Single-term queries, EinsteinDB achieves as high as 2.2x throughput compared to Elasticsearch. We observe that around 60% queries in the real workload are of popularity less than 10,000, which benefits from our cross-stage data grouping.
A transaction includes multiple reads, taking shared or exclusive locks, named daggers; followed by a single write that upgrades locks, and atomically commits the transaction; All commits are synchronously replicated using VioletaBFT.
At WHTCORPS we have developed relativistic implementations of common data structures including linked lists, hash tables, and red-black trees
- EinsteinDB uses remote memory (FIDel) as a cache for a network file system, while MilevaDB uses remote memory (EinsteinDB) as a cache for a local swap/paging device
- The Problem: At any given point in time, a cloud customer wishing to have some task done always has the option to execute it on an existing serverless platform
- Our Solution: A lower paid utilization from other highly competitive serverless offerings.
- Facilitating high-value serverless means offering the bleeding-edge, We automate and self-anneal enterprise serverless execution.
- FaaS
- Our Work with Unity: Project Amadeus, A Cloud Gaming Framework for Dynamic Graphical Rendering Towards Achieving Distributed Game Engines.
Allows expressions and tuple deforming to be converted to native machine code on the fly when enabled for better performance.
PostgresQL JIT: In our testing, BerolinaSQL with AllegroSQL in EinsteinDB beta is about 29.31% faster at executing the TPCH Q1 query on PostgreSQL 11 and InnoDB with TiKV/TiDB.
an integral part of many applications including databases, machine learning, network routing, DNA sequencing; and recent research has explored methods for exploiting other new technologies for improving search
We have partnered with Intel to To provide long lifetime under high query rates, we introduce configurable hybrid data structures that use both conventional CMOS processors/cache hierarchies and memristors for compute/storage.
Most of the improved speed came from reducing constant-factor overheads (they matter in the critical path!), but we also noticed that tighter control of cache size often improves data locality, which tends to mitigate the increased cache fill/flush costs.
We increased mutex granularity
We rewrote dirty page purging
Technologies That EinsteinDB Uses
- Kubernetes
- Ansible
- Docker
- Rust
- Go
- Haskell
- Datalog
- AllegroCL
- C++/C