![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
vvols?
- Thread starter Olga-SAN
- Start date
It takes significant work on the array side to support vVols. Plus, you don't throw something in to a storage array without extensive testing. VMware isn't pushing it hard since it's technically still a preview feature. I wouldn't be using it anywhere near production.
interesting...
did you see this:
http://cormachogan.com/2015/03/20/more-virtual-volumes-vvols-and-snapshots-goodness/
did you see this:
http://cormachogan.com/2015/03/20/more-virtual-volumes-vvols-and-snapshots-goodness/
defuseme2k
[H]ard|Gawd
- Joined
- Oct 7, 2004
- Messages
- 1,074
Preview feature? I'd say they're doing a pretty bad job calling it that, it looks released to prime time to me if the array vendor supports it. The upsides are enormous, I know several shops that are waiting to deploy this with baited breath -- ours included. If the risks are high they need spelled out, but it looks 100% GA, supported, the array vendors are spouting their support fully for it, and just about every prominent blogger is proclaiming it is the second coming of Jesus for storage.
We're looking to deploy it very quickly after going to 6, but given who you are I find it concerning enough to at least pause to ask why you'd say what you're saying.
They WANT to call it released to prime time.
It's a 1.0 product. It has a lot of limitations. Roll it into production at your own risk, and with a lot of testing - many things don't work right with it yet (replication, some snapshots, backups from some vendors, etc).
Be very, very, careful.
It's a 1.0 product. It has a lot of limitations. Roll it into production at your own risk, and with a lot of testing - many things don't work right with it yet (replication, some snapshots, backups from some vendors, etc).
Be very, very, careful.
The most compelling thing that VVOLs brings,outside of the technology, shows that VMware is committed to customers who already have investments in traditional storage and contrary to all the hype and communications with the "NO SAN" from others even inside VMware.
As others have pointed out, it's a 1.0 product but adds a significant amount of risk here, were talking about your data now. Quite frankly, if you are thinking about moving to this, then you need to test it like any other 1.0 product.
VASA 2.0 is a two-way street, meaning it's just as important that the vendors get it right as it is for VMware since now, the arrays are responsible for a lot more than just telling VMware what it's capabilities are. It's a complete redesign of how your data is stored.
I'll also mirror lopoetve's post and say there are definitely some interoperability issues right now that also need to be addressed before I would consider this ready for "prime time."
So keep in mind I don't speak for VMware...but we were told back in December that this would be released on a 6 month preview cycle like VSAN was. I think the problem is that VMware felt that preview/beta label on VSAN slowed its adoption down so they held back here. I wouldn't recommend a customer touch it for close to a year. New code on the array and in my hypervisor? Eh.....
Agreed, but let's face it, the Product Marketing message is a bit different. VMware has really pissed me off the past year with the different messaging about VSAN and now VVOLs. I think we can all agree its a great step to take for VMware, but at the same time, they need to get their messaging consistent, which , they used to be great at.
It's marketing. What do you expect
The preview, and GA upgrade issues, caused some serious bad feelings for many folks.
Remember too - VVOL was supposed to come out almost 2 years ago originally. It took this long to get it in a releasable state, and many vendors had a LOT of trouble implementing it (block arrays weren't designed for object storage, lets be brutally honest).
defuseme2k
[H]ard|Gawd
- Joined
- Oct 7, 2004
- Messages
- 1,074
OK how about with a design partner, like 3Par? We can wait, it isn't the end of the world, but we're moving to 3Par arrays it appears -- Given that we have many petabytes, I don't want to do it TWICE.
Many of the limitations are part of the spec - doesn't matter ~who~ you partner with.
Does anyone think that VVOLs will be the future for datastores? As NJ pointed out, it puts significant work on the array side. I haven't really looked into Tintri much, but it seems their storage is purpose built to handle much of what VVOLs is supposed to do and then some.
I am mainly interested as I test and plan for upgrading to vSphere 6 later this year. But for VVOLs I am kind of screwed since our NetApps operate in 7-mode. So I am trying to get an idea if I should start planning to make the move to Nimble, Tintri, etc instead of keeping NetApp.
I am mainly interested as I test and plan for upgrading to vSphere 6 later this year. But for VVOLs I am kind of screwed since our NetApps operate in 7-mode. So I am trying to get an idea if I should start planning to make the move to Nimble, Tintri, etc instead of keeping NetApp.
That is the tricky part. We have three NetApps in all different locations and use SnapMirror between two of them. The oldest NetApp is about 3.5 years old. Next one is 2.5 years old. And the newest is only one year old.
On the oldest NetApp we already renewed the support for an additional year which will expire in about 6 months. I don't know the 5th year cost for support on NetApp, but I'm sure it isn't cheap.
How long do people normally keep storage for before doing a refresh?
On the oldest NetApp we already renewed the support for an additional year which will expire in about 6 months. I don't know the 5th year cost for support on NetApp, but I'm sure it isn't cheap.
How long do people normally keep storage for before doing a refresh?
3-4 years.
That's one of the other advantages of the newcomers - support costs are way more sane
s/little while/since the beginning. That's how the platform is designed to operate. Per-VM only.
That's one of the other advantages of the newcomers - support costs are way more sane
s/little while/since the beginning. That's how the platform is designed to operate. Per-VM only.
3-4 years.
That's one of the other advantages of the newcomers - support costs are way more sane
s/little while/since the beginning. That's how the platform is designed to operate. Per-VM only.
Tintri has been shipping products for 4 years. In the world of enterprise storage that's "a little while".
EDIT: And a lot of the per-VM functionality was added 2 and 3 years ago....
I am guessing Tintri is somewhat expensive? I am going to talk with them in a couple of weeks to get more information about their product. So far the only storage outside of NetApp we have really looked at has been Nimble. They seem to have a decent cost, especially if we sign up to be a VAR. I have nothing against NetApp, but with all these additional support costs, licensing costs, etc. It does add up.
Well, now that you mention it. That is something I have been trying to determine. Nimble likes to talk IOPs all the time and claim how many IOPs their storage can deliver and NetApp can't really deliver the IOPs. I don't overly care for their sales pitch BS.
Our oldest NetApp is a FAS3240 which I believe at most would be up to 25k IOPs. So for Nimble I was looking at the CS300 which is claimed to be 30k IOPs. The next model up for Nimble is the CS500 which is up to 90k IOPs. This exceeds our newest FAS8040 which is rated around 50k IOPs, especially since we are stuck in 7-mode.
As long as that is accurate, we should be OK. I do like that if you get the CS300, you can easily upgrade to the CS500 in the future by just paying the difference in cost between the two. No need to pay full price for a brand new controller.
No, we're not that expensive.
The T820 (entry level) is 23T Usable, 40-50k IOPS, list price of $75k.
PM incoming.
Tintri has been shipping products for 4 years. In the world of enterprise storage that's "a little while".
EDIT: And a lot of the per-VM functionality was added 2 and 3 years ago....
It was always there under the hood - especially the workload profiling, which is pretty beyond VVOL as well.
And recently, 4 years has been a "Very long time" with all the changes happening.
that's not cheap keeping in mind you can get complete hyper-converged setup for fraction of that cost
?
?
No, we're not that expensive.
The T820 (entry level) is 23T Usable, 40-50k IOPS, list price of $75k.
PM incoming.
What hyper-converged product are you talking about? I looked at Nutanix and they were more than double the cost with three nodes and 12 TB of storage.
scale computing start as low as sub-20k for 3-node setup and they offer financing
nutanix is damn expensive and software licensing with it is kind of weird ((
nutanix is damn expensive and software licensing with it is kind of weird ((
you trust nobody? ))
scale computing see my previous post
but they are indeed weak because kvm is definitely not smb hypervisor and ecosystem around it sucks if not awol
scale computing see my previous post
but they are indeed weak because kvm is definitely not smb hypervisor and ecosystem around it sucks if not awol