Veeam Podcast on Storage FUD

Prologue:  Who am I?  Chad Sakac?!  This post got to be ridiculously long-winded, but it has tons of good info in it, and I certainly encourage a full read!  Sorry for the length, but I wanted to share as much as I could…

It’s no secret that I love social media.  Hell, just look at my page.  I accepted a long time ago that it is the way of the future.  You have two choices:  You can choose to be open and transparent, sharing all across all of the sites, or you can go along with the security/privacy fear-mongers and hide in your house with the curtains drawn.   Anyone that really knows me, even outside of work, knows that I’m a huge advocate of transparency, and just airing everything out there.  It’s healthy, it breeds conversation, constructive criticism and open thought, and does nothing but contribute to forward-thinking innovation.

I love social media.

It’s responsible for a lot of the growth, in my opinion, that virtualization has seen over the past 5-ish years.  One of the biggest mediums (or better, most rewarding) outside of the twitterverse, has been the VMware Communities Roundtable podcast, hosted by John Troyer (@jtroyer) of VMware.  This is one of those things that are quite literally on my calendar every Wednesday, because I refuse to miss it if I can help it.  I’ve been participating for several years now, and have transitioned from a customer role to a vendor role in between.  The beauty of the podcast is that none of that matters.  It’s an open chat about VMware, not about one vendor versus the other.

So when I see podcasts labeled things like “sorting through storage vendor FUD,” I immediately get a raised eyebrow.  These are the kinds of the things I certainly respect and appreciate outside opinions and views on, because ALL OF US in the vendor community are all too often, too close to the glass.

So I opened Evernote, and started listening.  And I wanted to respond to some of the things mentioned in the podcast.  I’m sure this will get read, for the most part, as a “Nick the NetApp guy” response, but keep in mind that a little over a year ago, I was a datacenter admin.  I was a generalist that ran with a Network Engineer, and we were pretty much a two-man shop.  A lot of my thinking and advice/criticism, mostly internally to NetApp people and mgmt, come from my viewpoints as a customer, and are usually anti-Marketing spin, and have to do with making what used to be “my day-to-day” easier and more efficient.

The podcast was hosted by Rick Vanover (@RickVanover) of Veeam Software, and his special guest was a VAR in Belgium by the name of Hans DeLeenheer (@hansdeleenheer).

In this episode, Rick Vanover hosts Hans DeLeenheer. Hans works for Ferranti in Belgium as a team lead for all areas of server, storage and virtualization. Hans is on Twitter and blogs at: http://hansdeleenheer.blogspot.com/. Hans and Rick discuss some of the challenges facing IT pros when it comes to purchasing storage.

In the interest of being transparent, I’m going to leave my notes VERBATIM here in the post, and I’ll address each point in more detail below.  To be honest, each of these points could quite literally be their own individual blog post, and who knows… I might do just that.  It just shows what a good podcast and overall conversation this was.

Enjoy!

———————————————

Notes from Veeam Podcast 53  – Storage FUD with Hans (HP/Dell VAR, Belgium)

Hans:  “If you’re not buying storage with tiers, you’re buying rubbish.”  I think that statement can be misleading.  If we think of traditional “tiering” mechanisms, such as FAST or Enterprise Vault, etc, I would disagree.  Using different types and speeds of disks to prioritize the hosting of different types of data and retentions is not necessarily the right idea.  Especially when it involves proprietary platforms and purpose-built appliances to accomplish it!

Hans:  “If you’re buying new storage solutions, don’t buy old solutions, such as an EVA.”

Highlight the new 2240 platform for the SMB market.  small size, small price, big performance.

Hans makes a good point, that host-based caching is going to commoditize the spinning disk, with things like Fusion-IO and SCSIExpress.

Rick makes a good point about controller software that bridges the gap between high speed SSD Host-Based cache and the commodity spinning disks behind the controllers.

Hans: “Make sure that every software product you have is usable by 3rd parties.”  He specifically called out vCenter plugins being available to the storage admins to manage provisioning of disks and user profiles!  YES YES YES YES!

Rick brings up an interesting point about politics.  Hans’ quickly counters by saying that “you sold them purpose-built storage!”  AND HE’S RIGHT!

Hans bring up the idea of DEDUPE, and how certain vendors are using broad ranges of metrics, such as 4:1 or 15:1.   The point is, there are too many “metrics.”  No other vendor has native deduplication on ALL storage at the volume level.  There are either agents or inline appliances or streaming processes involved.  The point here is:  THE DATA.  What are you trying to dedupe?  It’s not about who has it and who does it.  the difference is driven by the KIND of data, and the level of uniqueness of the data that determines what you will get in return out of your dedupe.

GENTS!  YOU DONT NEED DEDUPE APPLIANCES!!!  It’s the wrong approach and less effective.

Hans brings up active/active sites.  This is easier to accomplish in Europe due to the nature of the geography.  NetApp MetroCluster is MOST popular and MOST used in Europe.

Hans mentions DRaaS.  I agree.  I think this will be the first, overly accepted, use-case of “Cloud.”

Hans mentions Cloud Storage, or Storage in the Cloud.  “Are we going to virtualize the storage across all our types of arrays?”  Why yes, Hans.  We are. 🙂

“Cloud apps eliminate the need for local server/storage endpoints.”   Bingo, Rick.  Bingo.

————————————————

More detailed responses…

Storage tiering.  I think the conversation about storage tiering is over.  At this point, it’s legacy tech, and not something I would consider arguing about anymore.  Workload- or demand-driven tiering by moving entire datasets to different sets of spindles is the wrong approach, in my opinion.   Utilizing a more dynamic load/eject system with onboard caching based on demand is the way to go.  This ALSO allows you to do things like commoditize spending less on the performance on the spinning disk, utilize more dense, less expensive disks, and your demand loads with most often be pulled from dynamic flash configurations, not SATA vs. FC disks.  There’s an inherent difference in these two lines of thought.  While I agree that tiering can serve a purpose in long-term archiving, thats just it:  a purpose-built solution vertical.  Not something to be accessed regularly.

Hans continues to reiterate a point throughout the podcast.  Don’t buy legacy storage.  Buy as much as you need today, and what you need tomorrow, tomorrow.  Avoid things that are EOL, just because they’re a cheaper alternative.  My only counter to that, Hans, is that you have to be buying from a platform that supports that kind of M.O. in the datacenter.  Not all do.  Some technologies require tear-downs and rebuilds of disk pools to simply add storage to existing pools.  Otherwise you’re just creating a NEW pool, and paying the spare + raid disk penalties all over again.  This is bad.  ALL vendors do this.  We’re ALL guilty of it.  We ALL need to figure out a better way.

Hans continually referenced the SMB market throughout the podcast, and they are looking for alternatives to the big expensive enterprise arrays.  Hans, I’d like to introduce you to the latest in our series:  The FAS 2240.  Small form factor, 6000-series performance and architecture, small price tag.  Coupled with the OnCommand software suite (also demo’ed in the video), we have heard the pains of the SMB with limited staff, and we have responded.

Host-based caching.  Hans… You have hit the nail on the head with this one!  One of my personal predictions for 2012 (still need to finish and publish that post…) is that Host-Based Caching is going to be the “it” thing this year.  EMC is dropping Lightning here in a couple of days, and Fusion-IO is in a great place in the market.  But they’re not the only one’s.  There’s a host of others coming out with host-based caching technologies, including us here at NetApp.  You’ll hear more about “Mercury” very soon.  That said, Rick’s follow-up comment was that the pivotal key would be to extend the logic of the storage controller OS up to the card in-host, so that essentially the host-cards were “talking the same language” as the storage controllers.  Without saying more than I’m allowed to, Rick…. BINGO!  🙂   This is why I think you will see a continued dominance by EMC and NetApp as the top tier market share holders, even in this space.

This next part was my fav part of the conversation:  Storage Admins working in vCenter.  Hans specifically called out that “every software product you have has to be usable by third parties.”  I would love to have asked if he meant 3rd parties as in leveraging API’s?  Or non-admins of that software leveraging it and day-to-day working in that software?  I have been having some internal arguments about whether or not Storage Admins would/do want to use storage plugins inside of vCenter.  Personally, I am all for it.  I have been advocating for it.  So if you’re a storage admin out there, I (and our Product Mgrs!) would love to hear your opinion!  If we built MORE storage-centric functionality into the plugin (i.e. not specifically geared for the VI/server admin use), would you be comfortable provisioning storage from the vCenter console/VIC client? Leave me a comment below if you want to weigh in and drive direction!

I think I said all I need to say about dedupe in the original notes.  🙂  I’ll leave it at that to not turn this into a finger-pointing pissing contest about who does dedupe better.  I’ll leave it at “there’s things we could ALL do better than we currently do them today.”  How’s that?

DRaaS.  I think this will be the first non-consumerized, enterprise use-case of Cloud.  (also on my list of 2012 predictions…)  SRMaaS, anyone?  😉

Cloud storage.  This is a tricky one.  Public/Private/Hybrid?  Well, for Public, we’ve seen the wide acceptance of Dropbox, but they’ve had privacy issues with their TOS, stating that anything uploaded automatically becomes the property of DropBox.  The enterprise community was not amused.   VMware has introduced us to Octopus as well.  (personally, very excited about this!)  From a private perspective, I think when you talk about “private cloud storage,” you need to have the conversation in the context of “doing for storage what VMware did/does for servers.”   I would encourage you to read up on NetApp’s latest version of DataONTAP 8.1, currently available as an RC3 to customers and partners.  Imagine if you could “maintenance mode” a storage controller, and shuffle all workloads off to perform patches, hardware refreshes, etc…non-disruptively.  The 30,000-ft view is an image of immortality, and non-disruptive operations.  That’s the big picture.  We’re doing it.  More details on this when PR takes the leash off.  😀

To finish, I’d like to close with one of the things Rick said towards the end:  “Cloud apps eliminate the need for local server/storage endpoints.”  You nailed it.  We’ve got to get the OS out of the way.  Cloud Foundry was a big “first step” in this direction, and storage system OS’s need to become more intelligent and aware of cloud application foundations.  At the end of the day, from the controller down, we’re all just disks, and we manage 4k blocks.  Above that, we rely on software to drive the intelligence up the stack.

This was a great podcast, and one I’ll be subscribing to and listening to going forward.  Rick is a good interviewer, and whether they’re prepared ahead of time or not, asks the right questions to continue to move the conversation forward.   He has extended an invitation to me to participate in one of these in the future, and I’d be more than happy to, anytime! 🙂

-Nick