Businesses are investing more than ever on the devices their employees need to perform their work. But what is the overall cost and how do your device management choices affect the business?
I must start out by thanking Pure and Westcon for giving me the opportunity to travel to Las Vegas to attend Pure Accelerate 2023, and especially Carl and Dave for hosting us.
Events like these give me a great opportunity to meet a great bunch of new people and whilst the media will have already hit the news stands, here is my view of the key announcements of the event.
FlashArray//X R4 and FlashArray//C R4
No real surprise here, kind of expected as the R3 has been out for several years and technology marches on. The new release moves to new Sapphire Rapid Intel CPU giving ~40% increase in performance over R3 with no increase in price.
The X70 and X90 will also include the DCA (Direct Compress Accelerator) adapters that currently in the XL – giving even more performance.
There’s a now C90, and the numbers changed to align to the X series. So the C60 becomes C70 etc. The controllers are essentially the same, it's just the DFM/media and the tuning of PurityOS that are different.
Benefits for Evergreen//ONE customers is that when your array is due controller refresh then you are automatically upgraded, and if you have an X70 or 90 then you will get the DCA boards. I like the quote, "It's like driving your old car to work, but when you get out at the far end, you find you are driving a brand new car."
Whilst not technically earth shattering, there is a new bezel across all the ranges that is made from recycled plastic.
The new FlashArray//X
This one caught me a little off guard. But the promise is that this wee gem will:
- Designed to replace all spinning disk including the 7200 NL spinning rust, slower but higher capacity than the C
- 1-6PB (max 3PB raw)
- Basically same Purity and whilst has all the same functions as the //X and //C it's not designed for things like Virtual Machines. The //E is designed to deliver over Ethernet (no Fibre Channel option).
- Targeted to backup/Secondary storage, Medial Imaging etc.
I guess the real test will be when we get pricing for this.
75TB QLC and 36TB TLC DFM
A massive 75TB DFM, the biggest available flash module on the market.
QLC available for //C and //E arrays
That's 1.5PB raw storage in 3U! (20 DFM in an //E) And with dedupe and compression, those numbers are insane.
We also saw mock ups of the 150TB DFM (essentially 2 75TB boards piggy backed together in one module (due late this year/next year), and more than enough subtle hints that 300TB modules will be following soon. All this with no increase in energy requirements and no fancy new physics needed. There certainly seemed to be an expectation and 600TB and 1PB modules would follow.
Ransomware Protections SLA
If you have Evergreen//ONE then there is a new Ransomware protection SLA. If you get owned. Then Pure will ship clean replacement array within 48hrs (APJ) 24 (US/Europe), and migrate data @ 8TiB/hr
- Need to follow best practice in terms of Safemode, snapshots etc to qualify
- Allows you to keep original Array/Data for forensics for 180days before it needs to be returned to Pure.
- Pure1 also shows Ransomware detection alerts – i.e if the data reduction rate drops suddenly on volumes then get indication. Not designed as an upfront detection, but and in-depth response so you can see extent of issues
This could be really useful for customers to get back up and running again, whilst retaining original array(s) for insurance/law enforcement protection.
Keynote and Main Themes
Apart from Product Announcements there were 2 main themes.
Disk is done
With the //E family of products, Pure believe that they can truely replace all the spinning rust in a datacenter. My take is that the caveat here is that to meet the price point, some things have to be done at a certain scale. So with the minimum size of the FA//E being 1PB raw, then this may not replace every device in the datacenter, but I look forward to seeing what the options are.
As part of this there was also key messaging on why SSD are not the answer. Yes SSD contain NAND flash, but they also have a whole layer and interface (the Flash Translation Layer) which adds complexity and cost to the SSD in making the NAND flash look like a disk. By having the controllers access the NAND directly, then the Pure 'secret sauce' is more about what's NOT in the DFM rather than what is.
Did you know that for every 1000 bits of NAND memory in an SSD, you need 1 bit of DRAM? So open up a 1TB SSD and there will be 1GB of DRAM, this is why you don't see large SSD, you can't fit the memory in.
Environmental and Green Energy savings (ESG)
Worldwide, datacenters are responsible for about 2% of global energy consumption. And storage of data is responsible for 25-30% of that number. The move to all flash has the potential to reduce that number by 80%.
The future is here to stay, I lost how many times ChatGPT was mentioned, and the amount of data stored is increasing massively year on year.
Direct Flash can go a long way to reduce the energy footprint needed to store all this data.
I hope that gives you an insight on what goes on in Vegas, and maybe not all of it needs to stay in Vegas.