NetApp Announces First Shipments of FlashRay Systems

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
NetApp today announced its first shipments of FlashRay™, the company’s purpose-built all-flash storage array with the new NetApp® Mars™ operating system. FlashRay is designed from the ground up to improve the performance, efficiency, and manageability of all-flash storage architectures used in enterprise application environments.

The new system combines the benefits of all-flash storage technology with patented and patent-pending NetApp technology to provide breakthrough performance, storage efficiency, data protection, and data management. NetApp’s “no-compromise” approach to flash storage helps customers achieve maximum enterprise application performance at a lower cost. FlashRay began shipping this month to targeted customers and partners.
 
We want numbers!!! Performance numbers or it just another piece of overpriced corporate bling. :D
 
Tak Ne said:
We want numbers!!! Performance numbers or it just another piece of overpriced corporate bling. :D
Click to expand...

They don't give many numbers... All you can assume is that 50-55% of the entire storage will be reserved for WAFL, snapshots and overhead; granting performance on par with or slightly above their SSD PAM modules.

NetApp is great for CIFS/NFS and Data ONTAP v8 of their UI is pretty slick, but I would not touch it for block level storage as its performance blows more than the kissass at your job eyeing a promotion.
 
Tak Ne said:
We want numbers!!! Performance numbers or it just another piece of overpriced corporate bling. :D
Click to expand...

the only number they might give you is the price... then after you regain consciousness you'll look elsewhere lol

NetApp, you can buy better, but you can't pay more!
 
If you are buying EMC, HP or NetApp you are really paying for the established support of their products.

Otherwise, go the smaller name routes that may have as good or better products, but longer term support of the product is up in the air.

Companies with a lot of cash around are going to go with stability. Let's be honest, these days hardware is all a commodity. Whatever brand you buy is going to do the job, regardless of the feeds and speeds some SE is selling you on. But knowing a company is going to be around in 3-5 years to service your product goes a long way for some people.
 
We have a 48TB NetApp setup at work and so far we like it quite a lot. The all-SSD aspect would be great, but realistically I don't think we could generate a workload sufficient enough to utilize it. Even with spinning media the unit we have is damn fast for our workload and replaced a half-dozen individual devices, making the server room much less cluttered.
 
The all flash arrays I see going out there are for VDI deployments.
 
In reply to C0m0dor and FLECOM, I think your comments are based on what you've seen with the existing NetApp controllers and OS. This product is an entirely new hardware platform and an entirely new OS built from the ground up specifically for use with SSD Flash. While I agree that the proof is in the numbers I would suggest that you're rushing to judgement without seeing those results.

Full Disclosure I work for NetApp
 
Taco said:
If you are buying EMC, HP or NetApp you are really paying for the established support of their products.

Otherwise, go the smaller name routes that may have as good or better products, but longer term support of the product is up in the air.

Companies with a lot of cash around are going to go with stability. Let's be honest, these days hardware is all a commodity. Whatever brand you buy is going to do the job, regardless of the feeds and speeds some SE is selling you on. But knowing a company is going to be around in 3-5 years to service your product goes a long way for some people.
Click to expand...

really? tell that to everyone that invested in Sun equipment, you can get a service contract from Oracle now, but that contract is going to cost you more than 10x what it would cost to just replace the equipment

Duck said:
In reply to C0m0dor and FLECOM, I think your comments are based on what you've seen with the existing NetApp controllers and OS. This product is an entirely new hardware platform and an entirely new OS built from the ground up specifically for use with SSD Flash. While I agree that the proof is in the numbers I would suggest that you're rushing to judgement without seeing those results.

Full Disclosure I work for NetApp
Click to expand...

I don't doubt it's good, and I don't doubt it's going to cost more than a fleet of luxury vehicles but work great... if anyone could figure out how to use it, including your own people... we had a customer that bought a bunch of netapp equipment and paid for installation/support, we had 2 netapp guys in our datacenter for like 2 weeks trying to get ONE shelf working properly, I laughed
 
Taco said:
The all flash arrays I see going out there are for VDI deployments.
Click to expand...

Which doesn't make sense to me at all. ZFS+Zeusram ZIL+Good SSD L2ARC+dedupe+correct amount of ram=crazy speeds & crazy reliability.

You can put together a fully hardware & software supported, fully redundant ZFS SAN for ~$35k-$40k that will support a very extensive VDI cluster.

The only thing I see all SSD arrays really benefit is SQL. The ability to order an all SSD array from a vendor is a big step in the storage world. In the world of SQL proven performance & ease of deployment are king, which is what NetAPP will achieve with this product. When you have multiple data-centers with multiple SAN's a standardized deployment plan implemented by an experienced storage provider saves days of labor and thousands in transportation costs. The ability for an administrator to "order this level of performance & implement it in minimal time" is where this product will really shine.
 
Last edited:
FLECOM said:
I don't doubt it's good, and I don't doubt it's going to cost more than a fleet of luxury vehicles but work great... if anyone could figure out how to use it, including your own people... we had a customer that bought a bunch of netapp equipment and paid for installation/support, we had 2 netapp guys in our datacenter for like 2 weeks trying to get ONE shelf working properly, I laughed
Click to expand...

And I don't doubt that your story is entirely accurate, however I'm sure I could find a similar story for every vendor out there. We all have issues from time to time and we make changes to prevent those issues from ever happening again. Judging an entire company and an entire product line based on one event seems a bit extreme.
 
bds1904 said:
Which doesn't make sense to me at all. ZFS+Zeusram ZIL+Good SSD L2ARC+dedupe+correct amount of ram=crazy speeds & crazy reliability.

You can put together a fully hardware & software supported, fully redundant ZFS SAN for ~$35k-$40k that will support a very extensive VDI cluster.

The only thing I see all SSD arrays really benefit is SQL. The ability to order an all SSD array from a vendor is a big step in the storage world. In the world of SQL proven performance & ease of deployment are king, which is what NetAPP will achieve with this product. When you have multiple data-centers with multiple SAN's a standardized deployment plan implemented by an experienced storage provider saves days of labor and thousands in transportation costs. The ability for an administrator to "order this level of performance & implement it in minimal time" is where this product will really shine.
Click to expand...

fine, but what about the support costs and time? are you going to build it for me for 40k and let me call you at 2am when a drive fails or something breaks, 24 hours of the day, 365?
 
Duck said:
In reply to C0m0dor and FLECOM, I think your comments are based on what you've seen with the existing NetApp controllers and OS. This product is an entirely new hardware platform and an entirely new OS built from the ground up specifically for use with SSD Flash. While I agree that the proof is in the numbers I would suggest that you're rushing to judgement without seeing those results.

Full Disclosure I work for NetApp
Click to expand...

Duck, thanks for the response. I've been playing both sides of the NetApp/EMC camps now for 15 years and manage both architectures in my environment (Currently multiple FAS8040 and VNX2-5500s). I'm glad to hear NetApp is getting off its duff and making something new as it has been rather stagnant in storage innovation (playing catch up) for a couple of unforgiving years now. I just hope what ever has been developed by NetApp will carry over to mainstream and not just top-tier performance models.

When it comes to pure-SSD environments, I question both EMC and NetApp as the cost is still wildly outside of most budgets. EMC's XtremIO for example is stupidly expensive and EMC is generally a tad cheaper than NetApp, at least traditionally. NetApp has to beat that cycle.

On the other note: I understand that SSD behaves differently than traditional disk and requires different stripe sizes/configurations, thus different architectures are often preferred. Which is also why so many migrations from both camps have failed (Recent failure by EMC). So what you are saying is 100% accurate.
 
gigatexal said:
fine, but what about the support costs and time? are you going to build it for me for 40k and let me call you at 2am when a drive fails or something breaks, 24 hours of the day, 365?
Click to expand...

1st point: did you read the rest of my post? In no way, shape or form did I even imply that an "assembled SAN" is by any means a one-size-fit-all solution.

2nd point: If you have a hardware contract and a software contract (ex, nexentastor), a ZFS SAN can have full, 24/7/365 support. Pair that with a proper setup that can handle multiple drive failures per array, hot spares & auto-replace settings and you have a partially self-healing SAN. If my SAN can withstand 3 simultaneous drive failures per array, and I have 12 hot spares, waiting 6 hrs for me to swap a drive (to restore the system to 12 hot spares after the resliver is done) is acceptable. If you don't have an in-house storage administrator, then obviously a ZFS SAN isn't a solution for you.

3rd point: a properly designed ZFS SAN is 100% redundant. You can set up a system that can take a head failure, switch failure and a full enclosure failure simultaneously and keep running. I am not saying that a solution from a big-box provider can't do it too, i'm pointing out that a proper assembled ZFS SAN can also. 99% of the time the ZFS solution is going to be faster & cheaper, even when taking support contracts into consideration. The downside is they take longer to set up, you have 2 support vendors, and "the next guy" might not know what he/she is doing with it.

The all SSD product that NetAPP just put out will succeed in organizations with multiple SAN's and/or multiple datacenters, as well as businesses that don't have a storage administrator. Exactly NetAPP's target. The proof is in the pudding. I would be very interested in seeing some statistics with a very heavy SQL workload.
 
Last edited:
I didn't catch you were talking about Nexentastor. Everything I've said is moot then. I've had plenty of experience to know that a well designed ZFS system with a shit ton of ram should do just fine. At least for my client level work loads, I don't have any first hand experience at the enterprise. But hey it's good enough for high iops science installs like Livermore labs.
 
FLECOM said:
really? tell that to everyone that invested in Sun equipment, you can get a service contract from Oracle now, but that contract is going to cost you more than 10x what it would cost to just replace the equipment
Click to expand...

Always exception, but by and large the smaller names are going to collapse more often than the Suns. Beyond that Sun's decline was a long time coming and, aside from Cisco, no one out there is mirroring Sun's awful support contract structure. From my anecdotal experience the primary factor behind people going with solutions other than Cisco is to get away from their awful support contracts.

Which doesn't make sense to me at all. ZFS+Zeusram ZIL+Good SSD L2ARC+dedupe+correct amount of ram=crazy speeds & crazy reliability.
Click to expand...

Who wants to hire a staff to both develop the architecture around this and then support it after that fact? And the first time it's down get questioned by some CxO on why they didn't invest in an industry standard solution? You'd save money and gain job security spending through the nose for this NetApp solution.

Why do people pay Red Hat money when CentOS is free?
 
bds1904 said:
The all SSD product that NetAPP just put out will succeed in organizations with multiple SAN's and/or multiple datacenters, as well as businesses that don't have a storage administrator. Exactly NetAPP's target. The proof is in the pudding. I would be very interested in seeing some statistics with a very heavy SQL workload.
Click to expand...

No shit, I agree it's not made for whatever podunk company employs you. Most people who buy this have a specific purpose for it in mind, it's not meant for every app/role in the data center. I've sold many flash arrays, it's never been their only array. Hell in VDI alone it's generally 2 SANs, 1 for VM disks and one for persona data.

On the SAN admin comment, no clue what you are talking about with that tidbit.
 
Duck said:
And I don't doubt that your story is entirely accurate, however I'm sure I could find a similar story for every vendor out there. We all have issues from time to time and we make changes to prevent those issues from ever happening again. Judging an entire company and an entire product line based on one event seems a bit extreme.
Click to expand...

probably, I don't doubt that at all, but NetApp seems to subscribe to the:

"lets make it incredibly complicated for no reason except maybe job security"

I have worked with many disk arrays that are stupid simple to install, NetApp just makes me want to rip out my eyes

Don't get me wrong NetApp isn't the only one guilty of this, but it just drives me crazy that all this stuff is so much more difficult than it needs to be...
 
Maybe I should reconsider my 8040 cDOT configuration now....
 
FLECOM said:
probably, I don't doubt that at all, but NetApp seems to subscribe to the:

"lets make it incredibly complicated for no reason except maybe job security"

I have worked with many disk arrays that are stupid simple to install, NetApp just makes me want to rip out my eyes

Don't get me wrong NetApp isn't the only one guilty of this, but it just drives me crazy that all this stuff is so much more difficult than it needs to be...
Click to expand...
We'll have to agree to disagree on your first point, but I do agree with you that many storage vendors make the administration and operation of a system far more difficult than it needs to be.

Goaty said:
Maybe I should reconsider my 8040 cDOT configuration now....
Click to expand...

Quite honestly I've yet to see a workload that "requires" SSD or super fast controllers. SSD/SAS hybrid solutions are more than fast enough for the most workloads and the whole All-flash thing is selling mostly because of the "sexy" factor. The 8040 is a heck of a performer and properly balanced with spindle IOPS behind it it I'm sure it will work just beautifully for you.
 
what infrastructure is needed to support a ZFS san/das? it is just a filesystem that operates transparently to the apps that use it, I would think the only app side consideration would be whether or not you're doing a lot of synchronous or asynchronous writes
 
"first shipments of FlashRay" is an interesting statement.

There are no support docs, either for operations OR for sales/SE on NetApp's website. Basically they've very carefully qualified who the first units are going to, and are supporting them and the sales cycle directly.

This product is only very very gently baked.
 
guess the NSA needed faster storage?
 
gigatexal said:
what infrastructure is needed to support a ZFS san/das? it is just a filesystem that operates transparently to the apps that use it, I would think the only app side consideration would be whether or not you're doing a lot of synchronous or asynchronous writes
Click to expand...

A zfs volume is presented to a client device as a blank lun. That lun can be presented via iscsi, fc or fcoe assuming you are running an opensolaris based os on the zfs server. You can also use nfs or smb to provide read/wite access via a network.

I personally present luns to my esxi servers via 4gb fibre channel. I have 6 luns total presented to my 4 esxi servers. 1 lun is dedicated to each server for an esxi boot disk, the other 2 are shared storage for vm's. All 4 servers see the 2 vm data lun's.
 
Duck said:
We'll have to agree to disagree on your first point, but I do agree with you that many storage vendors make the administration and operation of a system far more difficult than it needs to be.



Quite honestly I've yet to see a workload that "requires" SSD or super fast controllers. SSD/SAS hybrid solutions are more than fast enough for the most workloads and the whole All-flash thing is selling mostly because of the "sexy" factor. The 8040 is a heck of a performer and properly balanced with spindle IOPS behind it it I'm sure it will work just beautifully for you.
Click to expand...


Yeah, on Flecoms first point id disagree as well. I have been working with netapp in the enterprise for 10+ years. They couldnt be easier to work with. Nothing is complicated. Whether you are CLI with ontap or using the old web mgmt or newer system manager. One unified management for all your SAN/NAS deployments. But, for an all flash array, even a company as large as the on i work for would only have a few use cases for. Especially at the current SSD prices(which are still far better than just 2 years ago).
 
theducks said:
"first shipments of FlashRay" is an interesting statement..
Click to expand...

They're talking about the first 25 customers who beta tested and were given the option to buy. NetApp doesn't expect many people to buy this first iteration of FlashRay because it's single controller at the moment and possibly also lacks snapshot capability (unconfirmed). Of course NetApp will add these features as soon as possible.

Why would they release an unfinished product? you might ask yourself. NetApp came a bit late to the All-Flash-Array party. Of course NetApp has had PCI flash and SSDs for a long time now, but it was impractically expensive to build an entire NetApp array from SSD. Then some stuff happened in the competitive landscape (Pure, XtremIO acquisition by EMC, etc) that forced NetApp into action sooner than they probably wanted. Releasing FlashRay isn't about the money, it's about sending a message.

At this very second if you think you're a FlashRay candidate you should probably be looking at NetApp EF series with SSDs, unless you already have FAS in your ecosystem, then simply add a FAS array with only SSDs attached and ONTAP will tune certain parameters and functionality for better reliability and performance and become an "All Flash FAS"
 
4saken said:
Yeah, on Flecoms first point id disagree as well. I have been working with netapp in the enterprise for 10+ years. They couldnt be easier to work with. Nothing is complicated. Whether you are CLI with ontap or using the old web mgmt or newer system manager. One unified management for all your SAN/NAS deployments. But, for an all flash array, even a company as large as the on i work for would only have a few use cases for. Especially at the current SSD prices(which are still far better than just 2 years ago).
Click to expand...

dunno, maybe the guys they sent out were just stupid? who knows, I'll never use any of it personally since their stuff requires signed drives
 
You must log in or register to reply here.
Back
Top