back to article Backers fatten up flash-disk mutant array, sic it on storage giants

Bulk up and get ready to take on the storage giants - that's what hybrid array startup Tegile's backers are telling it, and here's $32m to do just that. Tegile was founded in late 2009 to devise a hybrid flash/disk storage array, and its Zebi product was launched in February last year - background here. Together with fellow …

COMMENTS

This topic is closed for new posts.
  1. Anonymous Coward
    Anonymous Coward

    Not really

    The hybrid architectures that nimble and tintri have are unique. They are designed from scratch. Any mainstream vendor is bound not to cannibalized their existing business. Tegile isn't hard to replicate. From talking to my VAR it seems to be 95% the same as other ZFS based platforms. The quote they gave us has great $/GB though...

  2. Nick Dyer

    The above poster gets it - Nimble (my employer) and Tintri are unique as they've developed systems with no hangups of previous filesystems, internal/external requirements or politics - so they've had the ability to truly make something new which beats the hybrid offerings from Dell, Netapp, EMC etc in performance, usable capacity (and potentially other things such as cost, management etc).

    Tegile is an odd one as it's really just a ZFS storage array with the metadata placed on SSD. It can be compared to (and has all the same problems as) Nexenta, Greenbytes and other half-baked systems. However their cost is drastically low (so much that margins can't be that good).

    At Nimble, we've officially surpassed 1500 customers and 3000 systems, which is a very impressive statistic for a young upstart and who have been selling product officially since mid-2010... although what's even more impressive is this recent story detailing a non-disruptive upgrade to our first ever customer installed array from a 20K to 75K IOPS system... this is truly what makes us different to (and more appealing than) the Tier 1 guys. http://www.nimblestorage.com/blog/customers/non-disruptive-upgrade-to-nimble-storages-first-deployed-array/

  3. Rob Commins

    Oh Nick . . .

    Nick -

    Please don't try to differentiate Nimble based on having home brewed code versus Tegile's leveraging ZFS. Differentiate on customer value and stay out of the weeds. Here is where the battle is won. For example:

    Data Reduction with Dedupliction AND Compression

    Compression is great, but dedupe is where it is at in virtualized environments with redundant VMDKs and VDI images all over the place. Take 20-30% compression ratios (we all use very similar compression algorithms) and amplify them to 70-90% with dedupe! That takes the effective cache size and disk capacity far greater than compression alone. Our users get up to 200,000 IOPS at well below a dollar a GB from a single system.

    Unified Access

    Users want freedom of choice and flexibility. Why have iSCSI only when you can have FC/iSCSI/NFS/CIFS all on the same platform with no gateway? Database pointing to unstructured data? Sure thing. VDI desktop and data - you bet. No gateway required.

    Active/Active Controllers

    Funny - we use the same SuperMicro chassis, but Tegile users can leverage both controllers to maximize performance to all applications. That means twice the horsepower all the time. What's that? You're a service provider or an enterprise that uses chargebacks with really tight SLAs? Thats OK - you can put a Tegile array in active/passive mode if you prefer.

    Oh, and please don't compare Tegile to Nexenta and Greenbytes (you may as well throw Oracle in there too then); we've focused our developer's attention on improving and optimizing that ZFS platform you poke at. ZFS gives our developers the opportunity to focus on real differentiators as opposed to working on table stakes functionality from scratch, which takes 5-7 years to truly iron out. Trust me - the last two storage firms I was at did it from scratch - it takes a lonnnnng time to get it right.

    1. This post has been deleted by its author

  4. Nick Dyer

    Oh Rob....

    Hi Rob! Thanks for your reply. It's great that the VP of Product Marketing has time to find this post and reply to it on a public forum...(!)

    You're right! Customer value is more important. This is why Nimble has been so successful for the last 3 years of selling technology in the marketplace, and even placed on the Gartner Magic Quadrant for Storage this year in the “Visionary” category. I don’t recall seeing Tegile here (or anywhere in the Quadrant) at all. 1500 customers and 3000 deployments in under 3 years of selling is a huge achievement, so we must be doing something right.

    Just want to pull you up on a few of your marketing claims if you don't mind...

    1. Dedupe vs Compression

    We all know the operational overheads deduplication of data requires vs using inline compression. Heck, a lot of our founding engineers (ex Netapp and Data Domain, no less) do. This is partly the reason we chose NOT to do it; it allows us to run more important tasks on the system like data garbage collection & backend performance and cache-hit optimisations, meaning we can fill an array up to 95% in capacity before any overhead in performance. ZFS, as you know, has huge problems in this area - performance tanks to the floor once past 60% capacity on the box (and starts at 30%). The age-old problem of using a hole-filling file system, eh?

    However, the figures you quote seem wrong. Our customers (all 1500+ of them) see compression of 40-60%+ on their production environments, not 20-30%. And it seems your customer would agree - this one in particular see's far better compression than dedupe ratios on their system. PS - NICE GUI(!): http://www.iphouse.com/blog/mike/wp-content/uploads/2012/01/20120110-zebi-1-volume.png

    2. Unified Access

    Sure, having all the protocols under the sun is great. But if your delivery of said protocols sucks, then that makes you underachieve in everything, excellent at nothing. This is something heard a lot out in the field, where I am every day. Whereas we chose a protocol which we could optimise and build a solid foundation on and be the best in the field with that protocol. Which we are.

    3. Active/Active Controllers

    C'mon Rob... really? Any storage engineer/vendor worth their salt knows that running active/active controllers is a lot more complex to manage with protocol and volume distribution on the system (ever heard of LUN TRESSPASS?). It also means customers only ever run their controllers at 50% load to ensure if a controller failure occurs the system doesn't blow up when everything's running on one controller. It also means storage firmware updates are also FAR more complex, require lots of downtime and may even need engineers onsite to do it...

    Whilst we run active/hot standby controllers in our system (yes, data is mirrored from controller to controller in real time in NVRAM), a Nimble firmware update will take 5-6 minutes in total and will cause 4 PACKETS of downtime in the whole process of the update. 4 PACKETS! that's insane. Also, Nimble can upgrade a storage array from 20K to 75K IOPS by live-upgrading controllers on the fly without adding any further disks or SSDs needed in the system. Can Tegile do that? Didn't think so.

    4. ZFS

    ZFS has HUGE problems, lots of people/customers in the industry know this. You're trying to say that using a legacy code base with a hole-filling filesystem which is maintained by an open-source community with a few India-contractor band-aids is better than engineering and architecting a filesystem from scratch with full in-house engineers and support?! And you're trying to sell these arrays to enterprise accounts?!

    By the way, Nimble has been writing CASL (it's patented and proven file system) for over 5 years now with amazing success... so i'd guess you could say we've "got it right". Our large customer base (and world-class 24/7 support team) would.

    1. The Storage Dude

      Re: Oh Rob....

      What is really interesting here is the overstatement and understatement of a few points on both sides of this interesting debate. I have researched both companies and many more so as to have a better understanding of storage and how to optimize its' use in any environment.

      From what I have been able to discern, the Tegile array is using a heavily modified version of ZFS. They do not rely on the open source community for product development and have features that are estimated to be well beyond those companies that do. Making such comparisons as "Tegile basically uses ZFS", would be like saying all Linux is the same - Red Hat, Suse, BSD etc...of course we know there are similarities and we know there have been specific development efforts to overcome a number of things (not a linux differentiation post).

      There have been enough modifications to the underlying kernel, to include Tegile patented technology that it is no longer compatible with other ZFS distributions. We have seen this before with other storage companies, in that you cannot expect two systems with dissimilar capability to easily share resources or treat themselves as peers when they are not. I think Tegile ZFS is better than the versions out there from Oracle or the Illumos open source community.

      Yes, Nimble has spent a fair amount of time trying to develop a new OS/FS for this Hybrid approach. The Netapp and DD folks are trying to “Do it Again”. I agree with Rob, this is hard to do and get right and typically takes a lot of time. It would appear they have a foundation to build upon, but it also appears that the offering is severely limited. Block services only and Active/Passive architectures. Not sure the point here is well understood or why this is potentially important. More on that later... Also, with respect to both former companies of the founders..what a great statement made by Nick, paraphrase“…dedup is hard to do so they decided not to do it”, or on other words they have recognized what they brought to market before really didn’t work that well and certainly could not use that technology in primary application environments such as VDI and Exchange and others. Interesting how the honesty in that statement was cloaked in a dig at a company that seems to have figured out how to do it.

      Nimble is doing what most startups do, come out with an idea, bring to market too soon and with limited capability and try to gain traction until engineering can fill in the gaps. Build out a large Sales organization and attack the market - and they have done a pretty good job of that. On the other hand, If you are not careful, those product functionality gaps widen and the company becomes moot as newer, younger companies(Tegile) bring systems rich in functionality and capability and typically with less organization weight (overhead). This ends up forcing you to spend time scanning blog post and attempt to derail their progress because they have more than perceived value, but are delivering real value.

      The other comment about separating mete data from data – a complete lack of appreciation and understanding of the implications here. From deduplication to Read/Write to RAID rebuilds, but if you don’t care about data optimization along with speed and performance, then I guess there is no need to understand or appreciate this.

      What customers want…..

      Most customers are dealing with the economy and spending restrictions like everyone else. Budgets are tight, staff reduction take place and yet more IT services are needed to gain a competitive advantage to stay in business and grow further. They want full feature rich tools to enable IT and the business and not islands of technology that ends up costing them more in the end. Tegile’s multi-protocol approach with Hybrid SSD technology moves the bar significantly higher for those organization looking to do more with less.

      Is Nimble a better Dell EQ box, on that point I would agree. But they still lack the capability to transform the datacenter by providing a storage platform that can be specifically targeted for many different application use-cases – doing more with less. If storage is done right, then it’s not about the physical equipment being used, but how IT organizations are enabled by having the right tool in their tool-bag.

      Having versatility, flexibility and tool that allows you to get many jobs done at once, in essence allowing you to provide storage services for the application consumers in a way that is dictated by the business and not your storage vendor, is what customers want. Tegile makes a significant step forward in being able to provide that.

      Yes, I am more a fan of Tegile than Nimble!

  5. Nick Dyer

    PS

    One last thing - Marketing claims can only take you so far... I think you'll find the majority of storage/virtualisation admins look past the fanfare very quickly to understand the key differences in the systems, which in storage arrays is the filesystem & codebase which delivers the feature sets.

    if your system is built on a "house of cards" file system, no amount of claims is going to hide that fact. Regardless of how great "dedupe" and "unified storage" sounds.

  6. This post has been deleted by its author

  7. Harry Park

    Where to next?

    A fine achievement by Tegile with this new round of funding, money doesn’t grow on trees you know. The new partners coming on board , looks like Meritech and GreyLock. They are of course very experienced. Just look at how many companies they have grown and taken to the next level and IPO'd.

    As this was a finance article about $35M of investment it would be good to see a press release from Tegile. Maybe too early?. Storage is always super competitive and the technology continues to roll forward quickly. There is a stark contrast between some of these newer devices compared to the technologies of even just 5 years ago. So what’s next for Tegile and what are they going to do with this new money? My guess is as per the article, and that is to go and build out the organization and sell loads of it.

    I am not sure why Nimble feels the need to post a 'we are better' post an hour or so after this article was published. Scared much? It is either naivety that compels someone to duke it out here , or just in experience. I figure that Tegile would be happy to be compared to Nimble, similar upbringing, both Silicon Valley companies, ( same HW platform? ) ,but Tegile would seem to tick more boxes and one of those is Fiber Channel which is still a massive piece of the storage market. Nimble say they are the bigger startup. O.K, that would be true for now at least. Both companies are startups and really how much bigger is Nimble? Tegile would seem to have something that addresses a much larger part of the market. I think that FC would have been very high on the list of things considered by those now investing $35M and coming to the table with their skillsets. So it’s game on and no wonder why Nimble are paying attention.

    It’s a busy space, Pure Storage, Violin Memory and others. The hybrids are really interesting, again it’s about the addressable market.

    A good news day for Tegile and the fish are biting!

  8. vCenterMaster

    Enough

    Nick -

    You sound awfully desperate. Let's let this dialog rest and have the customers decide. Mudslinging gets no one anywhere. I, as an "anonymous coward" will let the truth come out in a bake off. It sure seems like Tegile comes to the table with a stacked hand, my friend.

    - James

  9. Anonymous Coward
    Anonymous Coward

    Op here...

    Never thought my benign comment on the architecture would generate this controversy and two new registrations for commenters for The Register (who does that ... Harry and James?)

    I am not going to disclose how we decided, but I am going to say it was tough. There was the upfront cost, the rack space (we are in a colo), and the time it takes to deal with storage.

This topic is closed for new posts.