Data center load shifting in distributed information systems: screening logs, study coding and evidence matrix

This record holds the data behind the review article "Data Center Load Shifting in Distributed Information Systems: A Structured Literature Synthesis and Research Agenda" by S. M. A. Aljaf, Y. A. Kravchenko and M. S. Ali. It contains the coding of the 162 included studies, one row per study with a DOI on every row; the full-text decision for each of the 283 records assessed, with the reason for each of the 121 exclusions; the seven logs written by the screening scripts; the five Boolean search specifications, with the query actually sent to the Crossref REST API and the count each returned; the screening flow as a table and as Figure 1 of the article; and a separate 100-record evidence matrix, which the article summarises apart from the screened corpus. The screening flow is closed: 283 full texts assessed, 121 excluded, 162 included, none undecided. The two identification routes are treated as separate sampling frames of 462 and 441 eligible records, and a stratified sample was drawn from each. 657 eligible records were never assessed, so the shares computed over the 162 are sample estimates. The coding, the evidence matrix, the search specifications and the flow were first deposited on 14 September 2026 as version 4 of an earlier archive, 10.5281/zenodo.22753882. This record differs from that version in five respects. The flow is closed: version 4 reported 116 exclusions with five records undecided, and all five were later excluded. Three DOI cells in the evidence matrix are corrected. The screening logs and the full-text decisions are added. The per-string counter that version 4's search specifications printed as duplicates of earlier strings is named for what it counts, and the file now accounts for all 1,500 records examined. The flow diagram is Figure 1 of the article; the diagram in version 4 placed the title-screen counts under the distinct works left after folding duplicate versions, so its boxes did not add up. The 162 included studies and their coding are unchanged. Version 4 also carries a correction notice withdrawing an 85-study coding file deposited in May 2026. That file is superseded by the corpus here and must not be used. README.md inside the archive lists every file and names the log behind each count.