Vector Database Storage Sizer

What the Vector Database Storage Sizer does

Vector search stores one embedding per item, so size grows with the number of vectors, their dimensions and the bytes per value. Graph indexes such as HNSW add links per vector. This calculator totals raw vectors, index overhead, metadata and replicas.

Formula

• Raw vectors = vectors × dimensions × bytes per value
• HNSW links ≈ vectors × M × 2 × 4 bytes
• Total = (raw + links + metadata) × copies

Worked example

Using the starting values in the calculator above:

How to read the result

Quantising from float32 to int8 cuts raw vector size by three quarters, usually with a small loss of recall. Many systems keep full precision on disk and quantised vectors in memory.

Reducing dimensions, where your embedding model supports it, shrinks both storage and query cost in proportion.

Limits to keep in mind

• Real systems add write-ahead logs, segment overhead and free space for compaction; leave headroom.

Related calculators

• IP Subnet Calculator: Calculate the IPv4 network address, broadcast address, usable host range, subnet mask and wildcard from an IP and prefix.
• Data Transfer Speed Calculator: Calculate how long a file or backup takes to transfer at a given connection speed, allowing for protocol overhead.
• IT Ticket Backlog Burn-down Calculator: Calculate how long it takes to clear a support ticket backlog and the closure rate or staff needed to hit a deadline.
• Penetration Testing Cost Estimator: Estimate penetration test effort and cost from applications, IP ranges, internal testing, reporting and retesting.

Browse every calculator in SaaS and Technology Calculators, or talk to us about web and SaaS development.