its block size and io pattern random/sequential. Performance also depends on how many concurrent i/o operation : are you testing the entire cluster with a single file copy operation or do you have many concurrent operations (other file copies, many vms in parallel..etc)
There is a benchmark page in the ui, your best bet is to test a 3 node test cluster and see what performance you get using different benchmark tests available from that page. PetaSAN is a scale out system, so the more nodes you add the more total iops/throughput performance you will get.
PetaSAN is a complex distributed system, its speed is not merely computed by adding speed of your drives.
Even the hardisks you mention have 200 MB/s, well this is just if you write large blocks like 4 MB in size, if you write small 4 KB you will likely get 0.5 MB/s so it is a factor of x400 depending on what you write, so it is also dependent on you workload, its block size and io pattern random/sequential. Performance also depends on how many concurrent i/o operation : are you testing the entire cluster with a single file copy operation or do you have many concurrent operations (other file copies, many vms in parallel..etc)
There is a benchmark page in the ui, your best bet is to test a 3 node test cluster and see what performance you get using different benchmark tests available from that page. PetaSAN is a scale out system, so the more nodes you add the more total iops/throughput performance you will get.
How to calculate disk speed/capacity. (iSCSI)
ihyuner
1 Post
Quote from ihyuner on February 9, 2022, 2:26 pmHello
Server 1:
SATA SSD 256GB (OS)
HDD 2TB * 24
NVMe SSD 1TB * 3
eth1 : 1Gbps LAN
eth2 : 40Gbps LAN
CPU : E5-2690v2
RAM : 32GBServer 2:
SATA SSD 256GB (OS)
HDD 2TB * 24
NVMe SSD 1TB * 3
eth1 : 1Gbps LAN
eth2 : 40Gbps LAN
CPU : E5-2690v2
RAM : 32GBAll disks are individually composed of raid 0.
I have a question when using PetaSan in such a system configuration.
1. Can eth1 be used as a management LAN card and eth2 as a data transmission (storage line) LAN card?
2. HDD and NVMe are installed at a ratio of 8:1 per server, so is that NVMe available only for 4k cache on HDD?
If not, I wonder if it can be used as a cache to raise iops.3. This third question is the core of this post.
How do you calculate the speed of the disk available in that configuration?
For example, assuming that each hard disk has a read 200MB/s write 150MB/s performance,
If you configure the 24 disks raid0, we will get read 4800MB/s write 3600MB/s (200MB/s*24 & 150MB/s*24).However, I will not use raid and plan to use pure PetaSAN only.
I wonder how to calculate the read/write speed and the available disk usage of the disk that PetaSAN can release when communicating from PetaSAN to CSI.
I'd really appreciate it if you could answer.
Hello
Server 1:
SATA SSD 256GB (OS)
HDD 2TB * 24
NVMe SSD 1TB * 3
eth1 : 1Gbps LAN
eth2 : 40Gbps LAN
CPU : E5-2690v2
RAM : 32GB
Server 2:
SATA SSD 256GB (OS)
HDD 2TB * 24
NVMe SSD 1TB * 3
eth1 : 1Gbps LAN
eth2 : 40Gbps LAN
CPU : E5-2690v2
RAM : 32GB
All disks are individually composed of raid 0.
I have a question when using PetaSan in such a system configuration.
1. Can eth1 be used as a management LAN card and eth2 as a data transmission (storage line) LAN card?
2. HDD and NVMe are installed at a ratio of 8:1 per server, so is that NVMe available only for 4k cache on HDD?
If not, I wonder if it can be used as a cache to raise iops.
3. This third question is the core of this post.
How do you calculate the speed of the disk available in that configuration?
For example, assuming that each hard disk has a read 200MB/s write 150MB/s performance,
If you configure the 24 disks raid0, we will get read 4800MB/s write 3600MB/s (200MB/s*24 & 150MB/s*24).
However, I will not use raid and plan to use pure PetaSAN only.
I wonder how to calculate the read/write speed and the available disk usage of the disk that PetaSAN can release when communicating from PetaSAN to CSI.
I'd really appreciate it if you could answer.
admin
2,921 Posts
Quote from admin on February 10, 2022, 7:23 pmPetaSAN is a complex distributed system, its speed is not merely computed by adding speed of your drives.
Even the hardisks you mention have 200 MB/s, well this is just if you write large blocks like 4 MB in size, if you write small 4 KB you will likely get 0.5 MB/s so it is a factor of x400 depending on what you write, so it is also dependent on you workload, its block size and io pattern random/sequential. Performance also depends on how many concurrent i/o operation : are you testing the entire cluster with a single file copy operation or do you have many concurrent operations (other file copies, many vms in parallel..etc)
There is a benchmark page in the ui, your best bet is to test a 3 node test cluster and see what performance you get using different benchmark tests available from that page. PetaSAN is a scale out system, so the more nodes you add the more total iops/throughput performance you will get.
PetaSAN is a complex distributed system, its speed is not merely computed by adding speed of your drives.
Even the hardisks you mention have 200 MB/s, well this is just if you write large blocks like 4 MB in size, if you write small 4 KB you will likely get 0.5 MB/s so it is a factor of x400 depending on what you write, so it is also dependent on you workload, its block size and io pattern random/sequential. Performance also depends on how many concurrent i/o operation : are you testing the entire cluster with a single file copy operation or do you have many concurrent operations (other file copies, many vms in parallel..etc)
There is a benchmark page in the ui, your best bet is to test a 3 node test cluster and see what performance you get using different benchmark tests available from that page. PetaSAN is a scale out system, so the more nodes you add the more total iops/throughput performance you will get.