buzzzo
3 Posts
Quote from buzzzo on March 9, 2017, 8:31 pm
Hi
How many numbers of nodes do I need at minimum in order to have a working system?
Hi
How many numbers of nodes do I need at minimum in order to have a working system?
buzzzo
3 Posts
Quote from buzzzo on March 10, 2017, 9:52 am
Hi
Thx for answer.
Is it a three nodes cluster with the right amount of ram suitable for a medium workload (eg: 4 hypervisor node cluster)
Thx
Hi
Thx for answer.
Is it a three nodes cluster with the right amount of ram suitable for a medium workload (eg: 4 hypervisor node cluster)
Thx
admin
2,930 Posts
Quote from admin on March 10, 2017, 11:28 am
The best thing is to try out your workload and hardware in a test environment and see if you are happy with the performance. In many cases it will be good.
Of course if you add more disks and nodes the faster the cluster will get, that is the nice thing about scale out. The exact number of disks that you can add per node is dependent on hardware used and also io workload: if you use ssds/nvme or have small io sizes, the cpu load will saturate with a smaller number of disks. We plan to write a documentation guide for performance tuning in the future.
Also we will be releasing next week 1.2.1 release which will have a new kernel ( 4.4 instead of current 3.12 ) which has better performance and we intend to stabilize on it for a long time so maybe better to use this for your testing.
The best thing is to try out your workload and hardware in a test environment and see if you are happy with the performance. In many cases it will be good.
Of course if you add more disks and nodes the faster the cluster will get, that is the nice thing about scale out. The exact number of disks that you can add per node is dependent on hardware used and also io workload: if you use ssds/nvme or have small io sizes, the cpu load will saturate with a smaller number of disks. We plan to write a documentation guide for performance tuning in the future.
Also we will be releasing next week 1.2.1 release which will have a new kernel ( 4.4 instead of current 3.12 ) which has better performance and we intend to stabilize on it for a long time so maybe better to use this for your testing.
buzzzo
3 Posts
Quote from buzzzo on March 10, 2017, 7:35 pm
Thx for reply
Your project seems impressive.
Do you plan to add an enterprise version apart the gpl one?
Thx for reply
Your project seems impressive.
Do you plan to add an enterprise version apart the gpl one?
admin
2,930 Posts
Quote from admin on March 10, 2017, 8:56 pm
Thank you very much for your feedback.
At some point we will offer a support option but PetaSAN will remain agpl and free to use.
Thank you very much for your feedback.
At some point we will offer a support option but PetaSAN will remain agpl and free to use.
tiagocapela
1 Post
Quote from tiagocapela on May 3, 2017, 5:49 pm
Hi!
I'm trying to test it but I only have Two machines for the storage and other two for the hypervisor.
How can i install it using only two nodes?
it is possible to do it?
thanks
Hi!
I'm trying to test it but I only have Two machines for the storage and other two for the hypervisor.
How can i install it using only two nodes?
it is possible to do it?
thanks
admin
2,930 Posts
Quote from admin on May 4, 2017, 8:51 am
The min is 3 since 2 can lead to split brain conditions, it is not supported
In your test setup, you can run PetaSAN virtualized or run your clients virtualized
The min is 3 since 2 can lead to split brain conditions, it is not supported
In your test setup, you can run PetaSAN virtualized or run your clients virtualized
