I know that this post will get me flames from the usual suspects, but I have to make it anyway. For the last several weeks my friend Mike at Zerowait has been asking questions in his blog that merit some attention here.
It has long been a question of mine why you couldn’t buy the exact drive used by the frame seller, but directly from the drive manufacturer — usually at a substantially reduced cost over what the OEMs mark sell them for a 300% mark-up is common).
This has not just been an issue with NetApp. IBM, EMC and a rash of other vendors do the same thing. They mark up the drives in their cabinets, but prohibit their customers from simply buying replacement drives from Seagate, Hitachi, Fujitsu, Western Digital or whichever vendor manufactures them.
The response I get when I ask this question of an OEM is always the same (and I paraphrase): we format the drive in our own way, using a certain number of sectors, to ensure drive resiliency or to ensure that they will work with some voodoo that we do with our storage in microcode.
Mike, who watches NetApp like a hawk, has seen another trend, that he emailed me about today. NetApp’s recent flip-flop from 520 Sector formatting (which they once told me was required for resiliency and compatibility with WAFL and RAID 4) back to 512 Sector formatting is viewed by Mike as a sham from a technical perspective. Here is a quote from Mike’s email.
You know this 520 to 512 is a big thing.
1) NetApp tells everyone that 520 Sector (BCS) is better
2) NetApp forces every one of their customers to buy new drives that support BCS
3) NetApp switches back to 512 sector ( ZCS)
4) NetApp forces new ZCS customers to upgrade to dual parity – wasting disk space because they are using less reliable technology.
I guess they will force all of their customers to switch to BCS drives again soon… increase sales again
Storage is now sold as fashion!
When is the proprietary block format no longer done for technical reasons, but instead used as a method to cajole users into upgrading their platforms? I have to agree with Mike and with the fellow from Hitachi who he quotes in his blog that this smacks of marketecture, not architecture.
I invite NetApp to explain the move in a response here, which I will promote to a full column topic. I need some clarification from Sunnyvale about why, exactly and technically, this is happening.