Get scalable, high-speed, cloud enterprise storage with Excelero NVMesh

As a software-only solution, the Excelero NVMesh storage platform is built to give customers maximum flexibility in designing storage infrastructures using bare metal, virtual instances, or containers.

The big data storage solution supports traditional network fabrics and protocols, enabling NVMe over Ethernet and InfiniBand, so you don’t have to invest in new networking technologies!

NVMesh is deployable on a wide variety of supported server platforms, so you can mix and match hardware from different vendors, with various capacities and from various generations.

This allows cloud-era application development and deployment by abstracting application requirements from the underlying hardware capabilities.

In public cloud environments, NVMesh supports instances, both virtualized and containerized, that feature NVMe drives. Public cloud instances with local NVMe drives have become widely available, allowing easy transition between on-premise and public cloud deployments of NVMesh.

Excelero NVMesh Nvme Storage supports the widest selection of network protocols and fabrics, including TCP/IP, InfiniBand, RoCE v2, RDMA and NVMe-oF.

A key component of Excelero’s NVMesh is the patented Remote Direct Drive Access (RDDA) functionality. Bypassing the CPU on the target nodes with NVMe drives, RDDA eliminates noisy neighbor effects on applications, enabling converged deployments, adding only 5 microseconds of latency overhead to the drive access latency.

The shift of data services from centralized controllers to complete client-side distribution enables linear scalability, provides deterministic performance for applications and enables you to maximize the utilization of your flash drives.

Excelero NVMesh is deployed as a virtual distributed non-volatile array and supports both converged and disaggregated architectures, giving you the architectural design freedom needed for cloud enterprise storage.

Excelero NVMesh’s architecture is designed for the modern data center from the outset, so it scales naturally. Targeting larger data centers with upwards of 100,000 converged nodes, the system uses three main architectural supports to avoid scaling issues:

1. Avoid any centralized capability for the data path

• • The system eliminates centralized metadata and lock management

2. Ensure that the client, target, and management functionalities are built as scale-out technologies

• • This is critical to ensure that the networking pattern is suitable for scale and that clients are completely independent with no cross-client communication ever required

• • Rare knowledge sharing occurs indirectly and anonymously via targets
  • Cross-target communication is clustered and with limited fan-out, reminiscent of the communication patterns of social networks. This communication is for data protection maintenance, ensuring data loss and corruption are avoided
  • The number of elements participating in data protection of any specific stored data element is limited to reduce system chatter and enable wide scaling

3. Intelligent clients communicate directly with targets

• • This reduces the number of network hops and the number of communication lines, making the pattern acceptable for a large-scale environment, where the number of connections are a small multiple of the domain size
  • This is in stark contrast to a system in which the number of connections does not scale linearly, for instance, where all targets communicate with each other or every client interacts with a significant percentage of target nodes

• Drives are perceived as resources that are pooled into a large storage area
• Logical volumes are then carved out of the storage area and presented to clients as block devices
• Volumes may span multiple physical drives and target hosts, but do not need to use entire drives, so a single drive can be allocated to multiple volumes

Volumes can be configured with any of the following redundancy levels:

• Concatenated – Data is laid out on single or multiple drives with no data redundancy. This volume type can be used for applications requiring temporary storage. Failures are typically isolated to a single device or host.
• Striped – Data is laid out across a set of drives and hosts with no data redundancy. This volume type can be used for applications requiring high-performance temporary storage.
• Mirrored – Data is protected by mirroring data across drive segments. To increase data availability, the drive segments are allocated from drives on different target hosts. These hosts should be connected to different power supplies, preferably in separate upgrade and availability zones, and any other zoning used for risk protection. The software’s management layer provides the agility to ensure such separation. Multi-way, active-active multipath networking is used for availability and performance.

• Striped and mirrored – DData is protected by mirroring data across drive segments and striping across these mirrors, then serviced from many drives and hosts, achieving high performance computing without sacrificing redundancy.
• Erasure coded – Data is stored on a set of drive segments, both for data and for parity, with an N+2 redundancy level. In case of a segment failure, either from a device or host, data is reconstructed on the fly from the remaining segments. This provides higher redundancy than is the case with both Mirrored, and Striped and Mirrored volumes while still providing high performance by spreading the workload across hosts and drives.

When deployed on public cloud instances with local NVMe drives, NVMesh pools these drives, preserving their performance characteristics, while providing data protection from both drive and instance failures, relieving concerns around data availability. Data can be spread across availability zones for an additional level of protection. Data compliance and security are customer-maintained and controlled by limiting access to storage and compute nodes using identity management systems.

Excelero NVMesh’s centralized management uses a stateless web serving framework (Node.js) and a back-end transactional scalable data store (MongoDB). This ensures that the management framework has the required scale-out and failover characteristics to support even the largest data centers that require IO intensive storage in the cloud. To facilitate maintenance, NVMesh features an intuitive GUI and a feature-rich performance analytics engine, which enables pinpointing anomalies quickly and at scale.

In addition, NVMesh enables users to analyze cluster-wide performance and utilization, and build a customized dashboard using a selection of data visualization widgets.

New modern workloads are engineered for containerized invocation while others are being redesigned. Excelero NVMesh provides a Kubernetes CSI plugin for simple generation and consumption of elastic storage.