For the P10400, we didn’t license many of the features that would have made this product really shine. We bought it for its advertised scalability, active-active controller design, and uptime (non-disruptive upgrades, port virtualization, etc.). We have not upgraded its capacity, controller count, or connection count since we’ve bought it so the scalability piece really didn’t make much difference for us. The uptime piece, however, has been 100% since initial start-up which is great. Although we cannot service the unit ourselves, cage level availability and the support team that takes care of the InForm OS updates and failed hardware have made it a great array for simply servicing I/O with nothing fancy added in.
The 7400 array is where things really start to shine for us on products and features. We initially bought this array as a production replacement of a set of EVA’s for file services and VMWare. As such it was important to us to have a reliable array that would balance itself, and was easier to upgrade than the P10400. It needed to support new OS’s fairly quickly and also have non-disruptive upgrades. We put that to the test when we added our test environment to the array. In that upgrade, only six months after initial purchase, it was very important that we could keep our test I/O from disrupting the production environment, so priority optimization became an important feature. We added two controllers, and doubled the size of the array to accommodate our test environment, and did so without downtime. The priority optimization is a real-time QoS component of the array that can be changed without disruption and takes effect in seconds. The six fibre ports per controller allow me to segregate I/O streams from VMWare and Windows to keep from having queue length and buffer issues or to simply balance out I/O load.
Improvements to My Organization:
3PAR, in general, has helped us speed up deployments and better track our I/O consumption. We are able to speed up deployments because we are no longer having to install a vendor specific DSM/MPIO on top of the Microsoft OS or Linux ones as we did with the EVA line. System reporter allows us a more in depth look at our I/O utilization patterns in an easy to gather and archive way.
Room for Improvement:
Licensing is still a pain point. It has gotten better, as any spindle count based licensing is capped at a certain number of spindles depending on the model of the array, but it is still expensive as you must buy every feature. Many of the arrays available from competitors come with just one or two licenses to buy which cover the entire array no matter the spindle count. The arrays are also not very user serviceable, which is something I miss about the EVA. A field tech must come out to replace a drive which is something I could do myself with the EVA.
Use of Solution:
I've used the P10400 for three years, and I would rate it 7/10. Alongside this, I've been using the 7400 for one and half years, and say it's 8/10.
Also, we are not currently using 3PAR flash storage but are planning an implementation. We are currently looking into flash to speed up our MS SQL for the logs and TempDB LUN’s. We are also looking into it as a possible future deployment for an as yet not deployed VDI implementation.
We didn’t encounter any issues with deployment. The pre-deployment guides were comprehensive and prepared us well for actual deployment of the product.
We have only expanded the 7400, doubling its size in disk and controller count, and did it without any issues.
High. We have “proactive” support, so many times they are calling me about issues before or right as I see the alerts. Our account support manager reviews our environment twice a year to discuss any issues we have had and any recommendations from HP on how the systems are performing.
High. Updates to the InForm OS and firmware have gone off without incident. When something was wrong with the HBA ports resetting, they brought in additional resources from other areas to resolve it. The issue turned out to be a problem with the fibre channel switches fill word instead of the array.
We were using HP EVA and direct attached storage. Our EVA’s were getting old and needed to be replaced within a year or two. The push to replace them sooner came after our main production array crashed due to LUN ownership issues between the controllers. A lot of I/O was coming into the non-owning controller and so it would switch ownership. Since ownership was set to fail-over/fail-back it would just keep hoping back and forth. It was a problem with VMWare losing its MPIO configuration but that caused major issues with the array. True Active-Active controller design really became an important criteria after this incident.
Both. The v400 was a complex setup where the 7400 was much more straightforward. The v400 had many requirements for how cables were run, specific hole sizes and cuts in the raised floor tiles, and the installation guide that the HP Field Technicians used was either out of date or incorrect in several places causing installation to take longer than it should have. In contrast, the 7400 had none of the requirements, as it was mounted in a more traditional style rack and the installation process was more mature. The service processor could now be deployed as a VMWare appliance and set up by the customer ahead of time.
We used a vendor team with in-house support where needed. The vendor team was much more knowledgeable on the 7400 installation than the v400. We had bought the v400 not long after HP acquired 3PAR so there were some issues there. However, the local installation team had a “quick” line into 3PAR support for installation so we weren’t waiting around for support to answer questions.
It was never calculated for these arrays outside of cost to buy new against the cost of continuing support of the existing arrays. It has been several years since we upgraded the class of array and type of support.
Cost and Licensing Advice:
The original setup was about $700,000 for the v400 and $350,000 for the 7400. The day to day cost is negligible in terms of support as they are really stable arrays.
Other Solutions Considered:
We evaluated the Hitachi VSP and Netapp FAS 6800.
Do your homework and follow the deployment guides, it will save time and headaches. If you do not license at least Direct Optimization the array cannot re-level the data so the array will eventually not preform to peak efficiency as you add and remove data.
Disclosure: I am a real user, and this review is based on my own experience and opinions.
Apr 27 2015