This post is more than 5 years old
9 Posts
0
15199
Combining PS-M4110E and PS-M4110X in a group with tiering
Hello!
I have two PS-M4110E and two PS-M4110X that I am going to deploy to two different locations, and I am a bit confused.
From the discussions here I have gathered that is not recommended to mix one of each, but from the TR1070 EqualLogic PS Series Architecture: Load Balancers document it states in the "Tiering with APBL" section: "When provided with tiered resources in a pool such as arrays with different spindle speeds or set to different RAID types, the APLB is able to use them to tier the workload."
My plan was to have one group at each location with one PS-M4110E and one PS-M4110X in one pool. I would then add two volumes of 10TB in each group and use one for production and one for replications from the the other site.
Question is then: what would be best to do in this situation?
Best Regards
Kim
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
1
March 31st, 2016 09:00
Hello Kim,
It depends entirely on the expected IO pattern and intended applications. An "E" model array with 7200 RPM drives that tend to be much larger than the 10K "X" model array, the "E" model ends up with more of the data. Since data is striped over members in a pool, more of the I/O load will be handled by the "E" array. APLB will work to keep the IO latency balanced. This works best when the load is sustained. If the IO spikes are very, very short, APLB efficiency is reduced.
What RAID levels were you planning on using on each member?
Re: volumes. Those are very large volumes. Is this in a virtualized environment? Generally speaking more smaller volumes perform better than one or two MEGA volumes.
Re: Replication. You need more space on the DR side to store the replicas. The min reserve percentage is 105%. Typically more than that will be needed depending on how many replicas you want to keep.
Regards,
Don
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
April 1st, 2016 09:00
Hello,
Then based on that information you can combine them if you wish.
Which hypervisor are you using?
If it's VMware ESXi, then please review the best practices docs for ESXi + EQL here:
http://en.community.dell.com/techcenter/extras/m/white_papers/20434601
Regards,
Don
Kihen
9 Posts
0
April 1st, 2016 00:00
Hello Don
Thanks for the quick reply.
IO does not tend to spike during normal business hours. I have an increased but sustained IO load at night due to backup.
My "E" models are actually smaller then the "X" modells. (9,59TB vs 11,51TB for the X models)
I was planning on using RAID6 for all. We dont have any IO intensive servers running off these arrays.
The arrays are for my virtualized environment.
Thanks for the info about replication! I will take that into consideration.
Currently Im running with the entire environment off one "E" and one "X" model with syncrep. Not ideal, but that was considered our best option at the time. Also I currently have more then enough space, so the meaning is for these arrays to cover our storage needs for the next 3-4 years.
Kihen
9 Posts
0
April 4th, 2016 04:00
Ok!
Tanks for the help!
Best regards
Kim
Anonymous
5 Practitioner
5 Practitioner
•
274.2K Posts
0
April 4th, 2016 08:00
Hello Kim,
You are very welcome.
Don