https://slotsdad.com/ - casino online slots

Parallel I/O in the Offing from DataCore Software

by Administrator on November 7, 2015

forgettheflashMy breakout session at Storage Decisions NY had the provocative title “Forget the Flash:  Going Parallel to Achieve Real Storage Performance…and Without All the Cost!”  Provocative because most of the other sessions at the show were heavily oriented toward the “flash for everything” craziness that seems to have descended on the industry.

Anyway, my preso was a mashup of some of the themes I have used in Hollywood, CA at the FUJIFILM Global IT Executive Summit and more recently in a webinar on Parallel I/O that I did with DataCore Software.  The illustrations were simply too good to pass up re-using for folks who hadn’t seen either show.

Basically, I started out with a movie homage to Storage 2015 — in the form of a satire of contemporary blockbuster films.

terminator_genisys_poster1Despite its poor showing at the US box office, the latest Terminator flick — which I renamed the Archivinator — featured a cautionary tale about entrusting everything to clouds (think “Skynet”) and suggested that technologies like tape were, to quote Arnold’s T-800, “Old, but not obsolete.”

oldobsolete

I reiterated the need for tape-based archive as a hedge against the zettabyte-sized data deluge that should break like a tsunami across IT infrastructure on or about 2020.  Folks seemed to get the joke, despite my poor Awnold accent.

I noted that the industry had, instead of solving the storage capacity demand problem, worked instead on addressing the problem of poor IT service performance.  I parodied Jurassic World (“Jurassic Infrastructure” in my deck) to make a few points.

jurassic-infrastructure-posterThe basic story line of Jurassic Infrastructure is the effort to tame the primal forces of the contemporary data center:  virtual servers, software-defined networks and software-defined storage using only poorly trained administrators armed with less-than-superb management software tools.

ji1

The dreamy goal of this “new-ish” infrastructure model is, of course, to achieve the airy goal of “agility” — the ability of IT to turn on a dime in response to changing business realities.  As represented by this send-up of the iconic poster from the blockbuster film that broke all records this Summer…

ji2

Of course, this new infrastructure needs a new lead vendor…a shy, quiet, behind the scenes type with technology to simplify operations and reduce both the CAPEX costs of server-based computing and the requisite IQ’s of IT staff…

ji4

ji3

So, here is the new and improved Jurassic Infrastructure, powered by VMware. But, this is a movie, so there must be drama.

For one thing, this Indominus Rex vendor is a lot bigger and louder and has more teeth than its predecessor, T-Rex (IBM), which dominated data centers back in the day…

ir trex

A little shout-out at this point to IBM, whose Z13 Mainframe is making a comeback in many data centers today. The original cloud!

Anyway, the new alpha vendor starts to do its own thing with infrastructure, forcing customers to go along for the ride.  And, of course, problems start to arise…

ji5

Storage is the big problem, of course.  Application performance issues trace to I/O and the alpha vendor does everything from offloading I/O to legacy storage to villainizing legacy storage and promoting its replacement with newfangled software-defined storage.  What is software-defined storage (SDS)?  Who knows.  It must be storage porn because we only know it when we see it.

I spent awhile catching people up on SDS and its cousin, hyperconverged infrastructure (basically an appliance combining a server, storage array and SDS software), noting that DataCore Software has become a bit of a darling in this space, appearing in the appliances of many prominent server vendors…including, most recently, Lenovo’s brand of hyper-converged infrastructure appliance…

hci1

 

This mention of DataCore also helps me transition to the final movie parody and the main topic of the session:  Parallel I/O.  I contextualize parallel I/O within the forthcoming blockbuster film, Star Wars:  The Force Awakens.  But in my universe, it is “Storage Wars:  The Cores Awaken.”

SWTFAposter1I proceeded to give a basic explanation of what has happened in CPU development that has brought us to the present situation — no increases in clock rates, but a constant increase in the number of integrated circuits on a die…the multicore processor.

Multicore architecture (with threading that multiplies the number of logical cores created from physical cores) reopens opportunities to apply parallel computing concepts to business computing…an idea abandoned in the early 90s because developers could not compete with the tick-tock of Intel unicore processor development…

The good news is that an old Jedi master, Ziya Aral, co-founder and top scientist at DataCore Software, still had his knowledge of multi-processor and has been busily adapting it to a singular mission:  take the I/O from all the workloads hosted on servers today and process the I/O in parallel.  Result?  Embarrassingly, ridiculously faster I/Os than just about anything in the market today and using inexpensive wares.

With Aral’s innovation, which is currently being audited and validated by the Storage Performance Council (watch this space), logical cores can be harnessed to the singular task of parallel I/O processing, thereby speeding up what appears to be the real log jam that has been slowing down virtualized apps.aralyoda  My head exploded when Ziya had me up to his Summer home to show me his accomplishment this past July.  Pity that it has taken so long for the SPC-1 benchmark test he did then to make its way through the SPC audit process.  I have about six articles hanging fire with SearchStorage and Virtualization Review that can’t be published until those stats have holy water sprinkled on them.  Also, Ziya has improved on his original code and is now seeing IOPS and price points that are twice as high and half as expensive as what were demonstrated in the first benchmark test.

This is really important stuff, folks.  Simply put, the DataCore experiment will shortly be a easily deployed software technology that will work in Microsoft and VMware environments, and for all I know, maybe in Linux OS and hypervisor environments in the future after that.  I finished with a couple of jokes that are worth sharing here.

I cautioned that parallel I/O could be weaponized, as suggested by this clip from the new Star Wars trailer showing a light saber design that could only be created by parallel beam emitters.

ols

 

As the technology comes into greater use, I asked, what would we see next?

Embarrasingly parallel lightsabers?

ezgif.com-gif-maker 

Ridiculously parallel lightsabers?

rpls 

 

Or maybe just some holiday-themed light sabers…

holiday1

You get the idea…

ezgif-218651962 

But at least my version of the upcoming Star Wars film at least tries to answer the time honored question, “How did the Millennium Falcon make the Kessel Run in only 12 parsecs?”

powerofcores

 

As I said, the preso went over well in SDNY, with lots of folks asking lots of questions afterwards.  Played right, DataCore may well have changed the tick tock of the whole storage industry with this innovation.  Can’t wait to watch it play out.

You can see my preliminary webcast on Parallel I/O on DataCore’s BrightTalk channel HERE.  Have a look at this previously recorded show (from a week ago) or register to attend the one we will be doing live for Europe on 24 November (very early in the AM in the US).

Cheers and thanks to everyone who attended my sessions in New York, Los Angeles and on the web.

Previous post:

Next post: