A node failure during product testing convinced the Yavapai-Prescott Indian Tribe it had found the best platform...
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
for its Microsoft Hyper-V Windows storage.
The Native American tribe in Northern Arizona started with a 2U demonstration model of all-flash Gridstore HyperConverged Appliances to support its Microsoft environment, which includes Exchange Server, SQL Server databases and SharePoint services. Gridstore systems optimize block storage for the Microsoft Hyper-V hypervisor.
Hyper-converged systems combine compute, storage and virtualization resources in a single chassis. The three-node Gridstore demonstration cluster included one node with compute and storage and two storage nodes. One month into the three-month demonstration project, a faulty backplane card resulted in one storage node getting knocked offline. The outage showed that Gridstore all-flash arrays could deliver fault tolerance and high availability, MIS network manager Chad Dixson said.
"You say 'node failure,' and people assume that means our demo project was a failure," Dixson said. "That wasn't the case at all. The Gridstore array was running in a degraded state, but we did not notice any drop in performance. The thing just kept running and running.
"When the one node stopped working, Gridstore wrote to the other two nodes, and it actually decreased the time to complete the transactions. Gridstore sent out hot-swappable replacement parts and quickly got it back to an optimal state."
Dixson since has purchased two additional three-node Gridstore all-flash clusters, giving it nine total nodes that collectively provide 30 TB of storage. Three nodes serve as Hyper-V hosts and six nodes run flash storage. Gridstore software presents storage as a virtualized pool to Hyper-V. Dixson said the scalable capacity allows room to grow beyond the 11 TB of effective storage presently in production.
"We started looking for 20 TB of usable capacity to fit a budget of $200,000. We got 30 TB from Gridstore, including three years of gold support, for $186, 000," Dixson said.
Gridstore HyperConverged Appliances replace consolidated storage
The Yavapai-Prescott Indian Tribe (YPIT) is one of three federally recognized North American tribes in Arizona. Roughly 180 tribal members and their families live on a 1,395-acre reservation near the City of Prescott. The reservation includes a 162-room resort, two casinos, a business park and shopping area.
Dixson's IT team manages the networks that support all activities of tribal government, with the exception of the casino operations. The Gridstore all-flash deployment marked YPIT's second experiment with the vendor's technology. The first test several years ago involved a Gridstore hybrid storage node that didn't suit its needs.
"To make use of the flash (in the hybrid cluster), we needed to have a lot of repeatable reads and writes," he said. "Ours were way too random, so we could never warm up the cache. The only alternative was to run it in volatile mode, but that meant none of the data was saved in flash."
He said the journey to Gridstore HyperConverged Appliances involved several detours. When first virtualizing its IT environment, YPIT used commodity disk storage and dedicated Hyper-V hosts on centralized NAS, with iSCSI connectors running across Gigabit Ethernet connections.
Hyper-V hypervisor upgrades in subsequent Windows Server versions enabled the tribal nation to migrate virtual machines (VMs) across disparate data stores. "That gave our VM farm the IOPS we needed. It was better than our centralized storage model, but we didn't have high availability," Dixson said.
Gridstore nodes spur SQL Server consolidation, VDI rollout
Dixson looked into EMC VCE, NetApp FlexPod and IBM VersaStack converged infrastructure products, ultimately concluding they were "interesting but complicated." Tests with software-defined storage products created fault-tolerant arrays, but "at a huge cost of storage."
Dixson said his Gridstore cluster provides scalable flash capacity that has kick-started dormant IT projects without requiring additional spending. He is planning to migrate Microsoft SQL Server workloads to Gridstore and use the HyperConverged Appliances for a future virtual desktop infrastructure project.
What's behind the rise of hyper-convergence?
Spelling out the differences between converged and hyper-converged
How a UK law firm used Gridstore appliances to upgrade its data center