Distributed pricing with loss-less curves

Quantlab 3.1.4155 introduces ultra-compact curve distribution

We are excited to announce a major efficiency upgrade in Quantlab 3.1.4155:
new built-in functions v_compress and v_decompress for distributing discount and forward curves.

What it means

• Any vector(number) can now be compressed into a compact binary blob.
• Blobs can be sent as raw binary, or base64-encoded for HTTP/JSON transport.
• Decompression is lossless, restoring the exact original vector.

The algorithm — DPCM²-RLE-Varint

Behind the scenes, Quantlab 3.1.4155 introduces a new codec we call DPCM²-RLE-Varint:

• DPCM²: second-order delta coding (delta-of-delta) exploits smoothness in financial curves.
• RLE: run-length encoding packs repeated second differences efficiently.
• Varint: ZigZag + LEB128 integer coding gives compact variable-length integers.

This combination is optimized for discount and forward curves, which typically change smoothly over maturities. The result is compression ratios far beyond general-purpose string or JSON encoding.

Real-world efficiency

In production testing:

• Naïve string encoding of a 365-point discount function: ~370 000 bytes
• Compressed blob with DPCM²-RLE-Varint: ~15 000 bytes
• That’s a 25× size reduction on average.

Network savings example

Suppose you distribute 50 different curves, each ticking every other second:

• Old method:
  50 × 370 000 bytes ÷ 2 s ≈ 9.25 MB/s (≈ 74 Mbps)
• With compression:
  50 × 15 000 bytes ÷ 2 s ≈ 0.375 MB/s (≈ 3 Mbps)

More than a 20× reduction in network load, without any loss of precision.

Where you can use it

• Quantlab clients directly (QLang, Excel add-in, Python integration).
• HTTP GET/POST endpoints — curves can now be served or retrieved in compact form.
• Internal distribution feeds where bandwidth and latency matter.

Quantlab 3.1.4155 makes real-time distribution of curves practical, efficient, and enterprise-ready.

Upgrade today and start feeding discount functions and forward functions across your network with minimal overhead.

‍