Quick Thoughts on ONTAP 8.1.1

Due to the interest generated by the podcast I posted on Monday night about Data ONTAP 8.1.1, I thought it might be a good idea for me to follow it up with a short blog post outlining what I like and don’t like about ONTAP 8.1.1 and what I think about the current state of play with NetApp ONTAP.  I’ll keep it to ONTAP as I am a technologist at heart….

For those who are interested in hearing the podcast, the links below will allow you to either play the audio directly from the website or to download it to whatever device your reading this from -

DON'T MISS AN EPISODE – SUBSCRIBE FOR FREE!

…or click to download this podcast

Infosmack Podcast MP3

 

To Cluster or Not to Cluster

While I appreciate that ONTAP 8.1.1 brings a lot of feature and functionality to cluster mode (c-mode) I’m not sure I quite understand where I would deploy cluster mode versus 7-mode.

In the EMC portfolio, for instance, I understand the use-case lines of demarcation between Isilon and VNX File.  One is better suited for general purpose file serving and virtual machine deployments, while the other is more suited to big data, oil & gas, scientific and scale out use cases.  I’m not sure I understand the lines of demarcation with ONTAP 8.1.1.  Hopefully somebody can clear this up for me.

When ONTAP was at 8.1 the choice was pretty simple. C-mode had pretty much zero feature and functionality in comparison to 7-mode.  Want any of the goodness that NetApp was known for, go for 7-mode.  Have a niche scale-out type requirement, go for c-mode, or another vendor. 

Now that 8.1.1 is announced, the decision is not so simple.  C-mode has a lot more about it with the introduction of 8.1.1 and appears to be looking more and more like the flagship half of the product (c-mode and 7-mode being two halves of the same product).

Will development in 7-mode see a huge nose dive now (if it hasn’t already)?

Is c-mode suited for general purpose file serving configurations, or just for more scale-out requirements?

If C-mode is the answer to everything, how do NetApp achieve the twin goals of being great at scale-out and those niche workloads while still being great at general purpose random access file serving?  The competition have separate products, one for each.  How does ONTAP achieve both with a single product?

Back to Innovating

I could well be lambasted for saying this, but it seems to me that NetApp have had their eye off the ball a little recently (too much internal focus on delivering c-mode?).  To me they look to be playing catch-up on tiering, flash usage, and innovating in general.

In saying the above, don’t get me wrong.  NetApp have been no slouches. But considering their pedigree in innovating I think they have slipped more than slightly.  One or two quick examples -

  1. No all flash appliance
  2. No server side flash offering
  3. No real block based tiering technology

All of the above are big hitters in today’s market and NetApp seem to be lagging.  If they;re not careful even HDS will beat them to items 1 and 2 above.

Does One Size Fit All

There’s no doubt, ONTAP has been excellent for NetApp.  But is it becoming a bit of a dinosaur?  Are they flogging a dead horse?

It seems to have taken an age to integrate Spinnaker (or bring scale-out/cluster-mode) to ONTAP.  Why?  I appreciate that ONTAP was feature rich to the extreme and integrating all of those features in a scale-out product is no mean feat.  But seriously, nearly 10 years to bring scale-out!?

Might it have been a better move to have two separate products (as EMC do with VNX and Isilon)?  Heck who knows that EMC won’t kill off VNX File/DART in the future and take more of the NetApp ONTAP route……

Final thoughts

ONTAP 8.1.1 looks good from what I’ve heard so far, but the devil is always in the detail.  Who knows what caveats there are behind things like Immortal Clusters , Infinite Vols and Data Motion.  Oh and ONTAP Edge Mode where ONTAP runs in a VM and brings ONTAP functionality to small branch offices with the option to replicate back to a Data Centre running ONTAP.  They all look like cracking technologies from where I’m sitting.  They just feel a little over-due.

Oh and if they can nail best of breed general purpose file serving along with scale-out, they will have one hell of a product!

Oh but time for one final question……Is it necessary to have to scale-out by adding HA pairs?  Surely HA pairs should be a thing of the past.  Surely they are a hangover from ONTAP that they couldn’t code out in time for release?

Listen Now

 

Download

3 comments for “Quick Thoughts on ONTAP 8.1.1

  1. TimC
    July 17, 2012 at 11:44 pm

    C-mode is the future.  You'd deploy it if it has enough feature parity for you to deploy on today.  The very name "7-mode" implies that it's legacy.  I'm willing to bet there were countless hours spent between marketing and engineering arguing about ever even having an ONTAP 8 "7-mode".  WIth a name like that, I'd imagine they had originally planned on there only being one version of ONTAP 8 (c-mode).
     
    As for host-side caching – bollox.  While they may be slow at getting it out the door, they've been working on caching for years.
    http://www.theregister.co.uk/2011/06/17/netapp_project_mercury/
    http://static.usenix.org/events/fast11/posters_files/Byan.pdf
     
     
    As for block based tiering, I'd actually agree with the NTAP CEO that tiering is the wrong choice.  Why tier when you can cache?  Does your laptop sample workloads and then at midnight every night decide which pieces of data to move from your hard disk to memory?  Or does it just do it real time?  What's more useful?  Give me caching over tiering every day of the week please.

  2. July 18, 2012 at 1:01 am

    Hi Nigel, Dimitris from NetApp here.
    Interesting post and assumptions. 
    Lagging? Doing things differently is not lagging. We may never have certain features some of the competitors have.
    Just because the architectures of other systems benefit from certain things, it doesn't mean our architecture will. And vice versa.
    It's all about solving the business problem – if I can solve the business problem in a different way, who cares if I don't have a certain feature a competitor may use to solve the same problem?
    In addition – waiting until we have a baked product instead of pre-announcing stuff (or selling stuff that's not baked) is also not lagging. However, it may be bad marketing… but that's not a technology issue :)
    Feature-wise, C-mode is practically on par with 7-mode (with the notable exceptions of Metrocluster and SnapVault as two major things it doesn't do yet).
    However, if you don't need those two features, C-mode today has many features 7-mode doesn't. And does scale-out, nondisruptive everything, including piecemeal replacements of cluster bits, and allowing non-uniform clusters (you can't have a VSP or VMAX with dissimilar engines, right? Or replace a bunch of disk shelves just because they're older now, all non-disruptively?)
    Why it took a long time to integrate the Spinnaker code with the ONTAP7 code: You do realize we are the only storage company that has accomplished something of this magnitude, right?
    Meaning, combining two completely different major storage products, based on different OSes with different kernels and pretty much zero common codebase. And ending up with an enterprise unified scale-out storage product capable of high performance for both block and file: http://bit.ly/Mp4uu0 and http://bit.ly/K2FBz1
    Imagine the coding and QA effort needed. While, at the same time, continuing to improve the feature set of the products (talk about a moving target).
    That is hard work and innovation.
    The competitors buy or develop other platforms, and have no near-term plans to meld the technologies. Their goal is that you buy storage islands from them, with zero commonality between said islands. 
    The whole point behind C-mode is creating a general-purpose product able to run block-based workloads very quickly and at low latency, while also allowing scale-out file storage, with all the goodies ONTAP is known for (write optimization, dedupe, snapshots, clones, megacaching) – plus allowing flexibility and ease of operation and upgradability. 
    A very tall order indeed.
    But think of the payoff and the potential savings for customers.
    Thx
    D

  3. July 19, 2012 at 7:27 pm

    Nigel, I appreciate your thoughts and look forward to speaking in depth about c-mode in an upcoming InfoSmack.  But first we need to clear up one fundamental piece – running c-mode does not necessarily mean you are clustering.  Clustering requires that you assign nodes to a cluster, and then create the cluster.  This is an option in c-mode, just like defining what level of RAID you want to run.  Now perhaps you can see NetApp's logic in developing a single platform, ONTAP, that can start with a single system, grow to multiple systems, evolving into clustered systems.  One platform to learn, one platform for the majority of your workloads, one platform that grows with you as it becomes your complete data infrastructure… is that innovation?
    Larry Freeman aka DrDedupe ( a NetApp employee)

Leave a Reply

Your email address will not be published. Required fields are marked *


*

You can add images to your comment by clicking here.