WADP is the world's leading intelligent data platform developed by ExponTech, aiming to reconstruct enterprise IT infrastructure with "data center" and build new infrastructure for AIs. Through the future-oriented unified architecture (Universal Storage), the product realizes the unified replacement and modernization of the traditional storage array, parallel file system and big data storage system, provides the enterprise with a unified data base covering the dual scenarios of "general computing" (traditional UniversalStorage) and "intelligent computing" (AI Infra), and helps enterprises seamlessly integrate core production data and AI capabilities.
Core competence
Full scene data fusion
A set of platforms, with full scene bearing: Supporting both core transaction data (high IOPS, ultra-low delay) and AI data (ultra-high bandwidth, AI semantics), breaking down traditional architecture barriers
Rich interfaces, seamless interworking: Support blocks, NVMe-oF, Vhost, etc., files, S3s, AI KV Cache interfaces and AI semantic interfaces, underlying data sharing, extreme space utilization and efficient data fusion and flow
+AI integration of production business: Realize the integration and sharing of enterprise production data and massive unstructured data, one-stop intelligent data management capability, support the storage and access requirement of full-life cycle data of AI data pipeline, and support the direct docking of production data with AI capability
Save and calculate synergy to save and calculate, unleash AI potential
AI training: Provides read and write bandwidth close to the physical network limit, accelerates massive data loading to GPUs, efficiently processes Checkpoint writes and reads, and greatly reduces GPU latency
AI inference: Based on high-speed network, build a super-large-capacity and globally shared KV Cache layer, realize efficient KV reuse across requests and nodes, significantly improve inference performance, and greatly save computational power and cost by "saving substitutes"
AI Agent/RAG/database/knowledge base: Provide high-performance storage sled optimized for database to meet the demand of vector database, graph database and relational database for extreme IOPS, low delay and high bandwidth, and database TPCC performance is better than local disk and high-end flash array
Superior performance and scalable architecture
Block storage performance exceeds high-end flash array; File storage single bucket can support hundreds of billions of files; KV Cache storage is based on random small IOs running full of physical network bandwidth
Infinitely scalable random IOPS and storage bandwidth that scales linearly according to cluster size while maintaining stable microsecond latency
Software-defined distributed parallel storage eliminates bottlenecks in traditional storage architectures
Better TCO
Software-defined architecture, independent of proprietary hardware
Multi-pool technology, full flash pool and mixed flash pool can be used in combination, data can flow online across pools
Through technologies such as NVMe over RoCE, efficient end-to-end NVMe transmission can be achieved without special hardware, and the cost of replacement and operation and maintenance of proprietary fiber storage networks can be reduced
Support the separation of storage and calculation, and reduce the hardware, rack space and power cost required for storage in combination with erasure codes
Enterprise reliability
Distributed architecture guarantees high reliability at system level and has stably supported the core production environment of large enterprises
Provides highly reliable technologies such as distributed dual-active, asynchronous replication, and provides multi-level system availability and data reliability for single hard drives, servers, cabinets, and data centers
Hardware Monitoring Based on Machine Learning, Subhealth Detection of Network and Hard Disk
With a unified, intelligent and unlimited expansion architecture, ExponTech WADP solves the data splitting problem of traditional IT and AI infrastructure, helps enterprises manage full-scene data at low cost and high efficiency, and realizes seamless integration of core production system and AI capability, which not only provides stable support for traditional business, but also lays a solid data foundation for enterprises to march towards the era of new quality productivity driven by AI.
