HDFS 1GB Benchmark — GLYPH v0.1

Dataset:

• HDFS.log truncated to 1GB
• source: Loghub / HDFS dataset

Query set:

• 100 unique blk_* IDs extracted from the 1GB corpus

Baseline:

time while read -r q; do
  grep -F "$q" bench_1gb/HDFS_1GB.log > /dev/null
done < bench_1gb/queries.txt

Result:

• grep -F repeated scan: 11.516 sec

GLYPH index:

• SA32u built
• BWT built
• FM-index built
• persistent backend: query_fm_server_v1

Index artifacts:

• corpus: 1.0GB
• SA32u: 4.0GB
• BWT: 1.0GB
• FM: 8.1GB

Persistent FM benchmark:

• queries: 100
• total_time_sec: 0.001673
• avg_ms_per_query: 0.016727

Interpretation:

GLYPH is not a replacement for grep on one-off ad-hoc scans.

GLYPH is designed for repeated exact queries over pre-indexed static corpora. In that mode, it avoids rescanning the corpus and uses a persistent FM-index backend.