Advanced scalable data infrastructure for organizations of all sizes
 Storage

GraphStream offers a number of storage options, based on software building blocks that manage scalable shared storage across clusters of servers.


WekaIO Matrix™ is an advanced infrastructure-software system for scalable data storage.

Relative to competing systems, Matrix delivers a radically superior combination of data-access performance, scalability, data integrity, data durability, availability, operational simplicity, and cost-efficiency.

Unlike competing systems with a design center around rotating-media storage devices and other decades-old technologies, Matrix is a clean-sheet design that was developed from the ground up to deliver to applications the full performance of current-generation datacenter technologies.

Technologies that Matrix takes full advantage of, include solid-state storage devices; NVMe interfaces for storage devices; and the highest-performing industry-standard datacenter networks, which currently support up to 400+400 Gbps full-duplex peak throughput per link.

The superior data-access performance delivered by Matrix, has been demonstrated via record-setting results on multiple industry-standard benchmarks, including IO500 and SPEC SFS 2014 software builds.

Matrix can support a wide range of application workloads. Examples of performance-sensitive and I/O-intensive application domains where the superiority of Matrix is especially prominent, include machine learning; life sciences; advanced analytics; electronic design automation; and advanced HPC simulation.

The Matrix software system runs on a cluster of industry-standard server nodes containing solid-state storage devices.

Matrix aggregates solid-state block-storage capacity across the cluster to create highly performant, sharable, scalable, and resilient file-storage volumes that fully support standard POSIX filesystem semantics.

Matrix sharable file-storage volumes can be accessed concurrently by multiple application clients. Each of these clients can run on server nodes within the Matrix cluster, or elsewhere.

Matrix cluster nodes, and any additional nodes running application clients, are interconnected via industry-standard Ethernet or InfiniBand networking.

Peak full-duplex throughput per network link can be configured according to access-performance requirements and cost constraints, from 10+10 Gbps up to the highest commercially available link speeds, currently 400+400 Gbps for Ethernet and 200+200 Gbps for InfiniBand.

Each client can use one or more of the multiple standard filesystem-access interfaces that Matrix supports. These include Linux POSIX native filesystem; NFS; SMB; and S3.

Unlike competing systems, Matrix supports parallel access to both data and metadata, which are both fully distributed across the entire storage cluster.

The Matrix POSIX native filesystem interface provides clients with radically superior access performance for both data and metadata operations, by bypassing the Linux kernel and communicating concurrently in parallel with multiple Matrix cluster nodes via standard network links.

Unlike competing scalable shared file-storage systems, Matrix can deliver filesystem data and metadata access performance that is significantly superior to that of a standard filesystem that is striped across multiple NVMe solid-state storage devices accessed directly within a single server node.

After an initial Matrix cluster deployment, access performance and storage capacity can both be scaled incrementally and non-disruptively by installing additional cluster nodes, and additional solid-state storage devices within cluster nodes.

Matrix includes a tiering capability that supports a single filesystem namespace across data with the most stringent access performance requirements, which is stored in solid-state devices; and "cooler" data for which lower access performance is acceptable, which is maintained in an object storage system that is external to Matrix and incorporates storage devices with lower lifecycle cost per bit.

Matrix incorporates an advanced cluster-wide data-redundancy system that is conceptually similar to erasure coding. This system can maintain full data durability and availability when one or more cluster nodes fail.

Matrix also incorporates an advanced end-to-end data-integrity system that checks all data at the point of delivery to a client, and thereby protects clients from data corruption.

Each Matrix cluster node is a physical or virtual server that is based on the industry-standard x86-64 instruction-set architecture.

Matrix supports deployments that can span one or more modalities, including data infrastructure that is operated by service providers such as Amazon Web Services; and infrastructure that is operated more directly, such as server clusters installed in colocation facilities or "on-premises" in datacenters managed by an end-user organization.

Matrix also supports a variety of cluster-node configurations. Pre-validated node reference configurations are available for Amazon Web Services virtual-machine server instances, and for physical server systems from leading suppliers including HPE, Dell, and Supermicro.