AegisDB

AegisDB is a fault-tolerant embedded database prototype built around LSM-tree principles, combining write-optimized ingestion with durable recovery and read-latency optimizations.

-40%

Read Latency

WAL

Durability

Embedded

Engine

Database Systems

What Wasn't Working

Continuous high-volume writes and mixed read workloads require a design that keeps writes durable and fast while avoiding unnecessary disk seeks during lookups.

The Solution

How I Fixed It

I implemented an LSM-style engine with WAL-first durability, in-memory MemTables, immutable SSTables, and Bloom-filter based file skipping to minimize redundant disk access, plus compaction for stale-version cleanup.

Stack

Technologies Used

C++11

LSM Tree

WAL

MemTable

SSTable

Bloom Filter

Compaction

Results

Key Outcomes

✓

Delivered fault-tolerant recovery via WAL replay

✓

Integrated MemTable-to-SSTable write pipeline with compaction

✓

Reduced query read latency by over 40% using Bloom filters

✓

Built a complete embedded KV foundation in strict C++11

Want something like this?

Let's build it. I ship fast and I ship clean.

View Code Hire Me