Information lifecycle management (ILM) is based on data having different value over time and then storing it on different types of disk: costly and fast for the mission-critical transaction data; mid-speed and mid-cost for the everyday data; and low-speed/high-capacity for the reference data.
EMC reckons that the concept of tiering network access to storage has value too. Ken Steinhardt is EMC's CTO of customer operations. He characterises three kinds of networked storage in enterprises: network-attached storage (NAS); iSCSI; and Fibre Channel (FC).
Steinhardt says of iSCSI: "It fills the space between NAS and Fibre Channel. It's the next logical progression for people outgrowing NAS but not ready for Fibre Channel."
The implication here is that NAS delivers data slower than a Fibre Channel can deliver block data from a storage area network (SAN). The implication also is that the really mission-critical data has to be delivered in blocks, that you wouldn't use NAS for the really important stuff.
This may or may not be true. When organisations take a marketing term to heart, as EMC is doing with ILM, then everything tends to be expressed in a way fitting that term. When did we not hear an HP product described as helping organisations to be more 'agile' or IBM products helping business to react better 'on demand'?
So ILM impacts most if not all EMC marketing pitches. Steinhardt says: "We position ILM as not only a storage tier thing. I believe the concept extends to the network stack and to data protection tiers."
It seems that tiering of storage networks is done on the basis of speed of data delivery in the EMC view. It sees iSCSI's role as a lower speed and lower cost route to block data in an IP SAN or as a lower speed and lower cost route to FC SAN data for servers not connected directly to FC SANs - FC SAN extension in other words.
We've ended up in the generally agreed places for iSCSI in our storage world, having started out from an unusual starting place, that of tiered networks.
How should customers approach these three network tiers? Steinhardt says: "The first question is 'Can we use NAS?' If yes, then do it. If not the second question is 'Can we use iSCSI?' If the answer is yes, then do it. If not then the service level requires the third question to be asked. 'Can we use Fibre Channel?' Historically, people have defaulted to Fibre Channel."
Now iSCSI provides an alternative to this default.
Steinhardt reckons that the network tiers can be independent of the storage tiers. He pictures a network fabric with, say, a Cisco director providing server access via NAS, iSCSI or FC, to back-end storage that could be Clariion or Symmetrix, the two poles of EMC's disk array offerings. The MDS box could be running InVista, EMC's storage virtualisation software platform.
Currently EMC virtualises NAS via its Rainfinity software and FC SAN via InVista. Steinhardt says they are "totally independent" and talks of being able "to ultimately abstract it completely and add in iSCSI."
Steinhardt says that EMC's multi-path file system can choose either a file-based or a block-based delivery route for NAS data. If a large data object is requested then block delivery over iSCSI could be quicker than file delivery via NFS or CIFS. He says that the iSCSI block delivery route could transfer the data up to four times faster.
What's happening is that "the (multi-path file system) software virtualises the ultimate delivery mechanism. I think it's pretty revolutionary."
It remains to be seen if other NAS/iCSI vendors provide the same virtualised network delivery method. Over to you NetApp?