Caches Write I/Os in flash in addition to Read I/Os
Highly available implementation
Requires only dirty blocks in flash to be recovered on flash failure as opposed to all disks with cached data
Teradata, EMC, Hitachi, etc. use flash tiering
Exadata Flash Cache adapts much faster to changing workload
Each I/O changes content of cache
If a new data is created it can be instantly cached
With tiering, data is slowly migrated from disk based on historical statistics
Flash tiering caches yesterday’s hot data, not necessarily today’s
Exadata Cache has much finer granularity of flash contents
Caching at 64K block level for flash cache, around 1MB for flash tiering
Caching is much more efficient at capturing hot data in flash while leaving cold data on disk. Multiplies the effective capacity of flash.
Exadata Cache doesn’t need to mirror data in flash
Can keep one mirror copy in flash to speed up reads while having the other mirror copy on disk
Flash cache can keep hot blocks cached forever
There is no need to ever de-stage them to disk. This is not a tiering advantage
All the existing Exadata flash advantages also apply
Scale-out architecture
InfiniBand 40Gb/sec connectivity
Flash PCI Cards are much faster than flash disk
Smart Scans
Smart Flash Logs
Smart Caching
Flash is shared across servers and works with RAC
Unlike server flash cards
Optionally Keep specific tables/indexes/partitions in flash with simple command
Scans run simultaneously on disk and flash and aggregate throughput
EHCC compression enhances flash capacity
Benefits from storage index
Decompression and decryption in cell CPUs during scans
In Netezza and Teradata the application must encrypt and decrypt individual column values explicitly on every SQL statement