HPE 3PAR Flash Storage Review

The flexibility and ability to adjust to business needs is where its robustness is

What is our primary use case?

We target 3PAR because of the availability aspects that it brings with a synchronous replication. I work with a lot of medical and larger business organizations which are looking for the ability to run stretch clustering between data centers. If they lose a full data center, they can use this to flap over live without application downtime. This is probably the biggest thing that comes into play from an availability aspect. Then, there is also inherently a lot of the performance which comes with it. E.g., if I have a lot of high demand applications, it is one of those where the system, the all-flash array, the 8450, and even the 9450 that recently came out, can keep the latency and response time down.

How has it helped my organization?

A lot of it has to do with its ability to stretch between data centers. It has helped a lot for times when our customers do DR testing. Instead of having to spin down and spin up, I can do it live and seamless. Therefore, I do not have to schedule downtime with an organization. Especially maintenance on arrays, if I need to do some maintenance, it could potentially slow down somethings or even take things offline. 

With 3PAR, I can transparently sort of flap over to that other data center and do all the maintenance I need to do (even if it means forklift upgrading things). I can do this without having to take applications down. For a place like a hospital, which is open 24/7/365, it can't suffer downtime. That is why this product is one of those nice game changers. 

It allows me to do so much and worry about taking care of the clients, instead of how do I keep things up.

What is most valuable?

Its biggest feature is the ability to do a lot of stretch clustering. When I look at a couple of the other arrays, sometimes I have to put extra layers on top where I can do this natively. It works with Hyper-V, VMware, and physical servers. I can keep that storage up transparently when it flops over and also to the kind of the way that it integrates with other stuff in the portfolio, like a Nimble and StoreOnce, to offload like snapshots. Therefore, I am not eating up a lot of what we call Tier 1 data for retention, when I am trying to keep data for archival purposes. I can offload it to less expensive storage. 

It all works in concert using Recovery Manager Central (RMC). HPE coordinates it all, so it is more of a solution instead of products trying to do things together.

What needs improvement?

Some people are talking about getting NVMe drives in with faster flash. However, I think that is on the roadmap. I was at Aspire this year, and they were talking about the next chassis and they are ready for it. It is just a matter of getting them in. 

Feature-wise, with the InfoSight additions, there is a lot of the stuff missing in the intelligent interface. As they grow and push, a lot of it will not tie into Hyper-V. I have a lot of clients with Hyper V, so having that put into InfoSight because I have a lot of clients who run half-and-half or a lot on Hyper-V, especially a lot of schools with public domains. There has been a shift more to Hyper V because the features are really good now, and getting those analytics would really help.

With OneView, there are some challenges. When I set up the Peer Persistence, it is very hard to manage zoning from OneView for arrays. We have eight 3PARs out there right now, and we do all our zoning outside of OneView, which is the opposite of what they say to do. 

What do I think about the stability of the solution?

Stability is nice. A lot of other systems have two to four controllers. From a performance perspective that gets into availability, it can suffer parts failures. The nice part about 3PAR is it starts getting into the six to eight node controllers, if I want to scale up. It is built-in there, and what helps with that is they from almost a managed services, or service provider background, before HPE even acquired them. They have a great structure, where if I have different business units, I need to get granular access. I can create separate domains for clients, but then manage it a little bit differently. The flexibility and ability to adjust to business needs is where its robustness is. Sometimes that adds some complications to the setup and the configuration, but sometimes that is needed based on what the business is trying to do with it.

What do I think about the scalability of the solution?

If I need to step up performance-wise, I can add additional controllers. As far as different systems, if I need to go between the 8000, 9000 or the 20000, I can get into the multi-petabyte and still maintain massive performance, if needed. Therefore, I can start low, even with small businesses. I can look at a two-node 8200, but then if I need to go big, I can get into the eight-node 20000. 

It is a nice family of products. It adjusts for even the smallest client and largest business clients.

How are customer service and technical support?

Usually if it gets to the point where we can't solve it, we will getting on the horn with technical support. Recently, we did have an under the hood issue which came up, and we know where to go or who to contact. Normal support has been good, but if I have to, we can get with their team that is developing the code. 

HPE even hooked me up. There is one guy who is writing their Remote Copy Software, and there is an inherent little bug we found. We have a pretty complex solution, and unfortunately, somebody always has to be the person to find the bug. The nice part is how they responded to it. Their team all came together, and everybody has been real responsive, even to this day. The VP and their product manager are emailing me, and I receive email updates, even as recent as yesterday. They are staying involved and care about the client.

Reactive is normal support, you do not go there. What will happen, and this is what we are excited about, InfoSight from the Nimble acquisition has a lot more of that predictive information because that is where I am pushing my clients to shift. They should be at the right OS level, so we can get those heuristics in there, because a lot of that information is will help. If something is going wrong, then we can identify it ahead of time, because it is easier to prevent than it is to repair. 

One of the reason why I think HPE bought Nimble is not so much for the Nimble Storage, but for the InfoSight part of it. A lot of the intelligence and data center will help resolve those last little issues regarding, "Why did we have downtime?" We should have seen this coming.

There have been a couple bumps with some of the support stuff, but HPE needs to sort that stuff out and that is where I hope Nimble will help because they are well-known for their support. For example, there were a few goofy things with support where we were trying to do OS upgrades and HPE MyRoom failed. However, they would not get on WebEx because they are only allowed to get on HPE MyRoom. Therefore, we had to reschedule an upgrade three times because of it. It was one of those where it was no one person's fault. It was just a policy and procedure issues. So, I am looking forward to getting some of these things cleaned up.

I would rate the technical support as a nine out of 10.

Which solution did I use previously and why did I switch?

It is the availability aspect. Over the last few years, I have done a lot of shifts where people come off of Dell, Hitachi arrays, or even some other products within the HPE line, and they need either the performance or the availability. The availability that they want to put in is a lot of time a type of Peer Persistence setup. We are selling at least two 3PARS where they have multiple data centers, so if an array goes down in the same data center, the customer can lose a whole array and still stay up because they have peer persistence set up. That has probably been the single biggest selling feature: availability. Nobody wants to be the person in the news that is down or costing their shareholders money. Thus, availability has been a real big push, and 3PAR does availability really well.

How was the initial setup?

If you have never set up a 3PAR, it can be daunting. There are a lot nerd knobs. There are a lot of things that we can turn on and adjust. It is easy to get lost in 3PAR. This is where our organization gives a lot of our time, in the setup. 

A lot of times people play with buttons just because they are there. That is where you can get in a little trouble with 3PAR. This is because there is plenty of stuff to do. That is where we try to get ahead of the game, and help them with planning and architecture, e.g., here is what we will do and here is what we will set up and do, because you can get yourself into trouble playing with everything.

We have a pretty good routine down now where the last two arrays that I did were all-flash 8450s, including racking time, and it took less than probably a handful of hours before we were up and running. We did about eight shelves. It is not overly complicated. It is more the OS and configuration which take more time. Usually, most installs never go beyond a day. The rest of it is just fine tuning and adjusting to the environment, depending on the size of the array.

What was our ROI?

Even as we have upgraded the product, one client that I work with a lot has eight 3PARs. They had two of the old 10000 arrays, all three tiers of disks, and they traversed about three to three and a half racks each. Therefore, we had almost six and a half to seven racks of storage. When we went to the 8450s, we went to all-flash and were able to go down to a 16U a piece, so that is 32U total. That is almost a six rack reduction. 

They are out in New Hampshire, in a very green state and very green conscious. I spent an extra 30 days getting the proper numbers from the state of New Hampshire about power consumption, even the air conditioning, such as calculating BTUs an hour. The benefit was already showing that they would save over five years over $400,000 from power cooling. That does not even get into buying less racks.

They were going to more of a fixed pod structure, like a service provider. This was six racks that I do not have to put PDUs into. It is less hardware to maintain. It is less likely to fail because there are less moving parts. Obviously, there are the SSDs, which was a big part of a green initiative (less waste and power being used). 

What's my experience with pricing, setup cost, and licensing?

I do not sell 3PAR all the time because it may not fit in with everything a client is trying to do. It is more about finding the right product for the solution. 

Which other solutions did I evaluate?

The big ones who comes up is obviously Dell technologies, VMAX and PowerMax, when I am looking for availability and performance. Another one that comes up a lot is Unity when we are looking at Nimble. Unity is a big player too right now. 

Also, what usually comes up quite a bit, is just Nimble. That is another one of those things where if I do not want all the nerd knobs, just a simple, great easy product that performs well, and if I want to be less focused on watching the data center moving forward, Nimble is a nice thing. The only thing that is missing right now is that synchronous for availability. They do not do synchronous replication, everything is asynchronous. Therefore, they are missing that availability, but it is on the roadmap for them. If peer persistence is not needed, Nimble is usually a great fit.

They chose HPE 3PAR because of the scale and the adjustability of it. If I need to get bigger, I can get pretty large with the product and still maintain good performance. There are not many vendors that go beyond the four nodes and maintain performance. Having the ability to scale from four to eight nodes allows that additional performance, because I can put 100 SSDs behind two controllers, but I will only be able to see about 20 disks worth of performance because I will outrun the controller. Therefore, getting that in the throughput  helps, but also in its ability to do Peer Persistence, which is the availability aspect where I can lose a whole 3PAR and it is seamless to the host. These are the biggest things: the availability and the speed of it.

What other advice do I have?

They do great things. The system is great. It is just a matter of cleaning up some of the support stuff, then the enhancements that are not there yet. Hopefully, they are coming in with InfoSight. That is where it could be better.

Never have blinders; everybody has their favorite product, but do not turn away from keeping your vision open about a solution. Keep that in mind when you are looking at a product. 

Most important criteria for clients when they evaluate a vendor: A lot of that gets into supportability. What do they have for a track record? How easy is support to work with? How efficient are they? When things go wrong, I do not want be scrambling. How easy is it to get to support and get them on the line (an efficient use of support). 

As far as feature-rich, how does it work with everything from an availability aspect?

Everybody talks about backup, and lot of times, they are talking about it after storage. They should be talking about it together, because storage is part of backing it up. The business should be asking: Am I backing it up fast enough? Are my RPOs and RTOs inline with what the business SLAs are? 

When we start talking backups and the availability aspect to a lot of businesses, they do not seem to be defining their SLAs. They often do not have any. That is where we find that we are having more of a discussion which helps drive a lot of what we need to do. 

You do not want buy stuff, then say, "What can we do with it?" You should be defining what you want to do with it, then purchasing. That is a lot of how we are changing the purchasing process. 

**Disclosure: My company has a business relationship with this vendor other than being a customer: Partner.
Add a Comment