EMC upgrading all-flash VNX with bigger SSD and drive box

EMC's VNX-F range of all-flash VNX arrays will soon scale up to 172TB of usable capacity, with a less than $5/GB cost, using a new drive enclosure and higher-capacity drives. A data sheet (pdf) on EMC's website talks about the new SSDs and drive enclosure. These are block storage arrays, accessed by 1/10GbitE iSCSI, FCoE or …

[tom 99]

> For comparison Pure Storage's FA-450 scales to 70TB raw, and a claimed 250TB effective capacity after deduplication.

But... if you have a large database to fit into this capacity, it WILL NOT dedupe at all! Every single 9kB Oracle block is unique.

[Anonymous Coward]

The VNX supports dedupe also, but EMC is honest enough to base their pricing on actual storage, rather than some 'make up whatever you want' dedupe or compression based figure.

Most of the usage scenarios dedupe excels at don't require the extreme low latency for random I/O flash delivers. Where you want that most is for a DB, which as you point out is not going to dedupe.

[M. B.]

The reason...

...for this is because it would butcher XtremIO sales, which does it's data reduction in-line like most modern AFAs.

The all-flash VNX is intended to provide the same feature set as the disk-based and hybrid VNX systems so existing customers can have an all-flash option and manage with the same toolset/skillset and same features (RecoverPoint, VPLEX, ViPR, etc)

I've seen even smaller VNX systems with MCx code and a bit of flash deliver some big IOPS with consistently low latency considering the requirements of who is buying them (and the latency of the old VNX with large flash pools), but they are intended to be a multi-purpose, multi-protocol Swiss Army knife of storage with disk foundations, not a hyper performance all-flash thoroughbred.

[Anonymous Coward]

Re: The reason...

Lets' not pretend the VNX-F is something it's not. EMC's dedupe on the VNX-F is post process and good luck with the compression if you're after low latency and yes that's also with all the supposed MCx goodness.

If these and other limitations weren't inherent to the VNX architecture then EMC would really have no need to push a one trick pony like Xtremio into the AFA market.

TBBH I'm at a bit of a loss to understand who other than a die hard EMC customer would even consider VNX-F, it just isn't competitive on any measure in the market they're trying to shoe horn it into.

[Najt]

Happy another vendor AFA customer

Well on our pure storage array the LUNs which holds MS SQL databases shows data reduction betwen 4.2 to 1 and 5.3 to 1.

Average read latency shown in vmware is 0 ms with occasional spikes up to 1 or 2 ms ( 2 or 3 spikes per 20 minutes interval).

So it seems that they are doing something right, afterall the controllers are compute monsters compared to those from VNX.

[Terafirma-NZ]

Re: Happy another vendor AFA customer

This is what we see as well.

[Anonymous Coward]

Re: Happy another vendor AFA customer

Careful with always data reduction with databases. You have effectively run data reduction on the transaction logs and critical master data of your dbs dramatically reducing their availability and recoverability. Undoing everything that microsoft has spent years putting in place to protect your data

*12 years as a db expert*

[Anonymous Coward]

Do remember that the new generation of all flash arrays typically offer De-duplication and also Compression. Databases can often benefit from Compression when de-dupe is not a good fit.

These should all be referred to as Data Reduction anyway.

Disclaimer... I don't work for any of these vendors, just have an interest in AFA technology.