NetApp AFF - All Flash Reviews

The valuable features of All Flash FAS, as well as the ONTAP, are the ability to have the storage efficiency of compaction, compression and in-line dedupe; being able to maximize the original investment for additional components to our Epic environment; also being able to SnapMirror and FlexClone to refresh our Epic instances in a streamlined manner that prevents us from having to do a lot of file copy.

We have consolidated on to UCS and Nexus on NetApp. The FlexPod model has made it very easy for our support staff. We don't have to support a large number of other types of vendors and such. Support from the two partners, including VMware, makes it easier for us to be able to manage it and get to the root cause of problems that we have encountered.

The way that we're using All Flash and FlexPod with All Flash is for an Epic environment. Because Epic dictates how they want things done, all the features that we're getting from ONTAP, for all the things that I’ve mentioned, really meet our needs.

One of the areas in which we are going to be looking at All Flash is for our MetroCluster environment. There is one feature that I would really want at this point: They are only talking about an eight-node MetroCluster for NAS, so I would want that also for SAN. We're very interested in moving towards All Flash for that over the next couple of years and we would definitely want to make sure that we can scale the MetroCluster beyond just four nodes; two nodes per site.

We've been up and running for over a year in production with Epic and we've had zero down time. We have been able to upgrade without impact to the application.

It's very scalable. The cluster will go up to eight nodes currently, and more. We can easily scale it, as well as being able to replicate it to our other data center.

We looked at VCE or the EMC equivalent. That was really the main consideration. HP was also considered, for 3PAR. Epic's recommendations for storage played a key role in the decision. Their comfortability with ONTAP and their flash. At the time, they were not very comfortable with the XtremeIO that was being offered up, what has happened with that product and the instability with that product. We're very glad that we did go with NetApp.

There were other factors too. Cost seemed to be lower with NetApp, but in the grand scheme of things the hardware component was a much smaller amount in the budget when you look at the entire cost of implementing Epic. Definitely cost plays into it. The elegance of the solution is another big key. The manpower required to administrate VCE and to patch it really requires someone to hand hold the entire upgrade process, whereas with NetApp it's a lot more flexible, it's intuitive and doesn't quite require that same level of administrative work.

I don’t recommend looking at any one specific vendor, but one of my biggest concerns is having a lot of different components that are brought together. I like having things simple, lowering the number of interdependencies for the storage platform; whatever makes that less likely and less prone to have failure. The other vendors out there that we have looked at have always been bringing different solutions together and having it be a construct of many parts. That played a big role; the most important thing for this hardware to do was to stay up and running, and required the least amount of manpower that we would have to hire and administrate. Ultimately, that's why we chose NetApp. It's an elegant solution.

Going from mechanical disks to flash was a huge benefit, speed-wise. A lot of big BI reports that we were running that would take hours, we can do in 10 minutes now. That was really the biggest impact. The user saw it immediately, the benefit of it.

We're an electronics manufacturer. Shop floor people rely on these reports to make decisions throughout the day and we can, instead of having a once-a-day refresh, they can almost get it on demand.

I would just like to keep seeing improvements in performance and efficiency, which it seems to have been doing between 8.3 and 9; it's getting better with every release.

The user interface is a lot better. I think in 9, we do a lot of command line stuff, so I'm not into the GUI too much.

We’ve been using it for six months. It's fairly new.

We've had no issues stability-wise; we've been a NetApp customer for 20 years and just rarely have any issues.

Technical support is getting better. Historically, it's been painful. We had some challenges with support but over the last couple of years, I think it has gotten a lot better. We have a really good SE now that we leverage and our partner's really good as well.

We knew we needed to invest in a new solution because we lease our equipment and it was due for release return.

Initial setup was easy. We had one small system. We have a lot of FAS systems; we have a single AFF right now. It's an 8080, with just one shelf. It was a very simple setup. We're familiar with cluster mode already.

Rack it and call it good.

We looked at several other options:Pure Storage, Nutanix, and Tintri.

We chose NetApp because all of our other storage systems are NetApp. We just liked being able to leverage the knowledge that we already had in house. We didn't see a lot of value in having another siloed storage system out there that we had to support. Price-wise, NetApp was very competitive, more competitive than we had expected.

Do it. You won't regret it.

I like the product, and am quite happy with it.

When I choose a vendor, some of the criteria I look for are support, the ability to execute and a mature product line.

For us, the most valuable features were the SnapMirroring, deduplication, and inline compression. Now with 9.0, the compaction system, that's actually the big thing that sold us on it besides just the price in general. It was a very well-priced system for what we got. The data dedupe and inline, we're getting substantial rates. I think it's about 60-65% in general. That's a massive savings over what you would get if you didn't have any of that stuff.

We have a job system that runs all the time; people can run what they call campaigns. It drastically increased performance. It decreased times by three times the amount. The amount of the CIFS shares increased from about 128 Mbps – it was only a 1-gig line anyway, to a 10 gig – to about 800 Mbps. The engine actually can't pull enough and it has caused a little issue here and there, because it's basically causing a race condition. We've had to program around race conditions because we haven't had a system that was this fast.

It saved us a lot of time as well, substantial savings.

If they could do the tabbing for the nodes, that would be spectacular. On 9, they offered more insight, so I can't really say that. We haven't upgraded both nodes. We have HA pairs, and one of them is still running 8.3.2. We upgraded our DR solution to nine first just to see if it causes any issues. So far, we haven't seen anything. They have a lot more insight into that; I wish they would have it on 8.3.2 but, what are you going to do?

So far, stability’s been excellent, and the update process was actually incredibly painless. We've upgraded twice now and I am surprised that it didn't cause any issues at all. Usually, you have to have some kind of user intervention. For this product, you just throw the image on there, click update and it's done. You come back about an hour later and you're happy.

The GUI is really good, but if you don't find the option in the GUI, then the CLI is amazing. You can hit Tab and just tab out. The only thing is, they haven't done that on the nodes themselves but I was told they're working on that.

We don't have that big of one yet. We originally quoted out a system of eight nodes, and it was going to be something like 12 GBps. That seemed like substantial amounts, considering what everyone else quoted. However, it actually was going to come in at about the same price for the AFF compared to everyone else's quotes for disks. The reason they went with it is because of the trust with the vendor they were currently using and they just didn't want to leave.

We were using NetApp before, but we evaluated EMC, IBM, HP, Pure Storage, XtremIO and Nimble.

It came down to XtremIO and NetApp. NetApp offered much, much more storage. And the cost difference to buy XtremeIO was huge compared to NetApp. NetApp just totally blew it out of the water on price. We got something like five times the storage for the price. It was really worth testing on that.

Try out what you actually want to do, because that's actually the problem we had; some of our people swore up and down that NetApp wouldn't be able to do compression at the new rates that they got, or that we got. They said that Oracle doesn't compress and so on. We ended up getting them to stick some of their machines on our NetApp, and we showed them that you actually do get it.

We actually bought ours and then we tried to show those other people before they got to the bidding table for theirs. They didn't really want to listen to the facts. They went with IBM. I wouldn't say they were not unhappy or anything. They realized that they could've gotten a lot more if they just went with our ideas instead of their idea. Actually, I was told it was more of a management thing; they actually didn't even want IBM, they wanted Oracle. It all comes down to what the boss says.

We use it for our high-performance requirement, low-latency requirement databases. That's at the core of the retail application; what we've connected are non-virtualized AIX databases running Oracle and DB2.

The valuable features are the high performance and low latency that the retail application requires.

We have a six-node NetApp cluster for our regular 8080 FAS systems, and we have two cluster nodes with All Flash FAS, so it enables us to manage this high performance, low latency, application workload in the same fashion as we treat all of our other data; the SnapMirrors, the SnapVaults, the snapshots, the user control. We can use the same toolkits for everything.

It provides ease of management and the ability to manage it as one unit.

One of the limitations we found with the All Flash FAS, using ONTAP version 8.3.1, is that we could not do foreign LUN import directly to the 8040. We had to stage that through the other cluster node before they ended up in the regular place. There were some limitations and some gotchas on the initial migration path.

It was installed about a year ago, and the full workload was deployed around March of 2016.

Stability has been good. We have had no stability issues at all whatsoever.

There have not been any latency issues.

So far, we have about 40 TB of raw space. On top of that, comes all of the inline compression, the dedupe and all of those features and functionalities. It's not a huge system but it's IO intensive. It's on the order between 40,000 and 80,000 IOPS.

Technical support is pretty good. We don't call on support all that often. We're well handled in house. For the AFFs, we haven't really had too many support issues at all.

We replaced 2 E-series, and the decision was made to get a larger cluster mode system with two nodes of All Flash FAS, specifically so it would be one cluster, and could be managed as one cluster.

I was involved with the initial setup. It was fairly easy; a little bit different from a traditional FAS but very well managed by NetApp as the install engineers.

I don't see the price of it, but my company must think that it provides plenty of value at whatever price we are paying for it.

Before choosing this product, I did not evaluate other options. We went with NetApp because we were already using NetApp. The strategic direction at the higher management level was to go with NetApp.

The mix we currently have with 8080 for traditional spinning disk workloads for VMware and file sharing – those kinds of things – mixed in a cluster with the All Flash FAS system, does everything we could possibly ever ask of the system. One set of management tools, one set of skills to manage all the capability, I think it’s an excellent solution.

It's the fastest that we can get through NetApp. We're deploying all these high-demanding applications and it's the best of the best, so of course we went with it, being a big NetApp customer.

In the transportation industry, we have a lot of demand for analytics and on-demand data, big data, and AFF provides what we need in terms of the quick read and write.

We spend less time thinking about performance and more time being able to worry about actual problems and the customer. Of course, the customer is the most important part of business.

I would rate it higher if it didn’t cost as much. It's a bit pricey.

Other than that, it's got what we need. I don't really have any suggestions for what it doesn't offer. I'm happy with what it has. I think it's only gotten better, especially with the 8.3 release and obviously ONTAP.

Both the GUI and the command line have exactly what you need and I have no problem navigating either of them.

I have not encountered any stability or scalability issues; absolutely no issues whatsoever. The only issue is how fast we can put them in.

We have rarely required technical support. Usually, it's just a one-off type thing and I've never had any issues getting what we needed out of them. They're always knowledgeable; never had a problem.

I was not at this company before they started using AFF.

I was not involved in the initial setup.

I have not evaluated other options.

Be sparing in capacity and don't just throw it around. Storage is cheap now, but AFF, as I’ve mentioned, is not cheap, so be cautious in how you use it. That’s something that needs to be analyzed before you start the process. It’s the kind of good homework to prepare. I think that goes for anything, but doubly for expensive flash. Just make sure that's really what you want and what you need.

When I’m looking at vendors, I need them to know exactly what they're selling to me does. I need them to know the competition, so they're offering a fair comparison and not just offering a vendor lock-in type situation.

The speed and stability are the two most valuable features. I've definitely have seen a huge increase in speed and bandwidth from when we've put it in production; it's been great.

We use a lot of bandwidth in our company, so user experience is really improved because of that. The IOPS are high enough that there's no latency anymore.

I’d just like to see them continue down the road of increased storage capabilities, bigger SSDs, and bigger flash. That's a problem for us; we use a lot of storage, PBs of storage. We definitely need to keep leveraging, expanding and increasing.

I haven’t rated it higher because we haven't had it enough to vet it. So far, it's been great; haven't had any issues at all with it. Then again, we've only had it in production for a few months. We just need more experience with it.

We have been using it for 6-9 months.

Since we've had it in the last 6-9 months, we’ve not had a problem at all with it. It's been great.

We haven't really scaled it yet. From what I've seen, it looks like it won't be a problem if we need to go down that road.

We have not had to use technical support yet thankfully, other than original installation.

Before we implemented the All-Flash FAS, we had, and still have, a mixed environment of mostly Hitachi and Dot Hill. We also have LTO tape storage that we implement. I'm all over the board.

Even though I was not involved in the decision process to invest in the All-Flash FAS, we selected NetApp and the All-Flash FAS because of the speed of the flash, the reliability and the stability; it stays up.

Initial setup was just racking and stacking, basically; that was about it. Rack and stack is very straightforward; we had help from technical support to cable it. From that point on, it was pretty easy.

I wasn't involved in the decision-making process, so I'm not sure who else we were looking at, at the time.

I would definitely recommend it. I think it's a great road to go down. Like I’ve mentioned, I haven't had any problems with it. The two things we were looking for, it does excellently.

The most important criteria for me when selecting a vendor to work with are availability and knowledge: if something helps me go down the right path and pick it, if someone gives the pros and cons for everything we need, and be able to get a hold of them when I need to get a hold of them.

The most valuable features are inline deduplication and compression.

It's enabled us to move all of our database tempdb locations to the AFF and save 70% on storage costs.

It's greatly improved our tempdb access. In our environment, we tend to use and abuse tempdb and as such moving our database tempdb locations over to that device has improved performance quite dramatically.

Beyond the setup complexity issues I’ve mentioned elsewhere, most of the things that I wanted to utilize – transparent vol migration, transparent LUN migration, reassignment of volumes from one HA pair to another – have all been solved with either cluster mode or 9.0. Those are things that we do on a daily basis.

It's just as stable as any other NetApp device, that is, very stable.

We haven't done a whole lot of scaling yet in our AFF solution. However, it appears to be quite scalable and now, with ONTAP 9, you can go up to 12 SAN nodes; it's been quite dramatically increased.

In general, I have not used technical support.

I was involved in the decision to invest in the All Flash FAS. We decided to go with an all-flash solution for our ESX environment specifically because we had a business initiative to virtualize our database platform. In doing so, it was not performing as well as we would like on the spinning disk. Moving to an all-flash solution has dramatically decreased the OS latencies and increased performance of the OS, which in turn improved the performance of the overall application.

We were previously using a NetApp FAS with the 10,000-rpm SAS disks; the 2 1/2" ones, the little ones.

A cluster mode setup is quite complex, generally speaking, and quite involved; not as intuitive as I would like it to be. A one-click install would be nice, something where you can just have a GUI-driven system where you put in the IPs you want to use and the interfaces you want to install them on and call it good.

We went to several different vendors; the two top contenders were NetApp and Pure Storage. Ultimately, we went with NetApp for a couple of reasons: 1) the scalability of the clustering system, and 2) we're already a NetApp shop and so adding on to an existing NetApp environment made it quite a bit easier, especially with replication and data management techniques that NetApp already employed. The storage grid that NetApp is deploying across the infrastructure makes transparency and migration of data from one device to another environment a lot more seamless. Whereas Pure Storage is fast, NetApp is faster and their devices are data islands. Taking a step back, we just didn't feel Pure Storage was going to work for us in the long run.

Our only experience with Pure is the demos that they brought us; nothing more than that. We talked to several of their customer bases and although they claim a lot of nondisruptive operations, they tend to be disruptive.

We've worked with NetApp and it's kind of tried and true. We do upgrades, we do hardware replacements and everything is transparent and doesn't affect the users, which is really nice, especially considering we're a software-as-a-service company. The less we can take our customers offline, the better.

If you've already got NetApp, you can't go wrong.

It's a fantastic system and it's solved a lot of our issues for application performance and it's probably one of the best storage systems I've worked with and yet the only reason I dock it a few points is because there's still the future. There's problems we have yet to solve, unknowns. There's always going to be issues in the future, we just don't know what they are yet, whether it's NPS storage, whether it's migration to the cloud. We have a business initiative to move to the cloud.

There are a few oddities, only because some of our systems are legacy. We have the 7-mode system, which is our primary platform, and moving to the cloud is a little bit painful for that system. You have to spin up the 7MTT tool to get it to transfer the data and the 7MTT tool was not designed with cloud in mind. It was designed for migration of a 7-mode system to a cluster mode system within the same environment. When you're trying to move it from one environment to another environment to a different site with a whole new IP scheme with a whole new infrastructure, it's just a little bit on the kludgy side. There are things that don't make a lot of sense on that front. For example, it limits SnapMirrors to four per cloud ONTAP instance. We want more than that. We want hundreds. By default, the cloud instance is supposed to support 50 and yet we can only do four with the 7MTT.

The most important criteria for me when selecting a storage vendor to work with are going to be speed, reliability and support. The better the support is, the easier they are to work with, the more likely we are to choose them.

NetApp All Flash FAS:

We're also using FlexClones, and we're able to flex clone our database to a test environment and a dev environment. We're able to keep it all underneath one storage system, so we don't have to manage multiple arrays to run test and dev.

Learn about the benefits of NVMe, NVME-oF and SCM. Read New Frontiers in Solid-State Storage.

NetApp All Flash FAS:

The speed and low latency are the most valuable features. My customers or users noticed an improvement in our EMR application, once we moved our SQL environment over to the AFF.

I think the OnCommand System Manager has been excellent. I like the newer version myself. We're not on ONTAP 9 yet, but I do love 8.3.

Pure Storage M20:

It's just simple. There's no clustering. You take that complexity away from NetApp, you get rid of the clustering. It’s a dual-node controller system. You can have dual or single aggregates, whatever, the same thing. But they don't do clustering. If you wanted to mirror that data off, you have to purchase another Pure, plug it into the expansion port and basically mirror between platforms. Whereas in the ONTAP, your data is clustered, you've got HA failover. You still have HA failover in Pure, but it is just on the controller only.

NetApp All Flash FAS:

I'm not involved in the price-making decision. I just throw the number at the manager and say, "Hey, this is what I want."

We're going to take a good look at Solid Fire for SQL environment.

I’m not rating it higher because when we piloted our AFF against a Pure Storage M20, we were getting much better deduplication out of our SQL database on the Pure product. I’m told that this has been improved in ONTAP 9. The deduping compression ratios are more on par with what Pure has been able to do. It's not an end-case decision. We have plenty of storage available. Our database is growing. We'd love to keep as small of a footprint as possible, but we still have overhead room in case it does expand beyond what we're expecting it to. Nonetheless, in the future, I'd like to see better deduplication out of SQL. That’s difficult to do; I get it.

Pure Storage M20:

I don't know what could be improved at this point. I haven't used it enough to know where they really are lacking in anything. It's fast, it's very easy to set up, it's very easy to maintain.

Nonetheless, there is no clustered data. Your data resides in a single point, so then it's up to you to mirror it, replicate it, copy it, however you do it; DoubleTake replication or if you buy another Pure product and do their onboard replication. I guess It all depends on your pocket book, really.

I would give it four stars because it's pricey. I do believe Pure was more expensive than the NetApp when we were pricing last time. Of course, that really varies on what time of the year it is. I think NetApp end of year is March or April. I'm really not sure when Pure's end of year is, but they call me every few months with a better price.

NetApp All Flash FAS:

It is very, very stable. I'm noticing a difference between it and my FAS 2230. Storage efficiencies are much faster. Deduplication is awesome.

Pure Storage M20:

We didn't see any issues with stability. Of course, we only ran a demo for 90 days.

NetApp All Flash FAS:

We're looking to scale it up because our FAS 3220 is coming up for maintenance renewal. We're thinking we'd probably be better off chucking a couple more shelves at the AFF, and running our vSphere environment off those shelves because we're not touching the controllers on that AFF. I hate to say they're running ideal with 2,000 SQL users on it, but it's running very well.

Pure Storage M20:

Scalability is pretty much the same as on NetApp, depending on what controller you buy, how many shelves you can attach to it.

NetApp All Flash FAS:

Technical support is hit or miss sometimes. Sometimes I get the run around. I have to go through multiple support engineers to help me out with whatever issue I'm dealing with at the time. Other times, they've been spot on: The first guy I get has said something like, "Oh I can fix that for you right now." I think it really depends on the complexity of the problem.

Pure Storage M20:

I did not use technical support for Pure.

Learn about the benefits of NVMe, NVME-oF and SCM. Read New Frontiers in Solid-State Storage.

NetApp All Flash FAS:

I was involved in the decision to invest in the All Flash FAS. We were modeling it against Pure Storage. We already had 3220 running for a couple of years, running VMware. I made the decision; I didn't want to split between multiple vendors. I wanted to keep it all underneath one hood. The AFF allowed us to do that. We could not put our SQL environment on spinning disk, obviously; not with the scalability that it's at or the number of users we have hitting it.

We were previously using Fusion IO cards, striped. They’re PCIe slot cards – some are on x8 slot, some are on an x16 slot – with Windows striped between all those cards. That was what I walked into when I was hired by AHN. They were using SQL mirroring. In the event of a system failure, they could always fire up the mirroring to resume production. Doing it with a NetApp has pretty much eliminated that all the way.

NetApp All Flash FAS:

I was involved in the initial setup, which was very smooth, very easy.

Pure Storage M20:

Initial setup was very straightforward. You plug in a management port, you plug in your iSCSI, your NFS or your fiber channel ports, and you're up and running.

Keep an open mind. Different vendors do different things in a different way. NetApp is highly complicated, it's very robust. In comparison, Pure's interface is about as simple as it gets. But they all support fewer protocols then NetApp.

The most important criteria for me when selecting a vendor to work with is approachability. I like people I can talk to; if you get overly technical and it's all technical garbage and you're not really a personal type person. I hate to say it, but I base a purchase off that. If I'm going to be working with someone for a number of years, I want to make sure it's someone I can relate to.