Generally yes, there being more write cache available. Also there are better availabilty features with the CX4 960, better IO module Front end bandwidth, ability to use Virtual Provisioning, Autotiering and Fast cache if they were purchased.
However if there is a particular lun in an overworked raid group, it probably will be still overworked and needs more disks for the lun
My response to your inquiry is that Yes, you should.
CX4 960 has faster CPU and more Cache that CX3 80 does that mean you have more capability to process IO request and buffer more data on SP cache for better response to later IO request. Therefore overall, the performance should go to next level/
But as Gearoid mentioned, any overheated device can lower your experience, you might use Navishpere to monitor and find who they are and consider to add more spindles or EFD to solve them.
Performance improvement is expected upon my experience. Only question is how much. That would depend on your current workload and its effect on your disks.
Cache has only a secondary affect on this. It is all about how fast the disks you have for this setup
If you have FC disks you expect x performance if you have Sata disks you expect y performance
Upgrading to a 960 wont change this, it will change how IO's are being handled coming in with more cache but if you still have slow disks handling the hosts this is the bottleneck
I recommend you read the EMC Storage System Fundamentals for Performance and availablilty
I didnt really mean that if your disks are slow, but more are your disks being overworked
For example if I have 5 luns using the disks in a raid 5raid group you need to consider the percentage of reads and writes to determine the best amount of drives in a RG
>>
Another question, a lot of raiod groups have a relatively small number of disks. some have 4, some have 6 disks.
should I delete these storage groups and make one storage group from more disks?
>>
Not 100% sure what you mean here, do you mean a storagepool? Possibly but its very difficult to say and normally requires correct planning before hand
You need to look at your hosts and determine the IO required by each lun and ensure that the luns together do not overload the StoragePool/RG
That best practises doc that I posted earlier is your friend again
Multiple buses is a good and bad thing depending how you have it laid out
It can be better for performance if your Backend Bus for one bus is saturated however if you loose LCC/Enclosure etc you can end up with loads of rebuilds
Normally this kind of discussion would probably be best discussed with someone who is familar with the environment/hosts and the storage. I would possibly recommend an engagement with EMC Professional Services for this to help lay it out as well as can be managed
The number of disks is a factor, and so is the size of the data store and number of vms. Also, what version of VMware are you using? Are you having response time issues on the storage group or array?
dynamox
9 Legend
•
20.4K Posts
0
May 19th, 2012 21:00
what else did you upgrade other then brand new SPs ?
tdubb123
2 Intern
•
470 Posts
0
May 19th, 2012 22:00
just the sps
GearoidG
251 Posts
0
May 20th, 2012 01:00
The short answer is it depends...
Generally yes, there being more write cache available. Also there are better availabilty features with the CX4 960, better IO module Front end bandwidth, ability to use Virtual Provisioning, Autotiering and Fast cache if they were purchased.
However if there is a particular lun in an overworked raid group, it probably will be still overworked and needs more disks for the lun
So it depends
Gearoid
reseach
2 Intern
•
225 Posts
0
May 20th, 2012 18:00
My response to your inquiry is that Yes, you should.
CX4 960 has faster CPU and more Cache that CX3 80 does that mean you have more capability to process IO request and buffer more data on SP cache for better response to later IO request. Therefore overall, the performance should go to next level/
But as Gearoid mentioned, any overheated device can lower your experience, you might use Navishpere to monitor and find who they are and consider to add more spindles or EFD to solve them.
tdubb123
2 Intern
•
470 Posts
0
May 20th, 2012 19:00
but i am still using the same raid group and disks.
reseach
2 Intern
•
225 Posts
0
May 20th, 2012 20:00
Performance improvement is expected upon my experience. Only question is how much. That would depend on your current workload and its effect on your disks.
ylinn
3 Posts
0
May 20th, 2012 21:00
There would be a little performance improvement due to more Cache memory.
GearoidG
251 Posts
0
May 21st, 2012 09:00
Cache has only a secondary affect on this. It is all about how fast the disks you have for this setup
If you have FC disks you expect x performance if you have Sata disks you expect y performance
Upgrading to a 960 wont change this, it will change how IO's are being handled coming in with more cache but if you still have slow disks handling the hosts this is the bottleneck
I recommend you read the EMC Storage System Fundamentals for Performance and availablilty
https://support.emc.com/docu32834_White-Paper:-EMC-CLARiiON-Storage-System-Fundamentals-for-Performance-and-Availability-%E2%80%94-Applied-Best-Practices.pdf
GearoidG
251 Posts
0
May 21st, 2012 12:00
I didnt really mean that if your disks are slow, but more are your disks being overworked
For example if I have 5 luns using the disks in a raid 5raid group you need to consider the percentage of reads and writes to determine the best amount of drives in a RG
>>
Another question, a lot of raiod groups have a relatively small number of disks. some have 4, some have 6 disks.
should I delete these storage groups and make one storage group from more disks?
>>
Not 100% sure what you mean here, do you mean a storagepool? Possibly but its very difficult to say and normally requires correct planning before hand
You need to look at your hosts and determine the IO required by each lun and ensure that the luns together do not overload the StoragePool/RG
That best practises doc that I posted earlier is your friend again
Multiple buses is a good and bad thing depending how you have it laid out
It can be better for performance if your Backend Bus for one bus is saturated however if you loose LCC/Enclosure etc you can end up with loads of rebuilds
Normally this kind of discussion would probably be best discussed with someone who is familar with the environment/hosts and the storage. I would possibly recommend an engagement with EMC Professional Services for this to help lay it out as well as can be managed
Hope this helps
Gearoid
tdubb123
2 Intern
•
470 Posts
0
May 21st, 2012 12:00
i dont think my disks are slow.
They are mostly sata 1TB drives and some fc 300gb drives.
another question, a lot of raiod groups have a relatively small number of disks. some have 4, some have 6 disks.
should I delete these storage groups and make one storage group from more disks?
does it matter if the raid group is on multiple buses?
cbshaw1
3 Posts
0
May 21st, 2012 12:00
The number of disks is a factor, and so is the size of the data store and number of vms. Also, what version of VMware are you using? Are you having response time issues on the storage group or array?
tdubb123
2 Intern
•
470 Posts
0
May 21st, 2012 13:00
i tried to do some analyze on navisphere but it says logging is disabled? I am using esxi5
cbshaw1
3 Posts
0
May 21st, 2012 14:00
You will need to enable logging and then you can get the performance information. It's good that you are on ESX 5.
Craig Bradshaw
tdubb123
2 Intern
•
470 Posts
0
May 21st, 2012 15:00
how to enable it?
cbshaw1
3 Posts
0
May 21st, 2012 15:00
Here is a link to another thread that has the detail:
https://community.emc.com/thread/124096
Thanks,
Craig Bradshaw