Attachment uploads are currently disabled. This is a temporary situation and will resume as normal in the coming days.
Solved!
Go to Solution- This is a part of the cache or a slot in cache ?
There is a part of the cache or a some slot reserved to a DA ?
What this means if the Prefetched Tracks activity is all the time equal to 0 ?
Can I undrestant that data are not sequential and are not present in cache ? - This is a part of the cache or a slot in cache ?
When a DA reads data, it must be placed in cache, random sequential or prefetched.
There is a part of the cache or a some slot reserved to a DA ?
Cache isn't reserved. It is just like any other read, just done ahead of the host actually requesting this data.
What this means if the Prefetched Tracks activity is all the time equal to 0 ?
Then there is no reads being done ahead of the host requesting them.
Can I understand that data are not sequential and are not present in cache ?
All data written and read goes into cache on Symmetrix. Random reads actually stay in cache longer than prefetched data. Prefetched data is usually sequential which is less likely to be re-read right away, so the Symmetrix uses a small circular buffer for prefetch activity. - If I undrestand no prefetched activity means no sequantial access, but data can be in cache or on disk.
I have a stupid question :
if the cache is not reserved to any part of the FA or DA what happen if I have many host on a particular port on a particular FA ( like FA 8 and FA 9 ) with many mapped and masked disk ( more than 300 disks ) ?
Thanks - All devices (many masked devices) share the same cache. The Symmetrix will decide what is the best cache data to chose for replacement should a new slot be needed from available slots. This is usually (but not always) older, more stale data. Dynamic Cache Partitioning can be used to allocate cache to some set of devices.
Quincy561
1281 Posts
495
0
Posted October 21st, 2009 13:00
Cache in the Symmetrix is shared by all devices. The cache algorithms tries to keep the data most likely to be reused in cache. If a read request comes in from a host, and the data is not in cache, the data least likely (usually oldest) is removed and the new data is read into cache. With 72 code we now have a feature called Dynamic Cache Partitioning (DCP) that can fence or reserve a portion of cache for specific devices.