The World's First relativistic Linearizable, Post-Quantum secure, Tree-SSD Hybrid HTAP

starOur Solutions

EinsteinDB is intended to be a flexible relational (not key-value, not document-oriented) store that makes it easy to describe, grow, and reuse your domain schema.

The power of Galois, the ergonomics of PowerGraph, and the flexibility of Snap. EinsteinDB and MilevaDB bridge the divide providing optimized data structures and computation models that do not require users to write low-level imperative code.

Stores data with nanosecond-precision timestamps.

EinsteinDB translates a GHD into a series of loops, aggregations, and set intersections using stochastic foraging joins.

View Details

Boasts 20.7 million writes per second and 45.8 million reads per second on a single Intel Xeon E5-2680v2 based cloud server with 60GB of RAM (EC2 c3.8xlarge).

EinsteinDB is the world's first main memory distributed computing platform that exploits RDMA to improve both latency and throughput by an order of magnitude relative to state of the art main memory systems that use TCP/IP. EinsteinDB and MilevaDB expose the memory of replicas in the cluster as a shared address space.

View Details

Delivers 4.93x compression, including the precomputed statistical records. Applications can use transactions to allocate, read, write, and free objects in the address space with location transparency.

WHTCORPS via EinsteinDB, MilevaDB, VioletaBFT, FIDel, and Noether provides automatic load-dependent scaling via AWS auto-scaling, which allows fully elastic provisioning of your storage and throughput needs.

View Details
icon Optimistic Locking

EinsteinDB employs optimistic locking which works best for read-only and low contention cases thrives with mixed workloads.

we have designed a Hybrid lock-free multi-tenant mechanism which supersedes two-phase commit without actually locking.

  • we are committed to building a self-driving database platform (SDDP) with FIDel: with capabilities of self-detection, self-decision, self-recovery and self-optimization.
  • Prolog File Based Database via SQLite interfaces: rather than directly interrogating the database, provided extensive support for translating combinations of arbitrary Prolog and table-associated predicates to optimised SQL queries
  • hybrid architecture that takes advantage of both shared nothing and shared storage
  • A schemaless database which allows any data, structured with individual fields and structures, to be stored in the database.
  • The GNU malloc is ineffecient for multi-threaded programs. jemalloc to the rescue!
  • integrated with RCU performs 3.9X to 28X better in micro-benchmarks.
  • main improvements can be seen in the write throughput (roughly 3×) compared to MonetDB, Citus' PostgresQL, AllegroGraph, SAP Hana, RocksDB, and PolarDB
  • Exploiting SIMD: The Battle With Skew
icon Services

Cloud Native and Serverless

EinsteinDB's data model is based on immutable causets (causal sets; partial relationships) stored over time, enabling a physical design that is fundamentally different from traditional RDBMSs. Instead of processing all requests in a single server component, EinsteinDB distributes causet queries, does indexing, and improves caching via RDMA to provide high availability, horizontal scaling, and elasticity with FIDel's Kubernetes scheduler. EinsteinDB also allows for dynamic assignment of compute resources via memristor-array based Merkle-Trees.

  • EinsteinDB follows a hybrid shared-nothing storage architecture. It consists of three layers: a SQL load balanced Proxy acting as a unified access portal, a multi-node database cluster, and a distributed shared file system FIDel.
  • EinsteinDB leverages multiple AWS storage options to satisfy its semantic and performance characteristics. Different AWS storage services provide varying latencies, costs, and semantic behaviors.
  • Data backup time on EinsteinDB has been reduced to mere seconds. With the help of the excellent sRDMA network offered by VioletaBFT (Multi-Raft/HoneyBadger/SpeculativeBFT) and the newest block storage technology, the backup time is unrelated to the size of underlying data.
  • All of the nodes in an instance, including read/write nodes and read-only nodes, are able to access the same copy of data on a storage node. However, the traditional cloud database model only allows each instance to get its own copy of data.
  • a complete management system based on Docker to handle instance creation, deletion, and account creation tasks passed down by the user.
  • When EinsteinDB receives a read/write request from MilevaDB, EinsteinDB uses a shared memory to send the data to MilevaDB. EinsteinDB is a background daemon process running on computing node hosts that receives all of the read/write block storage requests for the instances and tools on the host.
  • Furthermore, EinsteinDB uses technology similar to Copy On Write to support snapshot creation in seconds. This means that all of the data in a database can be quickly backed up regardless of how large the underlying data may be, allowing EinsteinDB to support hard drives as large as 100 TB.
  • We honor universal SQL adoption by providing ANSI SQL features alongside the full benefits of NoSQL paradigm.
featureEinsteinDB achieves the query performance of OLAP-centric systems such as SAP’s TREX and MonetDB and, in parallel on the same system, retains the high transaction throughput of OLTP-centric systems, such as Oracles’s TimesTen, SAP’s P*Time, or VoltDB’s H-Store.

By exploiting a fast network such as RDMA, a database can interact with the shared distributed storage layer the same way as with a single (shared) local disk.

On top of this shared storage, we can easily launch multiple compute nodes to create replicas of a single database, having the identical view on the same data. Therefore, requests can be distributed to different (read-only) nodes for parallel processing


shared address space.

FIDel is a kubernetes scheduler and relativistic load balancer that controls the process which assigns Pods to Nodes in distributed cluster replicas



If two synchronizing clients operate simultaneously, or begin recording data prior to their first synchronization, two divergent timelines exist.


Multiple timeline support

Callers can populate a watch set as part of a query, which can be used to detect when a modification has been made to the database which affects the query results. This lets callers easily watch for changes in the database in a very general way.


Relativistic Merkle Trees

allow wait‐free, linearly scalable lookups in the presence of concurrent inserts and deletes


Write‐side delay primitives

Writers, in order to synchronize with readers, defer the freeing of an object until there are no readers referring to the object


lock-free reads over RDMA

Efficient NIC-based Authentication and Encryption for Remote Direct Memory Access

teamTeam Members

Our Awesome Team

EinsteinDB is a smart technology provider under WHTCORPS and is built upon WHTCORPS Group's business expertise and technological accumulations in areas such as artificial intelligence, big data, cloud computing and the internet of things. It has established a technology ecosystem that delivers unmatched customer value through comprehensive services spanning from foundational platform building to business consultation and planning, business platform construction and operations and maintenance, and is driven by industry leading products that enable smart and digital enterprises and governments through solutions across a wide variety of scenarios.


Karl Whitford

CEO & Founder(Amazon, Microsoft)

Josh Leder

CTO, Co-Founder(Senior Eng at Netflix)r

Joseph Pollard

OSS provider; IC

Ligeia Mare

project Recent Projects

The EinsteinDB Enterprise edition EinsteinMAX (Coming Soon; 2021)

A Relativistic Distributed SQL Database That Scales



a fault-tolerant globally-distributed OLTP and OLAP database built


MilevaDB is a hybrid database that combines high availability, the scalability of NoSQL systems like Bigtable, RocksDB, and the consistency and usability of traditional SQL databases: PostgresQL, MySQL.


FIDel is built on top of EinsteinDB and MilevaDB. Both systems were developed at the same time and in close collaboration. FIDel handles lower-level storage issues like persistence, caching, replication, fault tolerance, data sharding and movement, location lookups, and transactions.


PUTs protocol buffers in the schema, removes this impedance mismatch and gives users a universal data structure they can use both in the database and in application code.


The use of hash partitioning allows us to implement an efficient distributed hash join operator and a distributed ag- gregation operator.


a resource-management model for ML-style programming languages, designed to be compatible with the OCaml philosophy and runtime model

Pricing Plans

The Production Topology provides a full featured, horizontally scalable, and highly available EinsteinDB, FIDel, MilevaDB, and BerolinaSQL Cloud system. Intended for production deployments, this topology provides a load balancer, auto scaling clusters with local SSD caching, and the ability to launch multiple query groups for different applications and environments.


What's the word?

blogHow EinsteinDB powers Predictive Analytics by hacking Stochastic Context Free Grammars

Latest Valuable Insights

EinsteinDB provides two mechanisms to improve performance where required: lock-free reads over RDMA, and support for collocating objects and function shipping to enable the use of efficient single machine transactions. EinsteinDB uses RDMA both to directly access data in the shared address space and for fast messaging and is carefully tuned for the best RDMA performance. We used EinsteinDB to build a key-value store and a graph store similar to Facebook’s


EinsteinDB: A Semi-relational knowledge base with relativistic causal consistent memristor array based Key-SSD sRDMA memory.

EinsteinDB and MilevaDB, along with FIDel (Kubernetes Scheduler) and BerolinaSQL(yatp/allegro/sql) create a succinct yet powerful interface library to the SQLite database system with LMDB/PostgresQL persistence. The single binary, server-less approach of SQLite along with the natural integration of relational data within Prolog, render EinsteinDB a useful addition to the existing database libraries in modern open-source engines.

Get Started