Dell is looking to add an all-flash array to its converged infrastructure storage portfolio, with an acquisition looking likely. Networked all-flash arrays deliver data from flash memory to servers very much faster than a disk drive array. The product space is led by Violin Memory with other startups involved as well. We'll list …
This is all wrong.
Back when I was alive companies like IBM, Xerox, De Havilland, Bell Labs, had vast R&D departments and invented their own stuff. Acquisition is not R&D and is susceptible to fashion and market distractions, it does not take the sum total of human knowledge into new places.
Michael: If you want new stuff, employ designers and build it yerself.
Go back 80 years and there was only a few places to go to actually innovate - IBM, Bell, General Electric, etc.
All the innovation came out of those guys. They were *the* places to work for engineers, physicists and such.
This was true, in my opinion, right up until the late 90's. Then the internet came along and revolutionised the way things could be sold. Suddenly small players didn't need bigger players to get their product out there and more and more people made that jump and started up their own businesses.
I'm not saying you're wrong - you're right....but some of these companies being acquired are very good, much better than what people like Dell could make if they threw a few billion at it. Think of buying in the market as outsourcing the innovation.
Yes, but outsourcing has been SO succesful for the railways, hospitals, HR departments, etc...
But Dell, M$, and all the aquisitive remainder (like GE, Siemens, Disney) are paying a lot for success and nothing for failure. it is close to risk-free.
The risk is being taken by the rest of us who invest in startups direct or through our pension funds, and are paying for the 80% failure rate. The acquirers are getting fat on the 20% success rate, without covering the cost of failure at all.
A bit late to the party but given how well Dell handled their previously acquired divisons...
...they have a good chance of succeeding here too, provided they found a solution with solid R&D and existing market share. With the same (careful) approach they did with EQL they can integrate it very well in a year or so.
The issue is rather the timing: they are late, they should have been among the first ones shopping for an all-flash solution... That being said I'm a bit sceptical of EMC's XtremeIO (they are inflating their results with dedup) and I don't know how RAMSan PCIe card numbers stand up in real tests (eg a similar but lot cheaper OCZ enterprise lineup couldn't even boot in Dell workstations when I tried to run them in pairs)
I think flash-only boxes at their current prices will make the most financial sense if they are *significantly* faster than their much-cheaper disk-based counterparts eg sustained bandwidth at least in excess of 4-5GB/s (and even that isn't much of a challenge when you can stick 3-4 PCIe storage cards into any Server 2012 box and run hundreds of them in active-active scale-out file sharing clusters)...
Based on my criteria I'd probably skip Pure - as far as I can tell they also have some dedupe-massaged numbers but at any rate 1.2GB/s per unit isn't really Earth-shattering today (downside of building on commodity controller architectures, I guess.)
Violin seems to have a better performing lineup in their 6xxx-family claiming 4GB/s but I'm not sure about the pricing (their lower-spec'd products are pretty much in HDD-based performance territory with smaller capacity though they might be cheaper, not to mention their latency advantage.) Certainly possible candidate...
Nimbus might be another valid option but it's not without caveats either: the tech seems to be solid w/ 10-12GB/s per box, using custom-mfr'd SSDs, cutting out the controller fat etc but AFAIK sold units are in the hundreds with eBay being their biggest customer w/ 100TB... I have to add that Nimbus wasn't too convincing for me few months ago either when after agreeing on an eval run I was sent a document to sign that said I will have to buy it if the unit arrives back to their Cali office after the 15th day (we are in NYC!), regardless of performance results (and the doc already sanctioned a mandatory purchase if my performance goals met)...ouch. I asked for the removal of these terms and Nimbus immediately backed out of the eval, citing 'difficulties to justify the cost of supporting the evaluation' (= ie. 'we don't have money to keep sending out demo units, we have to be sure you will buy it') because I don't have the money ready (we did, I just cannot get a budget approved without knowing the price tag of the right solution.) Not a red flag but neither exactly a confidence-inspiring experience - maybe with Dell's solid financial backing it would fly...?
GreenByte sounds interesting but IIUC it's a VDI-specific solution...? Never heard of the rest of the pack - how about a quick rundown on them, Chris? :)
Someone correct me if I'm wrong here, but does Whiptail actually have any unique IP or real Flash technology? I don't think so.
And to the first post, it seems you've forgotten Dell's original business model was never to develop real IP. They were a very good integrator of Intel/Windows technology, and then expanded into selling their own technology (via acquisition) a bit late, but with acquisitions that have in large been positive ROI.
Violin, Kamarino, and Pure are too high end for Dell and don't match the rest of their portfolio. SolidFire is a specific Service Provider play. Nimbus would be their best bet.
Well, out of all of Dell's acquisitions EqualLogic clearly worked out very well, Ocarina and Exanet were always about acquiring IP for cheap (Exanet was already busted so Michael Dell's ashtray change largely covered it) but Compellent seems to be rather, well, so-so, it isn't really the hit they've expected it to be among EMC upgraders but it's not CML's fault, they are simply not a replacement for top-tier EMC gear that Dell lost access to when they've severed ties last year...
OTOH Dell *is* doing some decent integration work, they just need to be more targeted and hire a lot more developers to accelerate the process because they seem to be stuck at building blocks level - they've bought CML almost 2 years ago, I've expected them to tier data between their NX-series, EqualLogic and CML by now but so far nothing in sight, even different EQL boxes weren't able to tier stuff between them a year ago IIRC...
Even basic upgrades, things that should be routine take forever at Dell: they were about ~2 *years* late on upgrading NX boxes running Windows Storage Server 2008 to version R2 - I wonder when they will move to Storage Server 2012, around 2015, maybe even later?
And anyone they acquire it will come, once again, with a *HUGE* software challenge, when they have to integrate ALL these solutions into one big, vertically compatible system... ideally you can just pick and choose which type of storage box you need and they will be able to talk to each other, even if at a limited level (for advanced mgmt I can imagine they would require CML or something similar.)
As much as I know about them I don't think Violin is too esoteric, after all Dell is trying to attack the high-end, what's better than doing it by using new tech at a very competitive price?
That being said I agree, Nimbus is probably their best bet - they have their IP that could greatly benefit from Dell's volume advantage (mfr'ing your own SSDs can get a lot cheaper when you order tens of thousands instead of hundreds per quarter, ahem), they have their own sw stack and they need money so they would be a lot easier to deal with than Violin which probably thinks they worth a lot more than Dell would pay... =)
Dell should roll their own
Supermicro makes a 4U 72HDD server chassis for $2500. Dell could probably slim that down to 3U and 96HDDs by rotating the backplanes 45 degrees. Stuff in some 500GB SSD drives for the price/performance/density sweet spot, lay in a server MB and some I/O. Now it's off to their own software guys to turn this into a 10M IOPs 40TB RAID6 storage screamer with quad 10Gbps iSCSI or some infiniband and network redundancy. Mark it up 3x and make some nice margins. Done and sorted for 0.04 billion dollars.
Re: Dell should roll their own
Right except it would be slow as a dog thanks to the stupid bottleneck of using SAS controller/channels, built on commodity hardware so it would be the same thing as someone buying a Dell server and fills it up with SSDs for 1/3rd of the price and of course, it wouldn't solve anything as it would have to run a commodity OS, resulting in another low-end, run-of-the-mill offering instead of a highly integrated, top-tier solution.
In short it's the worst possible idea, a solid failure.
Integration and Comments
I wonder who this reference is being tossed at???
All-flash array startups are hot right now, we suggested. "I can wait nine months until they are less hot."
He also said in passing that he doesn't like prima donnas.
The reason things take so long is they can't integrate for beans. What started as a hardware company will always be a hardware company, no matter how many acquisitions they make.
- Vid Hubble 'scope snaps 200,000-ton chunky crumble conundrum
- Updated + vids WHOA: Get a load of Asteroid DX110 JUST MISSING planet EARTH
- 10 years of Facebook Inside Facebook's engineering labs: Hardware heaven, HP hell – PICTURES
- Very fabric of space-time RIPPED apart in latest Hubble pic
- Massive new AIRSHIP to enter commercial service at British dirigible base