EMC CLARiiON – Stuff I’ve learnt

  • FLARE 30.5 adds support for FCoE
  • FAST Cache (also called Flash Cache depending where you look) – allows you to use solid state disks (SSD/flash disk) for write cache, up to 2TB. Because this cache is stored on flash disks in the event of power loss the data in the cache is inherantly safe, and does not need to be moved to the Vault disks as with the SP-based RAM write cache. I suppose it’s kind of giving your Clariion LUNS a performance boost in a similar way to how the newer hybrid SATA disks work. Can be enabled per Traditional LUN or per Storage Pool. Needs a licence (per Clariion).
  • The storage processors themselves actually run Windows Server 2003 x64 (though you’d never know it).
  • FAST moves data blocks between different types of physical disk based on their usage, so active blocks end up on faster disks, inactive or less active blocks end up on slower disks. Can only be enabled for LUNs within Storage Pools, not RAID Groups.
  • You can mix Fibre and SSD within a DAE
  • Auto Assign in the Ownership section of the LUN properties should be used for LUNs accessed by hosts that do not have Powerpath installed, e.g. vSphere hosts without Powerpath/VE, but should not be used for LUNs accessed by hosts that do have Powerpath installed.
  • Hosts can only be a member of one Storage Group at a time, but LUNs can be a member of many. This has implications for boot from SAN for vSphere hosts, as instead of having one Storage group with all hosts and VMFS LUNs in it, you’d need to create a separate Storage Group for each vSphere host, adding all the VMFS LUNs, plus one LUN dedicated to being the system/boot drive for the host. Thus I think local boot would simplify management of the LUNs on the Clariion, especially for large numbers of hosts.
This entry was posted in Hardware, Performance, Storage and tagged , , , , , , , , . Bookmark the permalink.

One Response to EMC CLARiiON – Stuff I’ve learnt

  1. rcmtech says:

    Having found out more from other courses, FAST Cache is not actually a cache at all, and does not work anything like a hybrid SSD+HDD.

    It analyses how “hot” the data on the FAST Cache-enabled LUNs is, and the hottest blocks are copied onto the FAST Cache SSDs, a bitmap of all the blocks for the LUN is then updated to mark which blocks have been copied to SSD. This bitmap is checked whenever the LUN is accessed to see where the blocks live, SSD or HDD, and the appropriate location is then accessed. If the Fast Cache SSD fill up and a new block becomes “hotter” than any fo the existing ones, the data for the “coldest” block on the SSD is copied back to HDD and the new hot block is copied over in its place. FAST Cache also takes time to “warm up” once it’s enabled due to how the algorithms/monitoring work.

    Thus it’s nothing like a “real” cache at all (e.g. the RAM cache on the storage processor, or any RAID controller), and will not get you out of intermittent high IOPS conditions. FAST Cache does NOT put the data onto SSD first, then migrate cold blocks to HDD. It can help a lot in certain scenarios – e.g. you could find that the blocks holding a heavily accessed SQL database table are promoted onto SSD, which is far more efficient use of SSD than using FAST (1GB chunks) or just holding the entire database on SSD. So it does have its uses, but soaking up bursty IO is not one of them.

    More info: http://www.emc.com/collateral/software/white-papers/h8046-clariion-celerra-unified-fast-cache-wp.pdf

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s