10 Responses

  1. Jason Boche
    Jason Boche at |

    Disclosure: VMware enthusiast employed by a storage vendor.

    I’d summarize that what I’ve witnessed from VMware so far is a similar level of boldness with VSAN as they had with the push of the VSA. I don’t have any numbers to back up this analysis but I think it’s fairly safe to say the VSA isn’t doing so well and for VMware’s sake I hope that history does not repeat itself with VSAN. VMware’s scouting reports incorrectly recognized market demand for a VSA touting mediocre performance which was pinned to both available and afordable storage types at the time, lack of protocol choice, significant overhead, limited scalability, and a price point that didn’t make sense when compared to other competitive options

    Providing the smallest amount of latency between demanding applications and storage is swell but we’re still talking about edge cases for the fit – just like FT, just like SR-IOV, just like DirectPath I/O, just like NPIV. VSAN is in need of serious road testing to build confidence as a bake-off contender for organizations in the mid to enterprise space. The silo breakdown hasn’t occurred yet to the point that storage SMEs have been displaced to look for new work elsewhere. Examining VMware’s history of licensing for new features, careful consideration will need to be given to attract and not immediately price themselves out of the SMB market. VMware has partnered with hardware vendors for many years to draw business investments in datacenter consolidation and convergence. I don’t see large opportunities near term for a VSAN fit. It’s going to take 3-5 years for current hardware trends to roll off their tech refresh cycle in order to make additional datacenter footprint space for boxes with the drive density to accommodate the VSAN architecture. It’s not rational to directly compare what we’re talking about here x86 virtualization 10 years ago which in comparison was a fraction of the complexity and the fit made immediate sense while not trying to snuff out Dell, HP, and IBM in the process.

    At the same time, I’m concerned that storage related API development and array vendor integration with partners who have had a head start in storage that can be measured in terms of decades, appears to be in a slump. If VMware’s intent is to sweep the datacenter floors with VSAN, that’s fine, but the timeline for that initiative doesn’t align with the timeline required to reach the market’s definition of storage maturity which is a prerequisite. In addition as you point out, the architecture is not unique at this point. Stiff competition for this revenue stream is around every corner.

    I <3 VMware

    Reply
  2. Evan Powell (@epowell101)
    Evan Powell (@epowell101) at |

    Disclosure – I pioneered software based storage way before it was cool as founding CEO of Nexenta. I’m now CSO of Nexenta.

    On the whole, my take on new storage architectures is GREAT – that’ll be an interesting alternative in 1 million+ hours of production usage and/or 5-7 years. Storage is the deep end of the pool.

    Having said that – if you have critical mass or a lower bar in terms of performance and data integrity and availability, you might get there more quickly. For example, I think Swift despite well known limitations is on a curve that gets it more and more relevant in the “niche” it fills. Keep in mind that in storage a “niche” like this is $1bn+ in TAM.

    My second overall point is that you still have the CAP theorem to consider. We see all sorts of architectures based on host based storage – and we sell one as well. These approaches typically sacrifice A or availability in order to get ease of use and scale out. I would argue that even DAS based approaches face similar trade offs.

    By comparison, back to Swift, it sacrifices Consistency. And that’s perfectly ok with their use cases. They are open about that. What does VSAN sacrifice out of CAP?

    Final point is VMware is perhaps verging on overreach. Every major OS and systems vendor over the years has, at some point, decided to do storage. I mean – “how hard can it be – and we have natural advantages.” So said IBM, Digital, Microsoft (still making progress in storage), RedHat, the late great Sun and others. Together their market share is tiny. Maybe VMware will the exception that proves the rule. However, it is a different set of users (historically – is that changing?), it is a different set of partners (don’t see this changing – its a new channel), it is a different set of field engineers (don’t see that changing). Historically – its a different company.

    Having said all of that – clearly some brilliant people working on VSAN. Just this afternoon I chatted w/ a friend at VMware who is really proud of the progress being made. If you read my note above – I’m talking about the reality of the market into which VSAN is being positioned. VSAN c/d the best storage product in history and I’d still expect it to take many years and a gazillion production hours to be adopted.

    Reply
  3. Nick Colyer
    Nick Colyer at |

    Disclosure: Customer and huge VMware fan

    We’ve been talking about this a lot, and really there are some key things said here which resonate with the people i’ve been talking internally about this.

    Here are some general themes..

    1. We will probably look to POC VSAN in the short term. Kick the tires on it and start to build some confidence between us and the storage guys. Thankfully we aren’t silo’d off so it’s an easy conversation between the Server/VMware guys and them. The storage guys are all for these kind of simple solutions at remote locations so they don’t have to support additional infrastructure.

    2. We would first be looking for VSAN to potentially replace VSA clusters of which we are currently in the middle of deploying a few. It’s a shame VSAN isn’t already GA as the migration may be a pain in the arse depending on how much storage we have left in the 3 node clusters we deploy remotely.

    We also are looking at what NetApp offers to remote offices for the exact same use case. The use case is only for remote offices which require an HA solution and do not have a high I/O requirement today, although it appears VSAN could handle that. All major applications though we try to centralize to the main datacenters where we already have a huge investment and this is just to support local infrastructure in order to save traffic going over the WAN (SCCM, AD, Qualys, etc.).

    3. VDI Rollout: We may have considered VSAN for VDI had it been GA already, but needless to say there are a ton of vendors all with great VDI options out there today that have built trust already.

    As for replacing anything larger, at this point it’s a trust thing. Would I want to throw all of my storage on VMware just yet, probably not. Are remote low risk sites a great place to start? We think so. It also begs the question in some people’s opinions of what does this mean for companies like Nutanix…Is there any point in considering them with VMware bringing VSAN out. I’m sure others have thought the same. Not to bash them as they have an interesting offering but VMWare having VSAN means they could be a risk if VSAN takes off.

    Finally, I completely agree it may take years to see larger adoption. We aren’t like to just replace our current storage implementations which are only a year old in favor of VSAN overnight. We are working on a ton of automation, of which policy driven storage decisions are a key piece. As we aren’t implementing VSAN today a number of these workflows will be built around current vendor offerings and integration. Not to mention policies and procedures for automated storage provisioning etc. Changing this over to another vendor would require significant time, be that VSAN or anyone else.

    Either way these are exciting times!

    Reply
  4. Esteban Ocariz (@estebanocariz)
    Esteban Ocariz (@estebanocariz) at |

    Disclaimer: VMware fan, ex-channel architect and employee of Nutanix

    Firstly, I love to see all of the different takes and opinions each person brings to the table. If there is one thing people can say about these conversations, it’s that they do not lack passion.

    Jason makes some great points and I do believe that there will be a time where most companies will have a mixture of server-based and SAN storage in their data centers, but many more will begin to move away from the traditional SAN we know and love. The Googles and Facebooks and Amazons of the world have set the bar and the benefits that server-based storage bring to the table are just too great to ignore.

    In response to Evan; all valid points obviously born from industry experience, but coming from the Nutanix side of the table, the last 20 or so months of selling have opened our eyes to demand for this solution. Granted, dev and engineering took years to get us to where we are and we will always be fine-tuning, but the fact that we are the fastest growing tech infrastructure company in the last five years speaks for itself. People want this technology and they want it badly. When large companies are displacing vBlocks and Flexpods (with many millions of dollars of investment) with Nutanix, people start to turn their heads, stop and think. I believe this is a wave and it isn’t even close to peaking.

    And lastly, in response to Nick; I believe that because VSAN is part of esxi now, we will see a ton of people doing just what you said, installing it in their environment and “Kicking the Tires”. Obviously, if VSAN can deliver, there will be adoption, if it can’t, maybe they’ll be validation enough to get Nutanix a win. Honestly, Nutanix really sees VSAN as validation to what we’ve been selling and shipping for a little while now. Do we see VSAN as competition? We would be stupid not to. Like you said, VMware has some brilliant guys working on VSAN. If when launch comes around, if they can get close to our performance, and close to feature parity with us, they’ll have a product people want and can replace their SAN with. I think VSAN has a steep hill ahead to burn the image of VSA from people’s minds and break the SMB/ Remote Office / Branch Office association, as well as some catch-up with performance and scalability/design of a product like Nutanix, but its a little early to tell.

    The bottom line is: there is room for everyone in the market (Evan said it with $1BN+). When they come out of beta we will welcome VSAN to the party and then its back to work.

    Exciting times indeed!

    Reply
  5. @vcdxnz001
    @vcdxnz001 at |

    Disclaimer: VMware Partner, Contractor, VMware Fan.

    It seems that the storage industry is going through a shift similar to what client/server was to mainframe. I think there will always be a requirement for scale up monolithic storage arrays at the top end and for certain use cases. But it seems that VSAN, Nutanix, Simplivity and others will all have their place in the market also. What VSAN really does is what Esteban describes, validates the architecture of scale out hyperconverged systems for everyone else. It gets a lot of mindshare for customers, which makes it easier for everyone. I see in the future a hybrid approach where the SDDC will provision the infrastructure required by the apps and it’ll have the different service catalog items of either scale up (traditional shared arrays) and scale out (hyperconverged systems). perhaps we’ll even see automation in the movement between these two approaches based on observed system utilization and performance. Even though some of these converged solutions have been around for a few years I would say it’s not even getting started yet, there is a long way to go before it gets anywhere near the peak. The opportunities are endless at this point. What’ll be important is matching the application workload requirements with the archtiecture solution that best suites them, in a declaritive way, based on the SDDC model. That is where the magic of VMware meets the magic of the ecosystem partners.

    Reply
  6. Josh Odgers (VCDX#90)
    Josh Odgers (VCDX#90) at |

    Nice post Chris (as always!)

    Disclaimer: Passionate VMware evangelist and Nutanix Employee

    I am of the opinion that traditional shared storage will in coming years no longer be the dominant architecture of choice and we will see a significant percentage of virtualization deployments running on Scale out , Shared nothing solutions like Google, Twitter & Ebay use today, and what VSAN is aiming to bring to the market alongside Nutanix who have for around 2 years had GA products shipping in this space.

    There is a huge market for storage technology like VSAN & Nutanix and the more players who enter this space, the more it validates the scale out shared nothing architecture VMware are building and Nutanix are using today.

    Storage has in my experience always been one of the biggest challenges in Virtualzation environments, and I think this is widely accepted and one of the many reasons there is a huge boom in the storage market at the moment.

    As a huge VMware evangelist, my concern is this is a major undertaking to develop a product like VSAN which is no easy task and I would hate to see a repeat of the VSA or for the resources within VMware to be taken off the core products which VMware are dominating the market with.

    A significant challenge I see is if a company deploys VSAN (or any similar product) in the ROBO, or for Test/Dev being two of the initial use cases, it will create another different storage tier to manage, and potentially change the way these environments are managed, eg: backup/recovery/DR etc.

    I think the strength of Nutanix is we provide a platform which can be used for everything from Business Critical Apps , Production servers, Test/Dev & VDI with DR/Backup/Recovery features built in like SRM support and replication.

    VSAN has a major benefit in being built into vSphere, and easy to deploy (once you have chosen & deployed your hardware) , so VMware have done a great job in this area. I agree with Nick & Estaban’s comments about Kicking the Tires, since VSAN is built in, im sure lots of people will test it, what will be interesting is what percentage of the tire kicking will result in a Production deployment.

    The fact VSAN can be deployed on existing hardware (subject to HCL) is also a great feature and will help adoption. I think if VMware can in the coming years develop VSAN to deliver and end to end solution for the datacenter & ROBO I think it will gain some considerable market share.

    Regarding the minimum size deployment of 3, with recommendations from at least one industry expert to use 4 nodes, I hear some comments that this will be a hurdle to adoption, but I think this objection will dissolve once the scale out shared nothing architecture is properly understood by the mass market.

    I don’t see 3 or 4 nodes as a problem at all – in fact even prior to joining Nutanix I was encouraging customers to scale out to get better efficiency/resiliency for their vSphere clusters. VSAN & Nutanix architecture aligns with what most Virtualization experts recommend which is a scale out architecture rather than scale up (small clusters) as scale out helps deliver the performance & resiliency we all expect from our vSphere solutions.

    To Mr Powell’s comment “We see all sorts of architectures based on host based storage – and we sell one as well. These approaches typically sacrifice A or availability in order to get ease of use and scale out. I would argue that even DAS based approaches face similar trade offs” – I am not intimately familiar with the Nexenta technology however Nutanix (and I) believe our implementation of Software Defined Storage using DAS we provide a solution which delivers both higher performance & increased availability/resiliency over traditional shared storage. I wrote this post showing a quick example of Nutanix resiliency as your comment confirms this is still a significant question in the market with this technology.

    http://www.joshodgers.com/2013/10/26/scale-out-shared-nothing-architecture-resiliency-by-nutanix/

    I don’t think performance is an issue for DAS, im sure Fusion-io and a like DAS vendors would argue (and I would agree) DAS should (and does in my experience) in a lot of cases out perform shared storage.

    I think the best use case for VSAN when it comes GA will be VDI, as a different silo of storage for VDI is common due to existing shared storage challenges, but I believe scalability limitations (hard and soft) as well as some of the underlying architecture of VSAN make it potentially more dependant on the network performance than other solutions on the market.

    From a scalability perspective, Horizon View supports 32 node clusters so VSAN really needs to match supported cluster sizes to be a true disrupting technology. It would be annoying having to reduce the scale of View clusters to meet storage constraints, since this issue has in recent times been resolved (8 node to 32 node View Composer clusters).

    With clusters of this size, I suspect the lack of data locality / distribution features may lead to performance/resiliency problems, but without have done testing on this scale with VSAN to date, I can only speculate and look forward to reading any case studies / experiences. I suspect this is one area VMware will work to address in VSAN over the next few releases if it proves to be a significant challenge.

    I like Chris’ comment “Does an entirely new architecture need to dominate the rest, or can we have data centers that go with server-side storage and others with shared array storage?” – I believe this new architecture doesn’t need to dominate the rest and even if VSAN got a 1% market share, I think it would be a success for VMware since they are not a “Storage” company as such. But I expect the new architecture will continue to take market share over the coming 3-5 years and I suspect it will eventually have a higher market share than traditional architecture, but we have to work through hardware life-cycles and educate the market before scale out architecture can potentially dominate. In saying that, there will likely always be requirements for large central storage solutions, what will change in my opinion is what data we store on them.

    In conclusion, its exciting times, and I look forward to being part of this disruptive storage market and seeing where VSAN and similar solutions end up.

    Sorry for the long comment!

    Reply
  7. Tommy Trogden (@vTexan)
    Tommy Trogden (@vTexan) at |

    Hey Chris – you might want to send Nutanix a Sponsorship bill for that ad Josh just posted as a comment 🙂 HAHA !

    My 2 cents – the largest amount of unused storage in a datacenter is still DAS and things like VSAN federate all that under a vSphere environment and makes it useable and easy to consume across a pool of servers. It’s the future and it takes time to bake out technology like this but rest assured, it’s going to be rock solid.

    It will win, it’s a freaking checkbox in vCenter for god sakes. “Mr Customer, you can buy my “Hyper-Converged Solution based on SuperMicro” or …or…..you can just check the box in vCenter and have the same thing. So, how many of my nodes would you like to buy from me 🙂 bahaha…i kid…i kid.. 🙂

    My humble opinion – we will never get to a day where 1 type of delivery of storage is the only way to go. There will always be the right tool for the right job and it’s going to take more than just 1 tool. Very similar to cars today, my wife and I drive totally different cars. She has the Suburban to haul around the kids – i have a truck..because I live in Texas and it’s required by LAW – but either way, we need 2 different types of automobiles to fit our requirements. Storage isn’t that much different. Sometimes we need to haul a crap load of stuff, and sometimes we need to get from point A to point B as quickly as possible with as little storage as possible. VSAN will just be another type of tool to utilize !

    @vTexan

    Reply
  8. Maxta Debuts Software-Defined Storage for Virtualization

    […] at the computer layer to form a distributed storage target. Sound familiar? The idea is similar to VMware Virtual SAN, which is the VMware specific product that builds a storage pool from many local storage devices at […]

  9. The New Haswell Fueled ESXi 5.5 Home Lab Build

    […] you plan on working with VMware VSAN, HP StorVirtual, or any other “disk + flash” storage using local disk, make sure to […]

  10. Welcome to vSphere-land! » VSAN Links
    Welcome to vSphere-land! » VSAN Links at |

    […] VSAN Public Beta (vNinja.net) vSphere 5.5 Improvements Part 4 – Virtual SAN (VSAN) (Wahl Network) In Which I Muse Over VMware’s Virtual SAN Architecture (Wahl Network) What’s New in vCloud Suite 5.5: Virtual SAN (VSAN) (WoodITWork) Introduction to […]

Share your point of view!