What is our primary use case?
In our enterprise, we support high-performance computing for academic research. So our flash systems are used for scratch temporary job data. We also use them for VMware support in our VMware enterprise infrastructure.
How has it helped my organization?
This solution allowed us to migrate all of our VMware infrastructure from direct-attached storage to NFS, which is hypothetically slower. But the ability to expand infinitely and quickly has been very important to us. So, for the VMware side, it's been very convenient. If we need another 50 terabytes or 100 terabytes of space, we just add it to our NetApp volumes, increase the export size to the VMware cluster, and we have more space. It's very easy for our VMware administrators.
And for the cluster side, it's just flash. It's NFS. We use it for NFS v3, v4, v4.1. It's a relatively quick NFS-based data store. That's what NetApp is really for. It serves NFS.
What is most valuable?
The most valuable feature is the solution's very well-defined and mature management interface. Their APIs, the GUI, and the CLI are very well thought out and consistent. So for us, it's really the management. It might not be the fastest solution out there, but it's extremely reliable and easy to administer. It requires very little personnel time.
What needs improvement?
Cost and speed are the biggest concerns. NetApp systems they're not slow, but they don't match the raw IOPs of some other parallel file systems design. The underlying architecture of the NetApp system isn't as conducive to pure raw speed, whether it's data I/O or metadata I/O, compared to some of their competitors. So that's a significant weakness. And they're extremely expensive. But that expense is justified in terms of management.
Cost and speed are the biggest concerns. The AFF A800 is an all-flash array, so it's not slow, but it doesn't match the raw IOPS of some other file systems that are designed for high parallelism. The underlying architecture of the AFF A800 is not as well-suited for pure raw speed, whether it's data I/O or metadata I/O. This is a significant weakness compared to some of the competition. And of course, the AFF A800 is also very expensive.
There are many features I'd like to see included, but most of them are niche features related to our use of QTree or other specialized aspects. There could be more types of migrations that can be done through the management interface, rather than relying on a client.
For how long have I used the solution?
We have an AFF A800, and AFF A200. We've been using the AFF all-flash arrays for about five years.
What do I think about the stability of the solution?
I would rate the stability a ten out of ten. We're achieving five nines at least.
What do I think about the scalability of the solution?
Scalability is relative. For example, if you need 500 petabytes, you're going to rate it as a two or something. For a moderate-sized enterprise or educational institution like where I work, I'd probably rate it as a seven out of ten.
You have to upgrade in increments, and it's not completely linear. There are some limits to namespaces and such. Nevertheless, I have expanded the system numerous times to a multi-petabyte system with minimal effort.
How are customer service and support?
Sometimes, it takes a while to get somebody competent on the other end of the line. They do have engineers in multiple time zones around the world. However, their level-one support is not always the best.
How would you rate customer service and support?
How was the initial setup?
The initial setup was very straightforward.
When we make a new NetApp purchase, a NetApp representative comes in and installs all the equipment. In fact, it's so easy for me personally that I handle most of the software deployment myself. NetApp is willing to do that.
But, we do have a very strong VAR that we work with. In the past, we purchased directly from the NetApp Factory, basically. And a couple of years ago, we moved to working with the VAR. They provide excellent support. But the system itself, I feel comfortable making major architectural changes to the system on my own just because they have very good documentation. Their support is good if you find the right person. And, the robustness of their CLI, API, and GUI is so great that I feel very comfortable making changes to it. So, it's very, very easy to deploy and maintain.
What was our ROI?
We have seen an ROI because we choose our use cases carefully. For example, for an academic institution, affording petabytes of all-flash storage from NetApp is not feasible.
However, we can afford it from some of their competitors. So, we choose NetApp, where their management is most applicable and where we don't need massive amounts of data. We have seven petabytes of NetApp spinning disks, but we couldn't afford to do that with Flash. Some competitors make it more affordable to have an all-flash system. So it's about choosing where you get the ROI.
What other advice do I have?
I would suggest to look at the ROI. If you need something that's extremely reliable, very easy to manage, has fantastic support from the vendor, and you can afford it, I would definitely recommend it. Unless all your requirements are solely based on throughput, it may not be the solution for you. But if you need something that you can manage with a minimal number of people and it's going to provide you with an extremely reliable, well-supported, and robust infrastructure, then I would definitely recommend it.
For us, reliability is the most important aspect. I'd have to give this solution a nine out of ten. Over the eight years we've been using NetApp with ONTAP, we've never lost a bit of data, and we've only experienced a few minutes of downtime in that entire time.
Which deployment model are you using for this solution?
Public Cloud