In VNX/Clariion environment we got Active/Passive ( SPA/SPB or Primary DM/Seconday DM ) mode operation, when Primary fails, Standby takes control of all luns/FS and process IO accordingly..
But in VMAX/DMX architecture why do we have Active/Active mode operation to serve workloads ?
Friend of mine referenced , that Cache plays a big role in VMAX/DMX architecture to serve IOPS. Where Cache has set of tables for each THIN device and performs mapping between Frond end IO modules and Back end IO modules. I am not sure if that is the method followed to serve IOPS in VMAX/DMX. Can experts clarify ?
Solved! Go to Solution.
A device in VMAX can be accessed by any director in the system at any time. We can have up to 16 director boards in the system and a device can be accessed from all 16 at the same time.
The cache is global, and accessible by all modules, FE, BE, RDF, etc.
All writes go to cache in VMAX. A read that can be satisfied by the cache will just be accessed by the front-end, the backend won't even know about that IO.
Thanks for your explanation Quincy56,
So you mean Cache slots reference table serves all write and read operations and it checks for the configuration database for the list of the mapped FA ports for that pa and then pushes the data through them to the Host .
Correct me if I am wrong
The user cache on VMAX systems is usually pretty large, and all of it can be used for reads. It uses an LRU algorithm to keep frequently read data in cache so we can get read hits even on random IO.
Configuration information is also kept in cache (meta data) separate from the user data, but both areas are accessible by all directors.
There is more detail in the product guide. The older guides had a lot more detail on IO flow, but a quick search didn't come up with the old guides.
Chapter 3 of this DMX3 manual has the detail on IO flow, which is pretty much the same as VMAX at the high level.