Skip to main content

Current release / ZJX 1.25.1 NIGHTMARKET

ZJX 1.25.1 NIGHTMARKET

Two explicit packing objectives: compact searches for the smallest complete archive; fast trades packing time for cost inside strict archive-size and restore-cost guardrails. The fastest schedule we tried was rejected for making restores slower.

RAW STRUCTURED HISTORY342.4MB
VERIFIED ZJX ARCHIVE4.41MB
98.7% smaller
342,409,609 bytes in 4,412,968 bytes retained

What changed

Compression is now a customer choice.

NIGHTMARKET introduces objective-driven packing. Compact is the canonical no-argument default: an exhaustive size-first search that keeps the smallest complete archive in the eligible candidate universe. Fast is a fixed reduced schedule, promoted from the former experimental balanced schedule and qualified independently under the unchanged fast contract.

Current releaseZJX 1.25.1 NIGHTMARKET

Qualification result

Measured, bounded, exact.

~2.75x faster
Relative speed
~63.6% less wall time
Less wall time
0
Measured size penalty
48/48 exact restores
Exact restores

Controlled NIGHTMARKET qualification campaign. See methodology and evidence.

Positive-median wall savings in 4 / 4 primary corpus families across 48 / 48 pack starts.

Methodology

Frozen before the first start.

Four corpora, six matched pairs per corpus, two arms per pair with alternating order, a deterministic paired-ratio sign-test interval at alpha 0.05, no outlier deletion, no post-result tuning, and no pooling with earlier measurements. The campaign ran on a dedicated, quiesced laboratory host.

Limitations

Campaign ratios, not guarantees.

Performance ratios apply to the frozen laboratory host, four corpora, commands, repetitions, and thresholds of the qualifying campaign. Fast carries a published size budget: the campaign observed no penalty, but other inputs may trade some size for time within explicit bounds.

Customer savings language

Model the runtime you could avoid.

Illustrative runtime model

At the NIGHTMARKET qualification ratio, a workload requiring 1,000 minutes in compact mode would require approximately 364 minutes in qualified fast mode, avoiding approximately 636 minutes of wall time.

364 min
Modeled NIGHTMARKET runtime
318.1 h
Modeled hours avoided / month
$318.15
Illustrative cost avoided / month

Modeled from the NIGHTMARKET qualification wall-time ratio. Actual savings depend on workload, hardware, configuration, resource utilization, and infrastructure pricing.

Historical context

Why 1.25.1 and not 1.25.

ZJX 1.25 was published as a private, authenticated release and withdrawn before public distribution: its runtime reported the fast objective as production-qualification-pending, contradicting the release claim. ZJX 1.25.1 corrected that status string, added a regression assertion for the exact API surface, and underwent a fresh exact-byte qualification cycle with new binaries, a new 48-start campaign, and new release validation.

Get NIGHTMARKET

Bring one expensive archive lane.

ZJX is offered through evidence-led evaluation: Greyforge measures your workload against serious comparators and returns an evidence packet for technical review.