Logo Index64

Index 64 - The fastest in-memory key-value store.

Index64 is the world fastest key-value store, outpacing its competitors by a factor going from 2 to 10.

Stable Release



Designed for performance

Index64 is an ordered, concurrent, in-memory key-value store primarily designed for High Performance Computing. It indexes key-value pairs on distributed architectures using lock free algorithms for accessing data, while ordering the keys.

Our benchmarks show that Index64 is the fastest in memory KV-store currently available on the market. When using the potential of multicore processing, Index64 is between 2 to 10 times faster than Redis, depending on the data structure (ordered keys or not) and the environment (OS, VM), for both readings and writings. Its outstanding performance under Linux VM makes it a perfect choice for a Data Center.

Unlike Redis or Memcached, Index64 supports concurrent accesses to a large amount of data within a single node. Keys are ordered a priori making group requests possible to execute in a single run at an amazing speed.

Key Features

Data access concurrency

Data stored within Index64 are never locked, so that reads and writes can be performed concurrently, taking advantage of multicore CPUs. Complex data can be read from several threads at the same time. Operations are not serialized. Latency is always reduced to its minimum by construction. Data operations are fluidized, no matter the data complexity and the context.

Group requests

Unlike any other existing key-value store, which all use hash-tables, Index64 uses an ordered data structure to store the keys. Data is ranked a priori, at the time of writing. This changes everything in allowing group requests such as COUNT or ORDERED SCAN, without having to go through the whole dataset, even when distributed, to perform the operation. This allows more complex requests over voluminous datasets while keeping the level of performance at its peak.

To perform a few very specific group operations, Redis propose a complementary structure called “Sorted Set“. Our benchmarks against “Sorted Set” show that index64 is 10 times faster.


We use an industrial development approach that allows us to deliver a fully engineered stable product. Our approach includes a set of tools such as a code generator, a specific debug tool and full coverage testing programs. This engineering approach is critical for high demanding IT architectures.

And also ...

Memory allocation using an anti- fragmentation mechanism, with no garbage collector for an efficient memory management.

Documentation (PDF)


  • Color Index64 Index64
  • Color Redis Redis
Download full benchmarks (PDF)

Read Operations

In a single thread index64 uses a simplified data structure for better performance

Write Operations

Any operation can be performed concurrently, lock-free and thread safe.

* Operations are performed on Intel Xeon 2690v2 3.0 GHz - DDR3 1866 MHz - KVM Redhat - Dataset of 10,000,000 strings

Our vision for Index64



* required

About us

Index64 has been engineered over a period of 4 years by two experienced developers, who anticipated the need for in-memory data access performance, and complex data requests. From the data structures to the implementation, going through the algorithm, all has been thoroughly studied and tested, with deep attention to details, to get the best out of the current hardware possibilities. We are not talking about academic research nor proof of concept, nor are we talking about an open source project. Index64 is a private engineering project which gave birth to a breakthrough technology. It was conducted with the aim to confer a decisive competitive advantage to its future owner.

Jean-Christian Llobet worked as an IT performance expert for 20+ years. He helped IT architects of large companies optimizing the overall performance of their infrastructure. As a true expert, Jean-Christian owns a top-down vision of his art. Among his key achievements, JC has developed many outstanding software and hacks in various environment. Jean-Christian is fond of processor manufacturing, compilation techniques, physics, science fiction and politics. He holds an engineering degree from Ecole des Mines - France and a MBA from the IAE - Paris institution.

Harold Chatenet followed a multi-disciplinary career path. He went through international sales in the software industry, before joining a management consulting firm where he led several strategic company reorganizations. Since 2009, Harold has developed several multi-million downloads iOS apps.
Harold is fond of marketing, finance and creativity.
He holds a masters degree in innovation management from Paris IX Dauphine and a MBA from HEC-Paris.