Open Rainbow Ltd

Most Performance Problems Don’t Require New Hardware just needs Re-Think! Workloads evolve. Infrastructure models evolve. Usage patterns shift over time. When the environment changes but the design stays the same, performance issues appear—long before the hardware is actually “old.” A recent optimisation project delivered: • 4× more storage performance • 30% lower latency • Reduced network load • Lower power & cooling (OPEX) • Zero new hardware The bottleneck wasn’t the array, the server, the network ...... It was the architecture. In this case, two small RAID-5 pools were rebuilt into a wide ADAPT layout, compute alignment was corrected, and network paths were tuned. Metrics confirmed the workload was never the problem—only the outdated layout. Result: mid-range hardware delivering far beyond expectations, and a six-figure refresh completely avoided. Infrastructure rarely slows down because it’s aged. It slows down because it no longer matches the workload. Before planning a refresh, optimise the estate. The gains are often far larger than the spend. #ITInfrastructure #StoragePerformance #InfrastructureOptimization #DataCenter #EnterpriseStorage #PerformanceEngineering #OPEXReduction #ITConsulting #HybridCloud #Virtualization #VMware #AIOps #TechLeadership #OpenRainbow

Everyone talks about GPUs. Few talk about the infrastructure surrounding them. This article shows why reassessing storage, networking and data pipelines is essential for any serious AI strategy. #AIInfrastructure #AIPerformance #AIFactory #DataPipeline #RDMANetworking  #NVMe #StoragePerformance #PerformanceEngineering #DeepLearningInfrastructure  #GPU #AIOps #SystemsArchitecture #LowLatency #Throughput

A deep-dive into why NVMe-oF/RDMA performance on VMware often hits a ceiling long before the array does. Our latest deployment revealed a clear pattern: when the storage path relies on the VMware HPP plugin, IOPS plateau around ~170K — regardless of backend capability. But bypassing HPP using SR-IOV instantly unlocks 850K on the same Pure FlashArray X90R5 hardware (vertical performance). This article breaks down: - Why this behaviour happens on VMware - What HPP adds to the path - Why NVMe-oF behaves differently from iSCSI/FC - What changes when switching to SR-IOV - Real test data (Vdbench / RHEL10 / Mellanox ConnectX-6 / 100GbE) - When VMware is the right choice — and when it becomes the bottleneck If you’re evaluating NVMe-oF for databases, low-latency workloads, or next-gen infrastructure, these findings matter. Understanding the storage path matters more than the brand of array. #NVMe #NVMeoF #RDMA #VMware #StoragePerformance #SRIOV #Datacenter #PureStorage #RoCE2 #Infrastructure #DataStorage #HighPerformanceComputing #vBench

Slow is the new outage. See how a switch swap cut storage latency by 38% — and why it matters from database to user experience.