Corporate IT Infrastructure and Service-Level Management (Rolik, 2012–2013)
A user-supplied compilation of six papers on corporate IT-infrastructure management, service-resource models, feedback and disturbance compensation, and the integration of service-management processes with operational control. Its enduring contribution is the service-to-resource reasoning model; its product surveys, forecasts, protocol emphasis, and centralized automation architecture are historical.
This archive preserves a 110-page compilation supplied in the PYEIIT/ folder. It contains work by Anatolii I. Rolik published in KPI journals in 2012–2013, with journal covers and publication matter between some papers. The papers ask a durable question: how can measurements and changes at network, compute, storage, application, and service layers be related so operators act on user-visible quality rather than isolated device counters?
The source is admissible research literature, but it is not a present-day protocol guide. Its service-resource and feedback models remain useful as ways to organize reasoning. Current IETF, OpenConfig, OpenTelemetry, and SRE sources determine protocol semantics, telemetry architecture, configuration safety, service indicators, and operational practice in the GTK entries.
Complete folder audit #
The supplied folder contains exactly three PDFs and 203 pages:
| File | Pages | Classification | Treatment |
|---|---|---|---|
merged.pdf |
110 | Compilation of six research papers and journal matter | Read in full, visually audited, archived, registered, and mapped below |
ПУЕ_ІТІ_II_01.pdf |
52 | Instructor-authored lecture deck on IT-infrastructure management | Read in full and visually audited; used for topic discovery, but not archived because it substantially duplicates the papers and contains dated protocol/product framing |
ПУЕ_ІТІ_II_02.pdf |
41 | Instructor-authored lecture deck on IT-service-management frameworks and processes | Read in full and visually audited; not archived because its ITIL v3/ITMS2 survey is mainly historical and contributes no unique load-bearing mechanism beyond the papers |
The exclusions are editorial, not access failures. No student work or laboratory manual is present in the folder.
Paper-by-paper concept map #
The PDF page numbers below refer to the archived compilation, not the printed journal pagination.
| Compilation pages | Paper or material | Durable contribution | GTK destination |
|---|---|---|---|
| 1–29 | Trends and Prospects in Information-Technology Management | Separates service, application, compute/storage, and network views; surveys monitoring and management functions | network telemetry and service dependency maps |
| 30–56 | A Concept for Corporate IT-Infrastructure Management | Connects monitoring, analysis, control, optimization, and planning around a logical-object hierarchy | model-driven network management and service dependency maps |
| 57–58, 70–75, 84–85 | Journal covers, publication records, and author summaries | Provenance only | Library record |
| 59–69 | A Decomposition-and-Compensation Approach to Corporate IT Service-Level Management | Decomposes a service target across managed layers and distinguishes corrective feedback from compensation for measured disturbances | SLIs, SLOs, and error budgets |
| 76–83 | Coordinator-Based Service-Level Management for Corporate IT Infrastructure | Coordinates several local control actions against a service-level objective | model-driven network management |
| 86–96 | Operational IT Service-Level Management Based on Service-Resource Models | Represents a service as dependencies on application, compute, and network resources; measures service outcomes and chooses a layer-local response | service dependency maps |
| 97–110 | An Integrated Approach to IT Service-Level Management in Corporate IT Infrastructure | Integrates service-management decisions with operational measurement and control | SLIs, SLOs, and error budgets and model-driven network management |
Mechanisms worth preserving #
A service outcome is supported by a resource graph #
The strongest idea in the collection is the service-resource model. A business-facing service depends on one or more application components; those components run on compute and storage; their communication depends on network nodes and links. Measurements taken at lower layers become useful only after the dependency is explicit enough to explain how a resource condition could affect the service.
GTK keeps the layered mental model but removes two constraints from the papers. First, a real dependency graph need not be a tree: shared dependencies create joins, redundant paths create alternatives, and retries or asynchronous workflows can introduce cycles. Second, dependency does not prove causality. A topology narrows the candidate set; temporal evidence and component-level signals must still distinguish cause, consequence, and coincidence.
Closed-loop control requires an observed outcome #
The papers repeatedly use a feedback loop: set a target, observe the actual outcome, compute deviation, select a control action, and measure again. They also distinguish feedback from feed-forward compensation. Feedback responds after an outcome deviates; feed-forward uses a measured disturbance, such as rising load or a failed resource, to act before the service indicator crosses its limit.
That distinction remains useful, but current GTK chapters add safeguards the source under-specifies: explicit desired versus applied state, schema validation, transaction boundaries, concurrency control, staged rollout, post-change verification, and fail-closed rollback semantics. Automation without those properties merely executes mistakes faster.
Measure at the user boundary, decompose only for diagnosis #
The source correctly prefers direct service-quality measurement when practical and derives a service estimate from resource measurements only when direct observation is unavailable. This becomes the modern distinction between an SLI specification—what outcome matters to a user—and its implementation—where and how the outcome is measured. Network and resource signals explain a bad service SLI; they should not silently replace it.
Thresholds need state, not repeated comparison #
The collection describes separate activation and clearing thresholds. That is hysteresis: once a condition activates, it clears at a different boundary so noise near one threshold does not cause repeated state changes. The mechanism survives in alert design, while threshold values and delay must come from the service's response time, measurement uncertainty, and error-budget policy rather than a generic product default.
Current-use boundaries #
- The source's FCAPS, ITIL v3, ISO/IEC 20000, CMIP, TMN, WBEM, and ITMS2 catalogues are historical context, not a current architecture. GTK uses task-specific protocol and data-model semantics instead of treating one classification as the system design.
- SNMP and syslog remain relevant, but they are not the whole management plane. Current coverage includes IPFIX, YANG-Push, NETCONF/YANG, RESTCONF, gNMI/OpenConfig, control-plane telemetry, forwarding-plane measurements, active probes, and packet evidence.
- Vendor and market forecasts from 2009–2012, including projections to 2020, are not retained.
- Product matrices, licence prices, named management suites, and the Forrester ITMS2 market taxonomy are not reusable knowledge.
- The proposed centralized, agent-based implementation and its .NET Remoting/WCF-era adapters are not treated as a recommended architecture.
- A sum of component delays is valid only when the measured components lie on the same request path and their intervals are defined compatibly. Queueing, concurrency, retries, fan-out, caching, and asynchronous work can invalidate a naive sum.
- A service-resource graph may contain shared dependencies and cycles; the source's hierarchical acyclic model is an analysis projection, not a universal property of production systems.
Visual audit #
All 203 pages in the folder were rendered and inspected. Load-bearing diagrams were then checked at full resolution.
- Layered infrastructure diagrams: the papers and first deck repeatedly show service, application, compute/storage, and network layers. The concept is retained, but the scanned originals are too dense and tie physical resources to obsolete product architecture. GTK uses a clean dependency graph with typed edges and explicit evidence sources.
- Feedback and compensation loops: several figures distinguish target, measurement, controller, managed system, and disturbance. GTK redraws the useful control semantics around desired, candidate, intended/applied, and observed state, using current NETCONF and NMDA concepts.
- Service-resource graphs and measurement paths: the strongest original figures map one service through application components and servers to switches, routers, and a communication channel, then show measurements returning to a management subsystem. GTK redraws this as a time-varying service graph and states that topology constrains diagnosis without proving cause.
- Process catalogues and product taxonomies: radial ITSM charts, process lists, vendor tables, and maturity diagrams are readable but primarily historical. They are intentionally omitted.
- Optimization plots and formula-heavy controller diagrams: the plots illustrate trade-offs and coordinated control, but their variables are not grounded in reproducible data and some assumptions are left implicit. GTK preserves the causal distinction between feedback and feed-forward without reproducing these curves.
No source image was copied. Every retained visual was independently redrawn for article-width legibility and checked against current primary specifications.