Pure FlashArray//X NVMe Review

We're able to get higher-density workloads on the same infra, with a smaller footprint


What is our primary use case?

We run all our Tier 1 and Tier 2 storage on it, our VMware infrastructure, all of that would be running on the //X70s. Our database workloads are on the new Xs.

How has it helped my organization?

The feedback I've had from the storage admins is that it's the simplicity. It's easier and quicker to allocate storage for us. We're able to get higher-density workloads on the same infrastructure, and we have a smaller physical footprint than we used to.

It helps simplify storage management a Database Admin perspective - there was a lot of thought that went into the size of the disks, how we allocate those, etc. Especially when doing maintenance or expanding disks. There was always performance issues when expanding disks on the old infrastructure arrays and allot of care had to be taken on offset sizes etc - Whereas now, we're finding that we can expand disks without having to remove and add disks again. That simplifies those admin tasks without any performance implications. From a Storage Admin side they love the ease of use and visibility the systems gives them.

What is most valuable?

  • Simplicity
  • Performance
  • Visibility

Those are the three most valuable features that I've observed.

The visibility you have on what the frames are doing, even your phone is amazing, very detailed information about the environment in real-time.

What needs improvement?

One area that I haven't been sold on yet is the POD replication. We've seen that when we create a POD in synchronous mode, it increases the latency. We have another array that our latency isn’t affected by replication and when compared to the X array with a synchronous POD setup – it’s faster to respond to the system, but as soon as the POD mode is Async – the X is faster again. We not talking huge numbers – with the POD in synchronous mode with talking under 1ms for most IO operations.(Distance dependent to the other array)  I do feel there is some gaps in my understanding of the POD setups in detail so perhaps its tuning gap.

What do I think about the stability of the solution?

The stability has been excellent, there was one minor issue when the M70s were first released – but they have been 100% stable since.

What do I think about the scalability of the solution?

The only aspect that I can talk to is the benchmarking that I did from a Database perspective, as we purchased the X70 for our Database workloads. What we want to see is as we increase workload the latency should remain stable. We are able to push the X70 way past our current needs from a throughput and IOPS perspective – without any degradation on latency. As we ramped up more machines into the tests, the only limitations we ran into were switch ports and FCoE saturation – the X had headroom.

How is customer service and technical support?

The technical support is very good. The teams are awesome. The guys in South Africa, and they're very, very good.

Which solutions did we use previously?

If you're looking at it from a database perspective, you obviously want to make sure that you can scale your workloads without a degradation in performance. What we've seen on the X’s, and the Pure’s in general, is that we can scale beyond what we need without any degradation in performance. The real-time monitoring I’ve seen via your phone is a differentiator alone – to get a heat map in the past was a big process and took place after the event.

In terms of the Predictive Performance Analytics, I haven't seen it myself. I know that the guys have just recently started looking at a lot of analytics.

Although I don't admin the solution, I would give it a ten out of ten - from the vendor technical expertise and helpfulness, it's simple, reliable, quick and predictable at the same time I’m expecting lower costs for us once we fully migrated.

How was the initial setup?

I didn't do the setup, but everything I heard was that it was straightforward.

What about the implementation team?

We used the main guys in South Africa, Data Sciences. They do everything, top to bottom if you want or with you teams. Our experience with them has been awesome.

What was our ROI?

I do know there has been a reduction in the total cost of ownership, although I can't say how much as we in a transition from our old arrays, but I do know the cost per gig is lower, the dedupe on the array helps drive this cost down – as well as the physical size is much smaller and uses less power.

Which other solutions did I evaluate?

The vendors at the time would have been Dell EMC and Hitachi.

What other advice do I have?

It depends through which lens you're looking at it. If you're looking at it from a database perspective, you obviously want to make sure that you can scale your workloads without a degradation in performance, your latency times. What we've seen on the Xs, and the Pures in general, is that we can scale way beyond what we need to without any degradation in performance. We don't need to sacrifice any of the performance as we scale up or scale to the side. There aren't many vendors that we looked at that can scale to the size of the operations that we needed, from an I/O-testing perspective. In real, day-to-day, we don't run that hard. But if we need it, we can. It's there. 

In terms of the Predictive Performance Analytics, I haven't seen it myself. I know that the guys have just recently started looking at a lot of analytics, but I haven't seen it myself.

Although I don't admin the solution, I would give it a ten out of ten. It's been awesome to work with. It's simple, it's very reliable, it's very quick. And we get excellent dedupe ratios on the machine without a sacrifice in performance.

Disclosure: IT Central Station contacted the reviewer to collect the review and to validate authenticity. The reviewer was referred by the vendor, but the review is not subject to editing or approval by the vendor.
Add a Comment
Guest

Sign Up with Email