Hello, my name is Axel, and I’ll be at CloudFest with our Partners Boston and Mellanox to teach you all about NVMe. Here are five good reasons to hang out with me at CloudFest:
1. I can tell you how to leverage NVMe to make your business smarter and grow your Cloud Services offerings
When our founders started Excelero in 2014, they looked into what challenges were keeping infrastructure folks awake at night. They learned that data center leaders were done with expensive proprietary hardware based storage and were looking for scale-out solutions that leverage state-of-the-art components in standard servers. This way they wanted to increase infrastructure efficiency and flexibility. At the time, standard SSDs (SAS, SATA, PCIe Flash) had become the standard in the datacenters and NVMe SSDs were just entering the market. Today, NVMe SSDs are seeing very strong adoption, but most NVMe SSDs are not used to full capacity due to the inherent characteristics of NVMe-based SSDs: NVMe provides blazing fast performance “locally”. NVMe provides unseen performance benefits for your applications, but they can’t be shared across nodes, which often results in wasted resources. Back in 2014 I did not know Excelero, but while working for Fusion-io – SanDisk – HGST – WDC, I was dealing those limitations all day. Yes, Fusion-io had a great All-Flash Arrays (AFA’s), but AFA’s have a scale-up design, exactly what customers with a long-term vision wanted to move away from. More importantly, AFA’s were designed around the legacy SCSI protocol, while the up and coming NVMe protocol was magnitudes stronger. As a matter of fact, some NVMe drives provided more IOPS than a full AFA.
Excelero has married NVMe with RDMA resulting in networked storage. Excelero’s NVMesh provides unlimited scale, multi-domain (hosts, racks, rows) failure protection, it ships with client-side integrated storage virtualization and active/active multipathing. On top of that, we added an out-of-band stateless Management service written in state-of-the-art language.
One more thing I think is particularly remarkable about Excelero is that the company waited for three years to come out of stealth. The reason for this was that the founders wanted to hit the market with a customer-proven product. At the time of the launch – now exactly a year ago, Excelero had production customers who had been using NVMesh for over 12 months. Since I joined the company, we have won several new customers across different verticals in Germany, the UK and the Netherlands.
2. I can teach you how to deploy NVMe at datacenter scale
25/50/100 Gbit networks have become mainstream in 2018. Software-defined Storage and Software-defined Data Center have become a reality. We designed our solution with design-partners that include Big Eight Tech Giants. To deploy scale-out storage, we leverage our customers’ ethernet fabric and make it their networked storage: the network is your storage. As NVMesh can be deployed in a architecture mode without impacting target CPU, there is no noisy neighbor effect. (New) storage can be deployed like (new) CPUs: just add NVMe devices to your nodes or add nodes to your NVMesh cluster on the fly to add storage and performance capacity to applications running your business. Consume networked storage without compromising on access time.
Cloud infrastructures require 100% uptime, but applications need the ability to move data as it gets cold. NVMesh allows remote storage to be shared across applications in a multi-write attach-mode through a low-latency network. No more headaches about keeping pace with many-to-many topologies. Our distributed architecture is designed for unlimited scale in all directions. NVMesh creates one big pool of high-performance block storage using distributed NVMe SSDs. Through the patented Remote Direct Drive Access (RDDA), applications can access NVMe SSD’s remotely. RDDA itself builds on top of industry-standard Remote Direct Memory Access (RDMA) networking to maintain the low latency of NVMe SSDs even when accessed over the network fabric. By pooling the NVMe SSDs, storage can be consumed as logical volumes by the applications. In a traditional hyper-converged model, storage sharing consumes some part of the local CPU cycles, meaning they are not available for the application. The faster the storage and the network, the more CPU is required to share the storage. RDDA avoids this by allowing the NVMesh clients to directly access the remote storage without interrupting the target node’s CPU. This means high performance—whether throughput or IOPS—is supported across the cluster without eating up all the CPU cycles.
3. Our partner Boston Limited (Booth G17) will be there with the Flash-IO Talyn
Boston Limited, partnered with Excelero to build the “Flash-IO Talyn” disaggregated appliance running NVMesh. Talyn was designed after the Micron SolidScale architecture and features Mellanox networking and Supermicro servers, providing the highest IOPS per U, but at the lowest $/IOPS. Join me to chat with Boston about best practices for NVMe deployment, infrastructure design and support. Georg Klauser, Managing Director at Boston Server & Storage Solutions, will be available to talk about solving one of the biggest challenges in computing today – shared access to NVMe flash storage volumes at the lowest latency and high-performance required by GPU-based applications, yet with minimum processor overhead. Using Excelero’s server SAN, along with other leading technologies such as RoCE v2 and 100Gbit/s EDR InfiniBand or 100GbE from Mellanox’s Connect-X 5 range, processor overhead and latency is kept low, whilst bandwidth is plentiful. Boston has partnered with Excelero to provide a scalable NVMe over Fabrics storage solution using Excelero’s award-winning NVMesh® server SAN software and adds another piece to our vision of the Software Defined Datacenter.
4. NVMesh will be featured at the Mellanox booth (Booth H23)
At Mellanox’ booth you can learn everything about the Mellanox ESF (Ethernet Storage Fabric) for CSP, MSP and Enterprises. Mellanox will demonstrate how to unlock your network for Flash storage performance. ESF is the fastest and most efficient Ethernet storage network available, using Mellanox Spectrum switches and ConnectX adapters. It delivers performance, efficiency, and storage optimizations unequaled by any other Ethernet storage fabric.
5. While you can hear & see Dave Navarro play live at the party, meeting the one and only Ax(e)l Rose(nberg) is way cooler.
If you miss me at the Boston or Mellanox booths, there are two more options to find me: come to our session at the Circus Celebration on Wednesday 11:05 where I will be presenting “Lessons learned from the Big 8: designing elastic storage architectures”. Alternatively, you can ping me on email@example.com to meet after hours at one of the events because… I’m a great storage networker (pun intended)!