Intro

Meshless - High-Performance Sidecarless Service Mesh

Speed up your cloud with the most performant service mesh.
99% less resource consumption than sidecar pattern of service mesh.
Completely eliminate the network latency caused by sidecar pattern of service mesh.
Even eliminate the network latency of direct communication.
Real-time optimal routing.

Problems of Sidecar Pattern of Service Mesh

High Resource Consumption

A single sidecar may not use a lot of resources, but the total resource usage of all sidecars is huge.
For example, even if you can reduce the memory usage of a sidecar to 50MB, the total memory usage of all sidecars on a node of 100 pods is 5GB.

Significant Increase in Network Latency


In the case of Service Mesh with Sidecar, all packets from or to the application must go through the sidecar, resulting in each inbound or outbound packet traversing the TCP/IP stack of pod for extra multiple times, resulting in significant increases in network latency ↗ .

Number of Times of Traversing TCP/IP Stack of Pod:

CaseDirectService Mesh with Sidecar
Pod-to-Pod on the Same Node26 (+200% of Direct)
Pod-to-Pod Cross-Node26 (+200% of Direct)

“Direct” means “without Service Mesh”.

Meshless Solves Problems of Sidecar Pattern of Service Mesh

99% Less Resource Consumption than Sidecar Pattern of Service Mesh

One Agent per Node
Instead of
One Sidecar per Pod

For example, on a node of 100 pods, this reduces the number of Sidecar/Agent from 100 to 1.

Completely Eliminate the Network Latency Caused by Sidecar Pattern of Service Mesh, Even Eliminate the Network Latency of Direct Communication


UDS(Unix Domain Socket)
Instead of
TCP/IP Stack

Number of Times of Traversing TCP/IP Stack of Pod:

CaseDirectService Mesh with SidecarMeshless
Pod-to-Pod on the Same Node26 (+200% of Direct)0 (-100% of Direct)
Pod-to-Pod Cross-Node26 (+200% of Direct)2 (Same as Direct)

Meshless wins in both cases, completely eliminating the network latency caused by the Sidecar Pattern of Service Mesh.
Besides, in the case of Pod-to-Pod on the Same Node, Meshless even eliminates the network latency of direct communication.

Real-Time Optimal Routing

When routing to a destination service, Meshless connects in parallel to all instances of that service, instantly uses the 1st (i.e. the fastest) successful connection, and aborts others.