VAST Data Platform is the first unstructured data platform all-flash storage appliance that delivers the performance throughput of modern applications, exabyte scalability, archival-like economics, and NAS simplicity.
VAST’s architecture employs QLC flash drives, storage class memory, and NVMe-oF, on top of which the proprietary SIMILARITY technology is used to unify a storage hierarchy that previously consisted of multiple products in multiple stages. The result is a significantly lower cost of ownership and up to 90% more cost-effective than traditional flash technology systems.
Whether for file or object storage, deep learning, multimedia, HPC, life sciences, big data analysis, or backup, VAST Data Platform is used worldwide in a variety of fields.
Storage class memory is used to significantly reduce the number of write cycles. Enabled the use of QLC drives, which are less expensive than today’s mainstream MLC/TLC drives, and achieved a lower price.
Despite being a high-end storage device that uses QLC, which has a short write life, it offers a long-term warranty similar to that of tape drives.
It supports the older NFSv3/4, as well as NFS+RDMA[+GPUDirect], SMB, and S3, allowing access to the same files using any of these protocols.
Rebuilds are performed automatically as long as there is disk space available. Full recovery is possible by simply replacing the disk without stopping all or part of the system. Even in the event of an abnormal shutdown, the system is atomically processed via non-volatile storage memory to complete the boot-up in the shortest time possible without rebuilding at reboot time.
Our unique SIMIRALITY technology, which simultaneously compresses and deduplicates data, determines similarities across the entire storage space and can store more data than the physical capacity of the storage space. It also supports preliminary verification of the amount of compression that can be achieved for existing data.
All management features are available at no additional cost, including easy-to-use GUI management screens and the ability to retrieve statistics. Over 1000 snapshots with no overhead are created per directory, visualization of data flow at each layer, backup to the cloud, and per-directory quotas are also supported.
What is NVMe-oF?
NVMe-oF (NVMe over Fabrics) is a mechanism that extends the NVMe protocol, a new transfer protocol designed for semiconductor memory, to enable systems to access NVMe SSDs over fabric networks, such as Ethernet and Infiniband, and PCIe connections to achieve higher speeds.
What is GPUDirect Storage?
This technology enables applications running on top of the GPU to access data in storage without the need for CPU or OS intervention. This enables higher throughput and lower latency.
What is Storage Class Memory?
In the conventional CPU-memory-storage configuration, memory is fast but its capacity is small, and storage is large but slow. Storage class memory is non-volatile memory placed between memory and storage, and is characterized by larger capacity than memory and faster I/O than disk.
VAST Data Platform takes full advantage of these storage-class memory advantages. VAST Data Platform takes full advantage of these storage class memory advantages, such as writing to QLC with optimal block size and maintaining distributed metadata.
VAST Data Platform consists of a VAST protocol server, network switches, and a VAST NVMe enclosure. It is a stateless protocol server and storage that can immediately free up server resources, and its separate functionality provides scalability and redundancy through its connection to the server.
With NVMe-oF, VAST DASE allows remote access to SSDs and storage class memory while decoupling the CPU from the storage device without compromising performance. In addition to performance comparable to a shared-nothing approach, VAST DASE provides greater fault tolerance by eliminating the need to reconstruct data over the network in the event of a protocol server failure. VAST DASE can fail multiple drives and continue to operate as long as there is space available on the other drives, and recovery does not require rebooting even a portion of the system. All modules, including power supplies and fans, are redundant.
VAST Data Platform increases throughput with more protocol servers and capacity with more NVMe enclosures.
Shared Everything is a data sharing scheme that allows direct access to data from all nodes. VAST’s distributed shared-everything architecture solves this problem by allowing NVMe-oF, storage-class memory, SSDs, and the VAST OS to process data through an optimal write buffer. storage class memory, SSDs and the VAST OS work through an optimal write buffer to eliminate this problem, creating a scalable, fault-tolerant system.
Files can be accessed with either NFS or NFS+RDMA[+GPUDirect], SMB, or S3 protocols. It also supports NFS multipath, which uses multiple I/Os for faster transfer speeds.
While traditional storage systems cannot access the same file using different protocols or experience significant performance degradation, VAST DASE can access files in the same namespace using any of the protocols without performance degradation. IPv4 and IPv6 are available.
You can configure a cluster with an unlimited number of stateless servers. Stateless servers specialize in protocol translation and do not store data internally. Therefore, in the event of a failure, data does not need to be reconstructed over the network. Because each server is loosely coupled and operates independently, all servers can access the shared global namespace on their own. This avoids crosstalk and error coding synchronization issues between servers and allows for virtually unlimited scalability.
NVMe enclosures can scale independently of the number of protocol servers and can manage exabytes of data in a single multiprotocol namespace. The enclosure has two fabric modules operating in an active-active mode, so that if one of the fabric modules fails, the healthy fabric module will continue to maintain atomic write consistency.
It is also possible to encrypt all data using FIPS 140-2 validated libraries if encryption on storage is enabled.
VAST’s management services are available via GUI, CLI, and API, and the GUI dashboard is designed to provide intuitive access to basic information, including IOps, latency, data compression rates, capacity estimation, and more. The dashboard can be customized to display more than 1000 graphs and data, including IOps, latency, data compression ratio, and capacity estimation. The graphs can be used to identify high bandwidth users and bottlenecks, and can be used to reverse engineer performance issues.
Detailed access restrictions, quota settings, QoS settings, and triggers to change restrictions can be set, as well as operations such as data replication and protection. For example, quotas can be set on a per-directory basis as well. Information can be displayed in real time or in the past from logs, and the retention period can be freely set. Multi-factor authentication is supported to protect management access, and more than 1000 snapshots can be created for each individual directory. Snapshots are located in hidden directories, allowing administrators to deal with data deletion due to user error without hassle. Encryption supports NFS KRB5p, TLS, S3 HTTPS, and SMB 3 (TBA). All features are available at no additional cost.
The VAST file system does not simply overwrite data, but rather manages data using VAST’s technology, which is minimized to make planned additional writes.
New writes are added to the write stripe via an indirect layer. Specifically, write stripes are created via pointers stored in temporary nonvolatile storage class memory for the data to be written, and all QLC writes are handled atomically.
Thus, no data corruption will occur in the event of power loss or system crash.
VAST Data Platform offers up to 10 years of maintenance service.
It uses QLC drives, but as mentioned above, the number of writes is greatly reduced by generating large block write stripes via storage class memory, maximizing service life.
VAST Data Platform is also designed to mix and match different generations of hardware. If you need to increase capacity during operations, you can add the latest protocol servers or NVMe enclosures to your existing system for expansion.
VAST Data Platform identifies common and similar parts of the data being written and reduces the data at the byte level in the flow of preparing the data for storage in the QLC. VAST Data Platform identifies common and similar portions of the data being written and reduces them to the byte level in the flow of preparing the data to be stored in the QLC.
VAST’s data similarity technology reduces the amount of data used by several times the disk capacity by determining the similarity of data globally and at the optimal granularity. times the disk capacity.
Life Science Data
GPUDirect is a technology that accelerates data transfers between GPUs and storage, working between local and remote storage without CPU intervention to achieve higher throughput.
The GPU server’s RDMA acceleration is four times faster than a typical NAS, and VAST Data Platform accelerates learning by providing low-cost, high-performance storage.
All-flash NAS offer high performance but at a high cost, and VAST Data Platform delivers fast workflows at a price point comparable to archival storage.
VAST DASE requires no time to recover from a failure. Petabytes and exabytes of content can be served without service delays.
Available at a price point comparable to archival storage, VAST Data Platform can streamline backups and allow for quicker recovery from fast flash drives.
The simplicity of the NAS platform allows you to take advantage of the performance of high-end parallel file systems. By eliminating data I/O bottlenecks in computing operations, overall processing speeds can be increased.
We can customize it to your needs. Please feel free to contact us regarding system configuration. *required