« What IT Leadership Can Learn From Manufacturing | Main | My Continuing Frustration With Automation »

February 27, 2014

Comments

Mohammed M

Thanks for this. My question. Is whether vSAN localises VM storage to the compute resources in use for both flash/spindles. If not, isn't I/o latency incurred by having compute/storage in separate boxes akin to the I/o latency realised when processing I/o through eg Nutanix's VM storage controller appliance?

Chuck Hollis

Hi Mohammed

VSAN pools and abstracts physical storage. VMs consume from the pool. Both VMs and their storage objects may potentially live anywhere in the pool. Taking that approach (vs. insisting on hand-placed storage objects) makes VSAN very simple to use.

But there really aren't many advantages to making storage and compute co-resident, if you think about it. For example, all writes have to be done to two nodes in order to be protected. And read latency/bandwidth over the recommended 10Gb network is pretty darn good -- just like the storage array guys use!

I've seen people at Nutanix make the hand-waving argument that they are somehow better because they put an application's data and associated compute on the same node.

But that can't be the case consistently, if you think about it. And there's no evidence whatsoever that doing this delivers aggregate performance benefits for a pool of applications, compute and disk. It certainly increases complexity, though.

Ask yourself, doesn't all data need to be written to at least two nodes to be protected? What if there's not sufficient storage capacity on the allocated node? And so on.

It's a good discussion -- thanks for the comment!

-- Chuck

iwan 'e1' Rahabok

Thanks Chuck, I hope the discussion remain technical and do not change into a religious one :-)
Are you planning to make a comparison with PernixData? I know it's a different school of thought since PernixData still need an external array for Data At Rest. But it's a kernel module, so all the advantages that VSAN has, PernixData also has, at least from performance point of view.

Can you also compare from Troubleshooting point of view? Meaning, how difference are all these architecture when "fire" happens (performance issue, availability issue). Often, "ease of troubleshooting" gives us a real picture on how elegance an architecture is. I believe in the converged world, a good architecture must lend itself to "debugability". I've configured VSAN, Nutanix and Pivot, and have not been able to master how I'd troubleshoot it if there is a fire.

Full disclosure: I'm a VMware SE since 2008, and the lab guy for VMware ASEAN.

Chuck Hollis

Hello Iwan:

I will do my best to stay on the technical side of the discussion.

Pernix is interesting to me, but not interesting enough. My impression is that it can do smart read caching, and do some write caching for transient data that doesn't need to be persisted. And it can do this for any external storage. I also note that there are a number of similar things on the market. Maybe they do it different, maybe they do it better -- I don't know. And I'm not sure the differences really matter.

Then I go look at something like VSAN, and notice it does smart read caching as well as smart write caching, including persisted data. But it doesn't do this for external storage -- that's part of the current design.

So you'd be interested in one, or the other, but not both.

When it comes to management and operations -- including firefighting! -- I think more fully integrated products like VSAN would have a distinct advantage. It captures the entire environment in one place: apps, server, network, storage, etc.

And getting a single, consistent and coherent view is sort of job #1.

Thanks for the note!

-- Chuck

Jerry

Hi Chuck,

Don't want to affend you but with all the discussion about the overhead of the hypervisor it looks like we are going back to pre hypervisor time. All arguments are the same, so why not run applications's native on hardware again. if it's better to put things native into the hypervisor this will work for everything. Is it's VMware's strategy to build everything in hypervisor? What can we expect. What about NSX? Is Vmware going to build that into the hypervisor? I am sure there are benefits to hypervisor integration but it will also lock you into VMware and create a thicker hypervisor. My guess is that hypervisors are mature and efficient enough to build complete functionality into virtual machine aka a software defined datacenter. I don't see the need for hypervisor integration altough there may be some theoretical benefits. Software defined functionality is a new way for providing hardware where the differences will be made in functionality, not theoretical efficiency.

Chuck Hollis

No offense whatsoever ...

Something has to sit between application logic and bare hardware. That could be an operating system, a hypervisor -- or, as you point out, a portion of the application itself.

I grew up in an era where operating systems (e.g. MVS, UNIX, VMS, etc.) were expected to be that abstraction. When the big shift to Intel happened, just using an operating system alone proved to be inefficient on many fronts. In addition, the introduction of hypervisors also created entirely new functionality (e.g. VMotion) that we hadn't seen before.

I'm not quite sure how to answer your other questions. More functionality will inevitably find its way into the hypervisor, but that doesn't mean "everything".

I would argue that the benefits are not theoretical, especially as one considers orchestrating multiple resources in a consistent fashion. Whether those translate for specific situations, we'll see.

-- Chuck

Verify your Comment

Previewing your Comment

This is only a preview. Your comment has not yet been posted.

Working...
Your comment could not be posted. Error type:
Your comment has been saved. Comments are moderated and will not appear until approved by the author. Post another comment

The letters and numbers you entered did not match the image. Please try again.

As a final step before posting your comment, enter the letters and numbers you see in the image below. This prevents automated programs from posting comments.

Having trouble reading this image? View an alternate.

Working...

Post a comment

Comments are moderated, and will not appear until the author has approved them.

Chuck Hollis


  • Chuck Hollis
    Chief Strategist, VMware SAS BU
    @chuckhollis

    Chuck has recently joined VMware in a new role, and is quite enthused!

    Previously, he was with EMC for 18 years, most of them great.

    He enjoys speaking to customer and industry audiences about a variety of technology topics, and -- of course -- enjoys blogging.

    Chuck lives in Holliston, MA with his wife, three kids and four dogs when he's not travelling. In his spare time, Chuck is working on his second career as an aging rock musician.

    Warning: do not buy him a drink when there is a piano nearby.
Enter your Email:
Preview | Powered by FeedBlitz

General Housekeeping

  • Frequency of Updates
    I try and write something new 1-2 times per week; less if I'm travelling, more if I'm in the office. Hopefully you'll find the frequency about right!
  • Comments and Feedback
    All courteous comments welcome. TypePad occasionally puts comments into the spam folder, but I'll fish them out. Thanks!