back to article Pure Storage's coming high-end array: We have the details

We can tell you a few more juicy details about Pure's coming high-end array, and we're deducing that SolidFire has a product capacity refresh coming. What we knew was that the new range-topper would boast up to 1.5PB usable capacity using 4 and 8TB 3D NAND modules (SSDs). It would have always-on QoS, support thousands of …

Silver badge

pretty conservative

Interesting how they (or maybe just you) chose to use a 2.9:1 reduction number. That seems much lower than what most folks (and even Pure themselves?) use.

My own 3PAR 7450 AFA gets between 1.5:1 and 1.7:1 dedupe at the moment depending on the data set. Used to be in the 1.8-2.0:1 range. 3PAR doesn't have data compression yet though. They first announced data compression at an event I was at in 2013, though taking a lot longer than I had hoped to get it out the door.

0
0
Anonymous Coward

Re: pretty conservative

Disclaimer: Pure Customer

Well, the 2.9 to 1 assumption is wrong, because the math works like this with Pure FAs:

522TB raw capacity - ~42% raid and system overhead = ~300TB raw usable

300TB raw usable * 5 (5:1 data reduction) = ~1.5PB usable

Cheers

2
0

Re: pretty conservative

I'm using Pure's supplied effective capacity figures. RAID and other system overhead has to be deducted from that. This gives me consistency across the table.

Cheers.

1
0
Anonymous Coward

Missing Reduction Ticker

Pure used to have a ticker on their web site showing the overall data reduction on their install base but I can't find it any more. Has this been moved or taken down?

0
0
Anonymous Coward

Re: Missing Reduction Ticker

http://www.purestorage.com/products/purity.html

1
0
Anonymous Coward

Re: Missing Reduction Ticker

Pure employee here.

You can find the data reduction ticker here.

http://www.purestorage.com/products/purity.html

People found it "too busy" on the homepage, so we moved it deeper into the site. Our average data reduction has hovered at >5-to-1 across the entire installed base for years now, and it is typical for us to see a 2X advantage compared to competitors - we encourage you to test head-to-head in a PoC. Data reduction has actually improved release-to-release and as array sizes have grown, but our workload mix continues to shift to mission-critical applications and databases (think compression value > dedupe for these workloads), leaving the net still > 5-to-1.

0
0

Hi, Why do you call it "usable" capacity when in reality it is not. Basically it is the new, inflated "effective" capacity. Do we know if Pure or any other vendor that touts "effective" can commit such levels in Writing (PO contract), and not only to their markitecture?

0
1

Disclaimer: Pure Storage Employee

With the 'Love Your Storage Guarantee' from Pure Storage - even the estimated effective usable capacity is guaranteed. If the FlashArray doesn't deliver, we'll provide the customer with the capacity to correct the situation.

-- cheers,

v

0
0
Anonymous Coward

Disclamer: EMC Employee here.

When we do XtremIO configs for customers we show the capacities in RAW / Useable / Logical.

E.G. 1 x 20TB Xbrick = 20TB Raw / 15.1 TB Usable / 45 - 75TB Logical (3x-5x).

Yes, we will put it in writing, but the data reduction rate is going to be dependent on the workload and analysis. We don't count zeros as Dedup so the overall data reduction is going to vary between workloads (VMs vs SQL/Oracle vs VDI).

1
0

That's really not a lot of comfort for customers. It just enables Pure to make some more aggressive claims knowing that they'll get the sale and can fix any mess afterwards. But for a customer that could mean going from 1 array to 2, hassle, inconvenience, embarrassment, etc. Much better to be more conservative / realistic with claims and anything better that is achieved is all goodness for the customer. The problem with Pure is that some of their arrays are quite small in terms of capacity so they need the big claimed saving to be competitive.

0
0

Useable isn't just a myth.

Disclaimer: EMC Customer

The problem with "usable" capacity is every workload is different. When we do RMAN dumps to our XtremIo we only get about 2:1 at best as it's already compressed. When we run our oracle DB's we get more of a 3-4:1 compression rate. So I don't think there is a meaningful way to test a compression/duplication workload. Maybe someone can come up with a testbed?

I know our EMC rep has been bantering about some patent case Pure lost to EMC. (Were about to buy a new array) I wonder if that has anything to do with the missing reduction ticker??

1
0

Re: Useable isn't just a myth.

For oracle databases (and just the source) thats about what you will get and seems pretty consistent with what I see. Where you can get higher efficiencies is in the SNAPS for test / Dev / QA. Each one of those are a thin (not dedup) copy and don't take up any space (at the time of creation, as you add unique data, naturally it will use some space). Not to mention each snap can have the same performance as production.

Not sure how long ago you purchased XtremIO but AppSync limited is now included with the base XtremIO and is retro for all previously sold units (talk to your sales peeps). That would help automate application consistent snaps for Oracle , etc.

0
0
Anonymous Coward

No Nimble, No Dell

Why are you making the same tired mistake (aka Gartner) of not including all-flash arrays from all vendors in your assessments and rankings?

I'm sure you've missed others too.

3
0

Re: No Nimble, No Dell

No Netapp AFF either (Netapp Employee)

1
0

When talking about capacity, I wish all vendors would just talk good old fashioned usable capacity. As someone has said, your mileage will vary with dedupe and compression so capacity only really becomes 'effective' once you've done a poc or proper analysis. Until then it's just 'marketing' capacity which is all risk for the customer.

3
0

This post has been deleted by its author

Anonymous Coward

Nitpicker

"FlashArray m20 - m90 can have Fibre Channel (FC) or Ethernet adapters plugged into IO slots; "gig" is shorthand space-saving for GB/sec."

Wrong - it's Gb/sec

2
0

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Forums

Biting the hand that feeds IT © 1998–2017