Last fall, NetApp published some eye-popping performance numbers for FlexGroup volumes. The test performed was a standard benchmark for software build workloads. These workload types characteristically include many files, high ingest rates, and a large amount of metadata operations.


Good news for NetApp: This is a FlexGroup volume’s sweet spot.


The numbers last year were so good, we decided to try it again, armed with a new ONTAP version, as well as the beastly A800 system with NVMe SSD drives. Last year’s tests were done with a four node FAS8200 (spinning SAS drives), so we wanted to see what an all-flash system could do. We were not disappointed.


There were two runs of the test, both using the new A800 system platform. One was with 4-nodes, the other with 8 nodes. These tests generated 1.27 billion and 2.42 billion files, respectively. This was designed to show how ONTAP clusters with FlexGroup volumes can handle billions of files with massive performance and scale it out even further just by adding additional nodes.


Not only did we beat our score from the previous year (expected, really, since it was a FAS8200 vs. A800), but we also achieved the #1 and #2 ranking, based on the sheer number of builds (2200 for four nodes, 4200 for eight nodes!) that the system was able to handle. Plus, our submission is nearly 3x more than the next closest competitor.*


SPEC SFS®2014_swbuild

* It’s important to note that who hasn’t submitted official results is nearly as noteworthy as who has.


Last year, NetApp achieved 520 SPEC SFS2014_SWBUILD SWBUILDs with an enterprise-leading 1.04ms overall response time (ORT). This year, the ORT was .732 for a 4-node cluster and .781 for an 8-node cluster. This was accomplished with all storage efficiencies enabled (ONTAP’s default setting). That’s something most storage vendors wouldn’t dare to try, as most platforms would see a significant performance drop. ONTAP sees very little performance impact when storage efficiencies are enabled.


We also didn’t modify any of the client-side attribute cache settings. Everything was left as the default values – this shows that we can achieve high performance without taking any shortcuts.

The Performance Results

The new results for a 4-node A800 cluster running ONTAP 9.5 achieved over 4 times the number of IOPS at a lower latency than the old results on the FAS8200 running ONTAP 9.2.


When the cluster scaled out to 8 nodes, the results got even better, as the performance for the benchmark scaled nearly 1:1, giving ~8x the number of IOPS than the 4 node FAS cluster gave us.


NetApp ONTAP with FlexGroup volumes also delivered the highest throughput of any of the submissions, at 27GB/s with the 8-node cluster and 14GB/s with the 4-node cluster. The next best number was 11GB/s.


NetApp has proven yet again that ONTAP with FlexGroup volumes can do more work at lower latencies and higher throughput for faster overall completion times. And we all know time is money, so expect return on investment over the life of the system as well.


These combined results show that with ONTAP and NetApp FlexGroup volumes, you don’t have to decide between IOPS, latency, throughput, or the number of builds a cluster can handle. You get it all. In addition to performance, you also get:

  • Cost benefits of storage efficiencies
  • FabricPool support (AFF only)
  • Nondisruptive scale-out and scale-up of file systems and clusters to grow with your applications
  • Enterprise-class Snapshot™ technology
  • Simple, streamlined deployment
  • Security and data protection

For more information, see:

Justin Parisi

Justin Parisi is an ONTAP veteran, with over 10 years of experience at NetApp. He is co-host of the NetApp Tech ONTAP Podcast, Technical Advisor with the NetApp A-Team and works as a Technical Marketing Engineer at NetApp, focusing on FlexGroup volumes and all things NAS-related.

Add comment