One of the interesting things in enterprise storage has been the FUD around iSCSI vs. Fiber Channel (FC). iSCSI has consistently been looked upon as the lesser option of the two storage transports. It’s debatable if it’s traditional enterprise storage vendors or “old guard” storage engineers resistant to change spreading the misinformation. It may be a combination of both but the consistent message has been the iSCSI is not an appropriate platform for large enterprises. The primary knock has been that IP based storage isn’t a mature platform with few examples of large enterprises with tier 1 applications running over iSCSI.
I lost a debate that iSCSI was mature enough to deploy within a green-field environment with a large enterprise due to my inability to show large-scale deployments within the enterprise. This may have been a chicken versus an egg conversation because most dominate enterprise scale products relied on FC. After this year’s Storage Field Day 5, I walked away with fuel on the side of iSCSI.
A little history
iSCSI is a relatively new protocol with the first draft standard being submitted circa March or 2000 by Cisco and IBM. The basis of the fear around iSCSI lies in the growing pains of any protocol. Storage has traditionally been a much more conservative area for innovation with a lower tolerance to “growing pains.” Therefore, iSCSI had been regulated to value plays and markets such as SMB and development and test environments. Not only were these networks typically slower 1Gbps networks of the generation but the products were under powered for the scale to meet the enterprise market. It would be the same as comparing an entry level 4 Gbps FC Array to an enterprise SAN running on an 8/16 Gbps FC fabric.
3PB SAN running on iSCSI
SolidFire, an all Flash SAN solution, made one of the more interesting claims by any of the vendors at Storage Field Day 5. SolidFire provided an example of a customer with 104 of their storage array nodes deployed at their largest client. Dave Wright, SolidFire’s CEO, mentioned that the average node is 30TB so that would put the overall storage footprint at ~3PB of SAN based iSCSI. This number automatically piqued my interest. I followed up with a question regarding what type of applications were being run by customers on SolidFire’s platform and Wright answered that they had customers running Tier 1 applications including Oracle and SAP ERP.
This got some immediate response from Twitter with some people questioning SolidFire’s roadmap that included FC. The questions were basically if life in the iSCSI world was so good why add FC. As a follow-on question isn’t waiting to add FC a demonstration of immaturity of the product. The answer to both questions was that SolidFire focused on their target market. Their target market is mostly cloud providers that were comfortable with iSCSI due to cost savings from CapEx and OpEx efficiencies associated with the transport. According to Wright, there was not large demand for FC from their customer base. This has changed and warranted a revisit to adding FC support.
While one vendor’s account of their iSCSI solution being leveraged for Tier 1 applications doesn’t completely dispel all of the FUD it is a critical data point. Prior to this reference, the only large-scale iSCSI deployment I ran across of a large “3 letter” U.S. agency that couldn’t be referenced directly.