Akash
LOAD BALANCER
⚡Ultra Fast | Built in Go
Design
We've Aura
Built for speed, reliability, and simplicity — explore what makes Akash powerful. Here are the list of features and benchmarks on Akash
Algorithms
1. Weighted Round Robin
Distributes traffic proportionally based on assigned server weights. Great for handling servers with different capacities.
2. Round Robin
Cycles requests sequentially across servers. Simple, effective, and fair when all servers have similar power.
3. IP Hashing
Routes requests based on client IP hash. Ensures consistent routing of a client to the same backend (sticky sessions).
4. Least Connections
Directs traffic to the server with the fewest active connections. Ideal for balancing uneven workloads.
Distributes traffic proportionally based on assigned server weights. Great for handling servers with different capacities.
Time for Benchmarks
Brace yourself...
What happens when Akash is pushed to the limits?
1,000,000
Requests
70,000
Req/sec
2.5ms
Median Latency
6.9ms
95th %ile
163ms
Max Latency
40 MB
Data Transferred
99.995%
Success Rate
Why Akash Stands Out
%
Success Rate
Concurrent Clients
That means Akash can handle massive traffic reliably, without dropping a single request.
Performance vs HAProxy
Benchmarking Akash against the industry standard L4 load balancer
Performance Metrics
Throughput
req/sMedian Latency
ms95th Percentile
ms99th Percentile
msMax Latency
msKey Insights
Akash matches HAProxy in throughput (within ~0.5%)
Latency distribution is nearly identical, with Akash slightly better at higher percentiles
For a Go-based project vs HAProxy's C engine, Akash proves competitive
Note* Benchmarks were conducted on an Apple MacBook Air (M1, 2020) with 16 GB RAM, 256 GB SSD, 7 GPU cores, 8 efficiency CPU cores, and a 16-core Neural Engine. Results may vary on different hardware.
Akash is Ready. Are You?
Meet The Developer
I'm an engineer obsessed with building things from first principles. Languages, databases — the closer to the metal, the better.
Notable Works
Akash — A Load Balancer built from scratch in Golang. Supports 4 Algorithms, Health Checks and Metrics
Prithvi — A Key-Value database built in Java. Supports HashSets, VecDeque with LRU Cache and TTL.
Ganges — A Sanskrit-inspired programming language. From lexer to VM, all handwritten.
Rooster — The simplest way to deploy your React app on S3.
Finalist at Smart India Hackathon 2023. Built under constraints. 5X hackathon winner.
"Don’t just use systems. Build them. Understand them.