HPE Primera Storage System: First Impressions and Customer Install Testing
In this article
One of the functions of the ATC in St. Louis is to perform early access testing and create a sandbox for partners and customers to innovate and experiment.
We are pleased to be one of the first customers to kick the tires on HPE's newest flagship storage system, HPE Primera. Primera uniquely provides the mission-critical storage for customers demanding the highest levels of performance, resilience and scalability with the option for the customer to self-install the system.
In the face of changes in the industry to the quick, easy and agile infrastructures, HPE developed Primera to be easy to maintain and to install; seeking a similar experience as their Nimble customers have enjoyed for many years. The challenge was to do this with a mission critical storage system.
HPE successfully developed an easy startup/install program and instructions on how to execute it. It was our task to test this new installation procedure.
First impression
We were pleasantly impressed with the Primera storage system. Compared to 3PAR, Primera is much denser, saving precious data center space. We choose the Primera 650 model with 4 nodes and 4 RU of rack space compared to the 16 RU on 3PAR 9450. The 4RU controller has space for up to 48 SSDs.
With the largest capacity 15.36TB SSDs, one can have 737TB of raw and over 1PB of effective capacity in 4RU. We opted for cost effective option of 1.92TB SSDs and only populated 36 slots. We still ended up with almost 70TB of raw and over 100 TB of effective capacity — not a bad system for a lab!
Initial installation
Since there was only the 4U controller unit, the physical racking was easy, but 36 SSDs in a compact form factor made the system very heavy. I am glad our ATC Ops team was prepared to install it with their server/storage lift.
Each controller node has an IP management connectivity. It is not required to have all 4 nodes connected, but highly recommended. Of all the management connectivity, only one is active at a time. There are two ways to connect to the web installation wizard.
First, for a Windows platform, there is a small executable that discovers the Primera array on the network. Once the Primera system is discovered, one can easily connect to it via web browser. Alternatively, on a Linux platform, the process is even easier. One just has to connect to the following URL, https://primera-<system-serial-number>.local. This procedure utilizes multicast DNS.
We did find that there were some details missing in the installation guide. One detail was how to setup the networking on the management PC. Without that detail, we were unable to get the multicast DNS to work. Another detail we learned was that the setup PC's network setting had to be modified after setting the network on the Primera management interface. Once these missing details were discovered and addressed, the installation went very smoothly.
Our installation took almost two days to complete because we ran into some internal networking issues and largely because we didn't know some of the crucial details. Murphy ruled most of the process, and we found just about every pitfall. To be frank, some of the roadblocks were our own doing in trying to guess what the missing details of the instructions were. If we had known the details, it would have taken us only a few hours.
Good thing HPE's knowledge management, support and engineering team was there to see this firsthand and use our experience to make this easier for our customers next time!
Primera OS upgrade
We learned that the upgrade of Primera OS is very easy — perhaps too easy! With the design changes in Primera OS, the upgrades and patching are very easy and non-disruptive. It's because the Primera OS now runs in user memory space/mode vs. Kernel/System memory space/mode that ran in the previous generation.
The previous generation required the rebooting of the OS when it was patched or updated, which was a disruptive event. With the new architecture, the Primera system does not have this issue. We were able to patch and upgrade the system while hosts were accessing data. The whole process is couple of clicks in the management GUI.
Architecture
The new service-centric OS decouples the data services, which are typically tightly coupled in monolithic high-end operating systems, so they can be independently deployed, upgraded and restarted. This again simplifies management and eliminates update risk.
InfoSight provides end-to-end visibility (even to your mobile phone app), predictive alerts and health notifications to reduce risk.
Performance
We were able to run our standard storage system benchmarks against the Primera system and were impressed with the results. It's on par with many of latest all flash systems. We can't wait to benchmark it more for customer specific workloads. The high performance levels observed were achieved without doing any kind of tuning.
The management of Primera system is much simpler; no more RAID or Pool configurations. Just create volumes and allocate them to the hosts. Easy!
Conclusion
We see many benefits to the architecture/design changes in the new Primera system. Those changes resulted in better manageability, serviceability, better uptime (100 percent availability guarantee) and quicker initial deployment. We are very impressed with the new Primera system.
We also think that allowing the option for the customer installation of Primera is a great improvement. Once the installation instructions are better documented, we see the value Primera can offer our customers along with the simplified installation, faster time to production and in turn, quicker time to market.
The customer self-installation gives options to the customers. The customers can still opt for the concierge treatment and let HPE or certified partners such as WWT do the installation.
We have a unique capability to integrate the whole solution from our integration facilities, including the integration of all the components including the third party equipment in a rack or multiple racks. See for yourself by scheduling time in our lab environment.