Driving the highest performance at an efficient cost with Excelero NVMesh technology and NVIDIA BlueField DPUs
It’s the end of the month, and you are in that dead period, waiting all those hours for your critical monthly business reports that everyone in the company is anxiously awaiting to complete. Hopefully, everything is square with the data, because we all know the difficulties if that report fails or needs to be rerun. Everyone is anxious enough, without having to take another half a day to the reports in their hands. It just makes me think about all the money that goes into these dedicated infrastructure platforms for critical application performance—platforms that often leave us wanting better performance. Don’t you wish the cloud could provide the necessary on-demand, pay-as-you-go performance for these sporadic yet critical workloads? Just dial up the environment for your monthly and quarterly analysis runs and get your reports faster. No need to unnecessarily pay for that huge fixed asset all the time.
This is where Excelero technology comes to help. Excelero offers NVMesh (NVMe technology) that delivers better performance than those dedicated infrastructure environments and can do it on any machine – and when I say “any,” I’m talking about both cloud virtual machines and on-prem machines in your data center.
Performance is of critical importance in so many applications today. If I take a business-critical Oracle database environment as an example, you need machines with compute horsepower, storage horsepower, and fast networking to hit those big performance numbers that cut down your report run times. In the cloud or the data center, there is a ton of compute performance available: you just need to choose the machines that meet your goals. But then you need extremely fast networking to pump through the amount of data that a typical monthly report requires. (This is where NVIDIA’s BlueField DPU technology comes in – I’ll get to this in a moment.) And then finally, you need high storage performance to feed the hungry CPUs and GPUs—storage that can be shared easily across servers and applications. This is often where the true bottleneck lies – especially in the cloud.
So how does Excelero solve this? NVMesh technology is a low-latency, distributed storage software that is deployed across machines with very high-speed local drives (NVMe SSDs, to be exact), enabling high speed compute and high data throughput that far exceeds anything achievable with other storage alternatives – and at a significantly lower cost. We take the NVMe drives from across all machines and pool them into data volumes that your applications – running on those same machines – need. So you get local drive performance for your applications, distributed such that the performance scales as you add more machines to the environment. A side benefit of spreading the data across machines is that we eliminate the single point of failure and potential data loss if a drive or the machine with your data goes down, making the environment highly reliable and always available.
This is where network performance becomes critical and why Excelero is working with NVIDIA and their BlueField DPU infrastructure-on-a-chip, plus NVIDIA DOCA software platform technology. To get into some of the nuts and bolts, NVMesh technology speaks native InfiniBand and RDMA over high-speed network connections to maximize the performance of the interconnect between the machines. With BlueField-2 delivering the interconnect performance and offloading all the networking and NVMe-oF details, NVMesh software can completely bypass the CPU and maximize storage performance. So by aggregating the performance of multiple NVMe SSDs, plus offloading and accelerating data transfers on the network, NVMesh and BlueField work together to ensure data delivery from the storage does not bog down precious compute power that your applications are so hungry to use.
What you get is storage access latencies less than 100 microseconds and throughput numbers well over 100GBps. In the aforementioned Oracle example, this means millions of transactions per second in both the cloud and on-prem data center, so you can now look at new options to deliver your critical business environments that can save you time and – more importantly – money!