"The reason that this is not a 50 per cent saving, as in the original guarantee programme, is according to Clifton because the V-Series has to use the third party suppliers' RAID technologies and not its own."

Something is a bit off here

By my math assuming best practices of both vendors (EMC DMX & NetApp) with 500GB sata drive. Configured with raid 6, 14+2 on both vendors (max size either one supports), both using snapshots.

500GB NetApp Drive = 423.8GB : Raw device usable (after "rightsizing" and byte conversion)

NetApp overhead:

Waffle = 10%

CoreDump = 1%

Aggr Reserve = 5%

Snap Reserve = 10%

Usable of 14+2 raidset = 4390.5GB

500GB DMX Drive = 476.9GB : Raw device usable (after "rightsizing" and byte conversion)

DMX Overhead:

Misc: = 0.02%

Snap Reserve = 10%

Usable of 14+2 raidset = 6007.6GB

The DMX is 26.9% more efficient at presenting the same physical quantity of disks than the NetApp to begin with, so they are already in the hole to begin with. Then they claim that they can get 50% more if they were to use NetApp disks instead of DMX disks... a total hump of 76.9%?

So if I allocate one terabyte of data on the DMX, NetApp is claiming that they would really only need 230.8GB of space. Color me skeptical of reality, I wonder what would happen if thin provisioning was turned on in the DMX...

Ohh... I see what they are doing. NetApp is allowed to use Raid-DP(and 4 if you read the guarantee DP is not a requirement for the guarantee), competitors are compared using Raid 1. NetApp is allowed to use snapshots, competitors are compared using full copies. NetApp is allowed to use thinprovisioning, competitors are compared only using thick volumes. So basically they are requiring their compeititors to use old tech (even though they have new available), but NetApp is of course allowed to use their newest tech.

Take it from me, having used both EMC and NetApp, the latter's RAID6, Thin Provisioning, Thin Cloning, Snaps and dedupe ACTUALLY WORK.

Try and configure a CX or DMX with RAID6 for production. Your users will lynch you. Same for snapshots & virtual provisioning. But I'm sure your EMC salesdroid will happily sell you EFD's to overcome their firmware deficiencies.

These are the same folks that claimed way back when that they could provide faster access via NFS over a 10 Mbps network than the server's locally connected SCSI disk.

Actually I have been running raid6 on my DMX's in production... and it would appear that I'm still around *not* lynched (you might note that raid calculations on a DMX are done on hardware so I don't have a penalty to move from raid5 to raid6). Additionally I'm running snapshots as well (and some hosts in production additionally mount those snapshots as their primary datasource) and it works like a charm. Since DMX snapshots are async there is no COFW penalty to the hosts, writes are commited to cache and the host goes on it's merry little way, clones as well are done async. (oh on a side note how do your users like your weekly waffle-iron defrag?)

Having ran both NetApp and EMC for about 8 years (early sym 3k boxes and FAS700 series), the DMX wipes the floor with the NetApp from a performance perspective (let's be honest now). So unless my numbers were off (which nobody seems to have noted, including you) the DMX is almost 27% more efficient on storage and it easily outperforming the NetApp as well.... makes for some interesting questions doesn't it? There are things the NetApp does very well (NFS, CIFS, snapvault, automating regular tasks) but there is a lot of things the DMX does extremely well.

I'm sure your NegtApp salesdroid (and pretyy much everyother storage company) will also happily sell you some as well, to overcome NetApp's own firmware deficiencies...

Both vendors post Exchange sizing information publicly and NetApp has enough confidence to post SPEC, SPC and other benchmarks (TPC in the past) as well. In the ESRP reports:

For proper performance, EMC configures RAID10 universally (never RAID6), whereas NetApp configures RAID6 universally (never RAID10 or RAID4, plus they don't support RAID5).

For predictability, EMC turns off snapshots and virtual provisioning, whereas NetApp leaves Thin Provisioning on and enables multiple snaps.

NetApp + All Features = Max Performance. EMC - Features or Efficiency = Max Performance. What could be clearer than that?

That would only be true if a NetApp + All features == a DMX in performance; which any NetApp employee who's not lieing will tell you is incorrect. A NetApp + All features still performance worse than a EMC DMX + all it's features. The thing is with the DMX features it can be either faster than a NetApp or much faster than the NetApp... if you are going to talk Clariion well that's a different beast.

Basically it's the definition of performance, the only way what you are saying is correct if they performan at the same speed when all the NetApp features are turned on and all the DMX features are turned off. Ask anybody and if you seriously try and make that claim you'll be laughed out of the room.

Well Coward, at least we agree the CX is overrated.

However until you show me some independent numbers, I guess we'll all have to "trust" your DMX claims. In the meantime my experiences are similar to what's posted on ESRP. NetApp systems outperform DMX on small-block random I/O (often generated by Windows | Linux | VMware hosts), while the DMX rules for DSS and cache-friendly FICON workloads.

In the past our shop was grossly over-provisioned with Symms and DMXs for all apps. That was economically untenable. We've replaced file svcs, WinApps, SAP dev-test, backup and archive with various NetApp models and frankly wish V-Series would have been available sooner to help with that transition.

OTOH - Our final set of mirrored DMXs remain in production because they're good at what they do and our internal support team has more experience troubleshooting DMX than NetApp. I won't push them to change becuase frankly I prefer a two vendor strategy, but it's now 80/20 in NetApp's favor as opposed to the other way around.