Hints for different disk size arranging
Ste
125 Posts
April 3, 2020, 11:31 amQuote from Ste on April 3, 2020, 11:31 amGood mornig,
in a short time I will fill my 3 nodes with the same type of disk, so at the end I will have 3 nodes with 10 x 10TB HDD, for a total of 30 disks. As a next step, I would like to add a 4th node filled with 14 TB HDD, what would be the best choice, regarding to performance and/or cluster stability ?
- Put all the bigger 14 TB disks all in the same node, so having 3 nodes with 10 x 10 TB and 1 node with 10 x 14 TB.
- Try to distribute more or less evenly the 14 TB disks across all the nodes, let's say having 2 nodes with 8 x 10 TB + 2 x 14 TB and other 2 nodes with 7 x 10 TB + 3 x 14 TB.
- It is a bad idea to have disks of different size and I must buy other 10 TB drives.
It is more important to have uniformity inside the single node, or across all the nodes in a cluster ?
Thanks for any hint. Ste.
Good mornig,
in a short time I will fill my 3 nodes with the same type of disk, so at the end I will have 3 nodes with 10 x 10TB HDD, for a total of 30 disks. As a next step, I would like to add a 4th node filled with 14 TB HDD, what would be the best choice, regarding to performance and/or cluster stability ?
- Put all the bigger 14 TB disks all in the same node, so having 3 nodes with 10 x 10 TB and 1 node with 10 x 14 TB.
- Try to distribute more or less evenly the 14 TB disks across all the nodes, let's say having 2 nodes with 8 x 10 TB + 2 x 14 TB and other 2 nodes with 7 x 10 TB + 3 x 14 TB.
- It is a bad idea to have disks of different size and I must buy other 10 TB drives.
It is more important to have uniformity inside the single node, or across all the nodes in a cluster ?
Thanks for any hint. Ste.
admin
2,930 Posts
April 3, 2020, 4:57 pmQuote from admin on April 3, 2020, 4:57 pmAs a rule, it is better to have you hardware as consistent as possible, disks, cpu, etc
A larger disk will serve more io, so will be loaded more that the other disks, which can make it become a performance bottleneck.
If you have mis-matched disks, you also do not want to put all larger disks in 1 node, the load on that server node will be higher, so it is better to distribute the larger disks across the nodes.
In your case, if you do not have 10 TB disks, i would go for putting the 14 TB in a new node, if your cpu ot network are stressed on existing nodes, maybe do reduce the count from 10 to 8 on new node.
Option 2 looks good in theory, in practice moving existing disks around to -other nodes will cause a lot of rebalance traffic.
As a rule, it is better to have you hardware as consistent as possible, disks, cpu, etc
A larger disk will serve more io, so will be loaded more that the other disks, which can make it become a performance bottleneck.
If you have mis-matched disks, you also do not want to put all larger disks in 1 node, the load on that server node will be higher, so it is better to distribute the larger disks across the nodes.
In your case, if you do not have 10 TB disks, i would go for putting the 14 TB in a new node, if your cpu ot network are stressed on existing nodes, maybe do reduce the count from 10 to 8 on new node.
Option 2 looks good in theory, in practice moving existing disks around to -other nodes will cause a lot of rebalance traffic.
Hints for different disk size arranging
Ste
125 Posts
Quote from Ste on April 3, 2020, 11:31 amGood mornig,
in a short time I will fill my 3 nodes with the same type of disk, so at the end I will have 3 nodes with 10 x 10TB HDD, for a total of 30 disks. As a next step, I would like to add a 4th node filled with 14 TB HDD, what would be the best choice, regarding to performance and/or cluster stability ?
- Put all the bigger 14 TB disks all in the same node, so having 3 nodes with 10 x 10 TB and 1 node with 10 x 14 TB.
- Try to distribute more or less evenly the 14 TB disks across all the nodes, let's say having 2 nodes with 8 x 10 TB + 2 x 14 TB and other 2 nodes with 7 x 10 TB + 3 x 14 TB.
- It is a bad idea to have disks of different size and I must buy other 10 TB drives.
It is more important to have uniformity inside the single node, or across all the nodes in a cluster ?
Thanks for any hint. Ste.
Good mornig,
in a short time I will fill my 3 nodes with the same type of disk, so at the end I will have 3 nodes with 10 x 10TB HDD, for a total of 30 disks. As a next step, I would like to add a 4th node filled with 14 TB HDD, what would be the best choice, regarding to performance and/or cluster stability ?
- Put all the bigger 14 TB disks all in the same node, so having 3 nodes with 10 x 10 TB and 1 node with 10 x 14 TB.
- Try to distribute more or less evenly the 14 TB disks across all the nodes, let's say having 2 nodes with 8 x 10 TB + 2 x 14 TB and other 2 nodes with 7 x 10 TB + 3 x 14 TB.
- It is a bad idea to have disks of different size and I must buy other 10 TB drives.
It is more important to have uniformity inside the single node, or across all the nodes in a cluster ?
Thanks for any hint. Ste.
admin
2,930 Posts
Quote from admin on April 3, 2020, 4:57 pmAs a rule, it is better to have you hardware as consistent as possible, disks, cpu, etc
A larger disk will serve more io, so will be loaded more that the other disks, which can make it become a performance bottleneck.
If you have mis-matched disks, you also do not want to put all larger disks in 1 node, the load on that server node will be higher, so it is better to distribute the larger disks across the nodes.
In your case, if you do not have 10 TB disks, i would go for putting the 14 TB in a new node, if your cpu ot network are stressed on existing nodes, maybe do reduce the count from 10 to 8 on new node.
Option 2 looks good in theory, in practice moving existing disks around to -other nodes will cause a lot of rebalance traffic.
As a rule, it is better to have you hardware as consistent as possible, disks, cpu, etc
A larger disk will serve more io, so will be loaded more that the other disks, which can make it become a performance bottleneck.
If you have mis-matched disks, you also do not want to put all larger disks in 1 node, the load on that server node will be higher, so it is better to distribute the larger disks across the nodes.
In your case, if you do not have 10 TB disks, i would go for putting the 14 TB in a new node, if your cpu ot network are stressed on existing nodes, maybe do reduce the count from 10 to 8 on new node.
Option 2 looks good in theory, in practice moving existing disks around to -other nodes will cause a lot of rebalance traffic.