Unsolved
This post is more than 5 years old
666 Posts
0
11484
Ask the Expert: All about VMAX performance best practices
|
|
Ask The Expert: Connectrix FOS 7.3 – Fabric Vision Proactive Capabilities |
Welcome to this EMC Support Community Ask the Expert conversation.This ATE session focuses on VMAX best practices. Within the VMAX product there are many options to consider when allocating hosts to an array. Depending on workload profile, host setup etc. it is critical to get the right setup in order to optimize performance. It is also key to use tools like Unisphere to monitor that workload to ensure best performance. The key thing is to specifically ensure that the FA (front end) and Lun configuration are optimum.
Your host:
Kevin Gleeson is a Technical Account Manager at EMC.
He has been with EMC for the last 18 years. Kevin has spent 15 years supporting EMC Symmetrix product range at many levels. For the last 5 years he has worked in the Symmetrix Level 2 group specializing in performance. In 2012 he moved to the role Technical Account Manager supporting an EMC Elite Customer. Kevin was also recently named as a member of EMC’s Technical Leadership Academy.
This discussion begins on March 24 and concludes on April 11. Get ready by following this page to receive updates in your activity stream or through email.
kgleeson
62 Posts
0
April 9th, 2014 04:00
The detailed product guide explaining the Virtual fabrics. Data compression etc is the one I am looking for. The recent version is very high-level and lacks details.
I will see if i can find a link to a doc that will cover that and share it with you
Raman_Chidirala
5 Posts
0
April 10th, 2014 08:00
Thanks, I have the older versions. I was looking to see the improvements on the 40K etc.
Raman_Chidirala
5 Posts
0
April 15th, 2014 08:00
Kevin,
I got this doc already. But this does not have the details like the previous ones. For example it does not show me the 4 Virutal matrices and the connectivity etc.
this appears more like condensed doc with only summary of most of the internals topics but not details.
Thanks,
Raman
kgleeson
62 Posts
0
April 15th, 2014 08:00
Apologies for the delayed reply please review link below
https://support.emc.com/docu47591_Symmetrix-VMAX-Family-Product-Guide-(Symmetrix-VMAX-10K-(SN-xxx987xxxx),-VMAX-20K,-VMAX-40K).pdf?language=en_UShttps://support.emc.com/docu47591_Symmetrix-VMAX-Family-Product-Guide-(Symmetrix-VMAX-10K-(SN-xxx987xxxx),-VMAX-20K,-VMAX-40K).pdf?language=en_US
kgleeson
62 Posts
0
April 15th, 2014 08:00
Hi Raman
let me do some further digging for you.
Thanks
kevin
kach1947_custom
24 Posts
0
April 22nd, 2014 08:00
-
Assuming the discussion is still open
-
Hi Kevin,
I remember the last Enguinuity upgrade had some kind of (zap?) e-pack fix for the forced stall in the I/O on VMAX 20K and 40K. Asking especially since some cases, SRDF and/or backup are all block based and happen at specific times of the day and could impact the BE disks (going through multiple spindles depending on workload and data size) as well as FE ports causing the bottleneck.
Is that something that could be fixed or is potentially something that could be alleviated by FAST/ VP?
Any specific recommendations, best practices and or any considerations that might need to be looked out for in terms of provisioning and allocation from the array perspective ?
Thanks,
KC
kgleeson
62 Posts
0
April 30th, 2014 03:00
Hi KC
sorry for the delayed response,
The issue you refer is no longer present on the VMAX systems asumming the arrays are at 5876.268 code which is EMC target which includes all the lattest fixes for VMAX. This is also a procative EPACK for 5876.268 code which addresses a recent ETA relating to ESX host.
Can fast help with the spikes.?? possibly yes. in that if it see regular and predictable occurences of these spikes, it will wherre possible insert the data the correct pool in order to improve response times.
There are some best practice docs available for FAST VP.
Start by Binding to FC ... Ensure that the PRC for FC is then set to 10% .
you can have multiple poilicies depending on requirement.
As regards general best practice . i strongly recommend spreading the workload onto all configured FE directors as evenly as possible.
If devices are synchronous SRDF use Striped Meta's to ensure better write response times.
Ensure that Back end configurations have equal number of disks PER DAF directors.
any specific questions please come back to me
again apolpgies for the delayed reply
thanks
kevin
RobertoAraujo1
718 Posts
0
January 14th, 2015 14:00
FYI everyone, currently Kevin Gleeson is answering question on a new ATE event titled: Ask the Expert: VMAX FAST VP and Performance. Don't miss it!
mallilinux
7 Posts
0
October 28th, 2015 02:00
hi,
what exactly reserved capacity means in vmax3? is that PRC value which means its reserved for the new host writes not for any FAST activities?
Here is CLI output:
# symcfg -sid 123 list -srp -detail -v
Symmetrix ID : 123
Name : SRP_1
Description :
Default SRP : FBA
Usable Capacity (GB) : 48401.7
Allocated Capacity (GB) : 1020.5
Free Capacity (GB) : 47381.2
Subscribed Capacity (GB) : 3130.5
Subscribed Capacity (%) : 6
Reserved Capacity (%) : 10
Usable by RDFA DSE : Yes
SRDF DSE Allocated (GB) : 0.0
Thanks,
Mallik.