Last year, I went to Costa Rica. The greeting they have for everyone was Pura Vida: “Good Life”. That has a new meaning for me now.
- Matt Kixmoeller, VP of Products Five years driving the Flash Revolution
- Neil Vachharajani, Principal Architect A Purity OE Deep(er) Dive
- Vaughn Stewart, Chief Evangelist Asking the right questions about Flash
The heart of Matt’s talk was about how “The Flash Revolution is Right Now”, and in the 5 years since the birth of Pure Storage, they have had a flash focused product. They truly have been revolutionizing how all vendors have thought about how flash should be used.
You change the industry by questioning yourself, and everyone else!
What if Flash could be everyday storage? Well, we have all heard the tagline “Flash for as cheap as disk”. We have seen the gorillas. This thought has been scaring all the other storage vendors, so much so that a few 500 pound gorillas have made a large portion of their keynotes about bashing the little guy. “It’s all about the dedupe”, and the compression, and working on a tiny scale (512b), and thinking of flash as individual cells, not just a “faster disk”.
What if Storage was fundamentally simpler? What, what? Simple? Yes. Simple. You don’t have to do upgrades. You pay support, make THEM do your upgrades! Click a button, open a ticket and say “Dude, upgrade my SAN at midnight this Saturday” and they are all like “Dude, consider it done!” Non disruptively! I’ve done it.
You want a new lun (Volume as it’s called on a pure array) it’s as simple as 2 clicks. Yes 2 clicks. Easy enough, even a windows guy can do it! Ok, you do have to type in the size of volume you want, but I’m sure even a windows guy can figure that out eventually! (Sorry to those clicky wizard windows admins out there for poking fun at you. Ok, no I’m not 😉 *cough* @johnobetto *cough* )
I have one in my lab. It works. This very moment I am pushing 4.88GB/s with 291K IOPS mixed size workload. (3 VMs, one with an 8K SQLIO workload, one with a 16K SQLIO workload, and one with a 32K SQLIO workload). Those are not B.S. 4k block size “HERO” numbers. That’s a heavy IO workload!
Setting up the unit in my lab was simple. It took a fraction of the time that installing Windows, SQL server, and SQLIO did.
What about simplifying how people bought storage? Well, a good starter is only having 2 line items on a quote. Hardware and Support. Not this pack and that pack and this software and that software. And, heck, pay for support for years 4 and 5, and they will throw in a free controller upgrade to what ever is the latest! You mean they don’t want to make money from you on a 3 year refresh cycle? Nope. That’s cool! That’s Forever Flash!
What about support? It’s pretty damn awesome. They typically contact me. Yeah, I don’t open tickets much, and I do some crazy crap with my lab unit. (Let’s say every part in it that I can easily get to has been randomly yanked out at least once or a dozen times. Just to see if they are paying attention. “HE DIDN’T FALL? INCONCEIVABLE.” )
This is from an actual ticket recently:
This is Paul from Pure Storage. We’ve seen multiple failures on array “ha-pure”. I wanted to find out if your office is just doing some testing or if there is a real issue. If there is a problem, could you please open a Remote Assist so we can start doing some research on this issue. If this is just testing, we’ll know
FYI a remote assist is like a reverse shell to them if I want to give them remote control of the unit for investigation.
What happens if you happen to kick off clone/deletes of 500 VMs in a batch, and batch it thousands of times? Well their support analytics would see some wonkiness in the trend, and they would proactively open a ticket and email you to make sure everything is ok. Yep, been there, done that. Did I mention I put stuff in my lab through some insane stress tests?
What about multiple consecutive 10,000 mailbox JETSTRESS 2013 testing for weeks on end? Again, proactive notification “Good afternoon! We have been monitoring your array ha-pure and found an overabundance of metadata created on the array… is everything going ok?”
I like a storage company that has my back.
P.S. There’s no shutdown procedure. You just pull the plugs. How’s that for “built for the worst possible situation”!
What if we could build a better storage company? How? Start fresh, with the best possible team. Learn from the mistakes and successes of others.
Pure was originally going to be an investment into an existing company with the best possible flash storage offering. Investors went to John “Coz” Colgrove (One of the VERITAS founders, Architect of VxFS/VxVM, and one of my storage heros), and asked “What’s awesome in the industry with flash right now?” He was honest. No one had a good, long term, viable solution. He was tasked with making the best team, and helping design the best product. The brain trust that they put together is incredible.
One of the co-founders, and craziest hair in Silicon Valley, is John Hayes. He was a former CTO at Yahoo. Seriously, look at that hair! Imagine the size of the brain that all that hair has to be covering up! It’s mind boggling! (Said as a bald guy with the utmost respect for the hair!)
Other notable figures in the brain trust:
- Matt Kixmoeller, VP of Products, formerly a visionary at Veritas/Symantec (who presented first, and I’m sure would win the award at Pure for the most perfectly sculpted hairdo)
- Scott Dietzen, CEO, PhD, who held CTO positions at little companies you know.. WebLogic & Zimbra, yeah those little billion $ companies!
- Dave Hatfield, President, Veritas (seeing a trend? Veritas knew storage! ), Akamai, Limelight Networks
So, they have a bunch of smart people, with lots of experience. What does this mean for you? Every one of them have played in the Enterprise space before. They know how critical it is to have a stable product, a product that “just works”, and a product that they would be willing to risk their names on!
The technology part of the presentation, by Neil, went into the architecture and designs. It is way too much to put here, but there are numerous videos on youtube with deepdives. “Let me ‘splain…No, there is too much. Let me sum up.”
One detail which people had a bit of a hard time understanding was the Flash Personality Layer. A less than perfect statement of “reverse engineering the firmware” was used to try to describe what it was. It is way more than that. They are taking the functionality and tuning of the flash OUT of the firmware, and putting it into the controller space!
Overview of the FPL (From a Pure presentation in 10/2013)
- Goal: encapsulate SSD properties in a software layer
- Enable per-SSD model optimization for
- RAID / data protection
- Data layout & striping
- I/O timing
- Allows Purity to use multiple drive types efficiently with minimal software change
If you knew intimate details about every SSD model you used, and made those characteristics into a software profile for those drives, you could easily make your IO work how you want it to, or even change how IO flows in realtime, to best make use of the flash!
The final session from PureStorage was given by Vaugn Stewart, about “Asking the Right Questions”. Ever important.
How Simple is the array? Is it transparent to the Application Owner?
Is VVOLs supported? Yep, it will be!
Is there a SPOG (Single Pane of Glass)? Yep, coming soon! Watch the presentation for a demo! What I am really excited for, as a partner and VAR, is that I will have visibility into my customers array performance (if they opt-in), so that I can help them out if needed!
Great new things coming soon from Pure Storage!
Have a workload you want tested on a Pure? Want a demo? Let me know!
*** Disclaimers: #1) I am extremely opinionated and pessimistic about new things. #2) They gave me a SAN for my lab because they knew once I touched it, I would like it. #3) I am a VAR for multiple platforms. #4) I tend not to favor hybrid storage designs, unless it is the absolute best and cheapest solution. My clients live and die by the 5% of cache misses IMHO. This is not a hybrid solution. #5) I am an advocate for @Netapp as a member of the #NetappATeam. I love Netapp. I pick the best solution on a case by case basis for my customers. #6) I do not let 1, 2, 3, 4 or 5 influence what I say or how I feel about a great or terrible feature or great product! You never know who will buy who in the future ***