IT and maintenance terms A to Z

Backup rule of thumb: 3 copies of the data, on 2 different media, 1 copy offsite (or air-gap). Minimum standard for ransomware resilience.

Physical or logical isolation of a system from the network — standard for backup archives that need to be ransomware-proof.

Storage system exclusively with SSDs/NVMe — no rotating disks. Sub-millisecond latency, standard for modern databases and VDI.

Recording and maintaining all hardware assets with serial numbers, locations, contracts, EOSL dates. Basis for maintenance inventory and budget planning.

Mandatory contract under Art. 28 GDPR between controller and processor. For TPM maintenance necessary when service technicians may have access to personal data.

BaFin requirements for IT security in banks and financial service providers. Among other things, regulates documented maintenance processes with defined SLAs.

Physical server without hypervisor — the application runs directly on the operating system. Maximum performance, no virtualization overhead.

BCP: plan for business operations during major disruptions. DR: technical restart of IT systems. Together the basis for KRITIS- and BAIT-compliant IT.

Routing protocol of the internet — and in large datacenters also used internally for spine-leaf topologies. Considered complex but proven.

Modular server in a thin "blade" form factor, several within a shared chassis. High density per rack unit, shared power supplies and cooling.

Out-of-band management chip in the server. Allows remote management (power cycle, console access, sensors) even when the main system is unresponsive.

Method by the German BSI for structured IT security. Modular structure, basis for ISO 27001 certification per BSI methodology.

Load test over several days to expose latent hardware defects before productive use — standard after refurbishing or in tier-1 banks.

BSI requirements catalogue for cloud providers. The C5 attestation is effectively mandatory for cloud contracts with German federal authorities and KRITIS operators.

Structured cabling in the rack — labeled, color-coded, with strain relief. Clean cable management halves repair time during component replacement.

CAPEX: capital expenditure (hardware purchase), depreciated over useful life. OPEX: ongoing operating expense (maintenance, cloud, power). Maintenance contracts are OPEX, refresh is CAPEX.

Hardware-refresh strategy that prioritises CO₂ footprint: refresh only where the new generation delivers > 30 % energy efficiency gains, otherwise extend the lifecycle via TPM.

Shifting workloads to times/regions with green electricity. Example: backup jobs at night when wind power is available. Increasingly relevant for ESG reporting.

ITIL process for the controlled introduction of changes. Every hardware maintenance with system impact should have a documented change request.

Documented plan for how a workload exits the public cloud. Required by BAIT and KRITIS for critical applications.

Central database of all IT assets (servers, storage, switches, contracts, maintenance contracts) and their relationships. Prerequisite for serious lifecycle management.

Hosting your own hardware in a third-party datacenter (instead of your own server room). The provider supplies power, cooling, network connectivity and physical security.

Consolidation of multiple maintenance contracts to a common end date so renewal negotiations can be conducted bundled.

Component replacement only possible after a system shutdown (e.g. CPUs, mainboards, RAM without hot-plug support). Requires a maintenance window.

Classic three-layer campus network topology: access switches (user ports), distribution (aggregation, VLANs), core (backbone, routing).

Specialized air conditioning for datacenters — precise temperature and humidity control, higher cooling capacity than normal office equipment.

EU directive for sustainability reporting. Required from FY 2024 for large enterprises — including Scope-3 emissions and circular-economy KPIs of installed hardware.

Storage directly connected to the server (SAS, SATA, NVMe), no network in between. Fastest variant, but not shareable between servers.

Software for centralized monitoring of power, cooling, capacity, asset tracking and cabling. Bridges the gap between IT monitoring (servers) and facility management (power/HVAC).

Duplicate data blocks are stored only once. Reduces storage requirements typically by a factor of 5–20 for backup workloads.

Structured fault-finding path: from symptom via logs (BMC, syslog) and POST codes to the root cause. Saves swap attempts and escalations.

German standard for secure data-carrier destruction in 7 security levels. Banks/authorities require levels 5–7 with proof of destruction.

Liquid-cooling variant: cold plates sit directly on the CPU/GPU. Lower risk than immersion, no full submersion. Development path for AI workloads.

Hardware that arrives already defective. Replaced immediately via the RMA process, usually within 30 days of receipt.

EU regulation for the financial sector from January 2025. Requires explicit continuity plans, third-party registers and stress tests — TPM providers must actively meet DORA requirements.

EU-wide digital product passport with origin, material and lifecycle data. Becoming mandatory step-by-step from 2027 for IT hardware (EU Ecodesign Regulation).

EU-wide data protection regulation (GDPR). Maintenance contracts must also cover GDPR compliance — TPM providers, for instance, may not dispose of data carriers without a DPA.

Server RAM with built-in bit-error detection and correction. Mandatory for productive workloads — standard desktop RAM does not have it.

Data processing close to the source (branch, factory, remote site) instead of centrally in the cloud. Important for low latency and data protection.

Date from which a manufacturer no longer sells a product. Maintenance is often offered for several more years (see EOSL).

Date from which a manufacturer officially no longer provides maintenance, patches or spare parts for a hardware model.

Expiry of the manufacturer warranty supplied with the device. After that: either costly OEM maintenance extension or switch to a TPM provider.

Electrostatic discharge. Technicians wear ESD wrist straps during hardware work — even a 50V spark can destroy chips, often latently without immediate failure.

High-speed protocol for SAN connections (8/16/32/64 Gbit). Dedicated HBA cards and FC switches, separated from the Ethernet network.

Installing or updating hardware firmware (BIOS, BMC, RAID controller, disk firmware). Often part of the maintenance intervention to fix bugs.

Datacenter cooling using outside air (instead of mechanical compressors) — possible on cool days or via water heat exchanger. Saves 30–60% in electricity costs.

Field-replaceable component — hard drive, power supply, fan, memory module. TPM technicians bring the FRU as a spare part.

Systematic reduction of the attack surface: disable unnecessary services, change default passwords, keep firmware up to date. Mandatory step at every commissioning.

Compute, storage and network in one software-defined box — examples: Nutanix, Dell VxRail, HPE SimpliVity. Easier scaling, but vendor lock-in.

Server-row arrangement so cold supply air enters on one side and warm exhaust air exits on the other. Standard in modern datacenters.

Component replacement during operation without server shutdown. Standard for hard drives, power supplies, fans in enterprise servers.

Mix of own datacenter and public cloud. Sensitive workloads stay on-prem, elastic workloads run in the cloud.

Mix of SSDs (for hot data) and HDDs (for cold data) in the same array, with automatic tiering. Cheaper than all-flash for large data volumes.

Global cloud providers at mega-scale: Amazon AWS, Microsoft Azure, Google Cloud Platform. Typical: pay-per-use, global availability, broad service portfolio.

Software layer that abstracts multiple virtual machines on one physical server. Examples: VMware ESXi, Microsoft Hyper-V, KVM, Proxmox.

IDS detects attack patterns in the network, IPS actively blocks them. Today usually integrated as a module in NGFW or XDR platforms.

Servers are submerged in dielectric fluid — no fans, no air conditioning, significantly higher power density. Advanced variant of liquid cooling, common for HPC/AI clusters.

Structured handling of unplanned outages. Prioritization by impact (P1/P2/P3) determines SLA response and escalation path.

Block-storage protocol over normal IP Ethernet. Cheaper alternative to Fibre Channel, but higher latency.

International standard for environmental management systems. In IT relevant for hardware disposal, energy efficiency and ESG reporting.

International standard for information security management systems (ISMS). Often required as a prerequisite for enterprise contracts.

German law since 2021 tightening KRITIS provisions. Lowers thresholds, mandates BSI-certified components, and introduces a "state-of-the-art" obligation.

Best-practice framework for IT service management with standardized processes (Incident, Problem, Change, Capacity, Service Level). Industry standard in larger IT organizations.

German regulation for operators of critical infrastructure (energy, water, healthcare, finance). Tightened security and reporting obligations under the BSI Act.

Planned replacement of aging hardware with a new generation, usually recommended by OEMs on a 3–5-year cycle — TPM economically delays the refresh.

Direct water or glycol cooling at the hot components (CPUs, GPUs) instead of air. Standard for high-density racks above ~30 kW per rack.

Distributes incoming requests to multiple backend servers (Layer 4 or Layer 7). Examples: F5 BIG-IP, Citrix NetScaler, HAProxy.

Loan device during the repair of a customer system. TechCare premium SLA tiers include a loaner pool for critical workloads.

Central collection of system, application and security logs (syslog standard). Forms the basis for SIEM, forensics and compliance audits.

Industry standard for magnetic tape backup (LTO-7/8/9 ≈ 6/12/18 TB native). Lowest cost per TB, ideal for long-term archives and air-gap backups.

One-off payment to the OEM for a longer guaranteed maintenance period instead of yearly extension. Rarely economical compared to TPM.

German BaFin regulation for bank IT outsourcing. Requires written contract, clear responsibility, right to instruct/audit, contingency plan, continuity assurance.

At least two independent login factors (knowledge + possession, e.g. password + hardware token or authenticator app). Mandatory in BAIT/KRITIS environments.

Continuous monitoring of hardware and service status via SNMP, syslog or vendor APIs. Examples: Nagios, Zabbix, PRTG, Prometheus, Datadog.

Statistical average of how long a component runs between two failures. Published by the manufacturer per model.

Average time from fault detection to recovery. Decisive for system availability calculations.

Deliberately using several hyperscalers in parallel — to fight vendor lock-in, for regional coverage, or to combine best-of-breed services.

File-based storage over an IP network (NFS / SMB). Simpler than SAN, ideal for file shares, backup targets, home directories.

SLA tier: response or spare part by the next business day. Typical for non-critical systems or mid-tier maintenance packages.

Firewall with application awareness, IDS/IPS, SSL inspection and threat intelligence — beyond classic port/protocol filtering.

EU directive with national transposition from October 2024 — extends KRITIS to 18 sectors with ~30,000 mandatory companies in Germany. Requires SBOM, incident reporting in 24h, supply-chain security.

Fast SSD connection directly to the PCIe bus, significantly faster than SATA/SAS. Standard for modern all-flash storage and high-IOPS workloads.

Manufacturer of the hardware (HPE, Dell, IBM, Cisco, …). OEM maintenance = maintenance contract directly with the manufacturer, usually significantly more expensive than TPM.

On-site maintenance work at the customer — fault diagnosis, parts swap, cabling. Counterpart: offsite (drop-off repair in the workshop).

Two distinct end-of-life dates: software patches (security/OS) often expire years before hardware EOSL. Asynchronous lifecycle is critical for compliance.

Structured process for applying security and bug-fix patches. For hardware: firmware updates for BMC, RAID, disk; for software: OS, hypervisor, applications.

Power distributor in the rack — either a simple power strip or a managed variant with current measurement per outlet and remote switching.

Time-limited pilot of a provider with real workloads to verify functionality and SLA in practice. With TPM often a 30-day fixed-price pilot before a full contract.

Contract penalty for SLA violation — e.g. credit note for every hour of exceeded response time. Important negotiating position, often 5–15% of the monthly fee per violation.

Diagnostic routine the server runs at every start. POST codes (beep or LED patterns) help technicians identify defective components without booting.

Phase after factory warranty expiry, when hardware remains in operation and requires maintenance — the most common TPM entry point.

Ratio of total power to IT power. PUE 1.0 = perfect (no overhead for cooling). Modern datacenters reach 1.2–1.5; old server rooms often >2.0.

Height unit of a 19″ rack (1U = 44.45 mm). Servers usually come in 1U, 2U or 4U; a standard rack is 42U high.

Classic server in 19″ rack form (1U, 2U, 4U height). Standard form factor for datacenter hardware.

Multiple hard drives as a logical group — for redundancy (RAID 1/5/6/10) or performance (RAID 0). Standard in server and storage systems.

N+1 = one reserve component on top of N required (e.g. 4+1 power supplies). 2N = full duplicate provisioning of every component.

Planning rhythm for hardware renewal, often a 3-, 4- or 5-year cycle. OEMs push for 3 years; TPM strategy extends to 7–10 years.

Fully overhauled hardware: tested, cleaned, with updated firmware. Often used as a spare-parts source for EOL models where the OEM no longer produces originals.

Mandatory labelling (1–10 scale) showing how easily a product can be repaired. Arriving in 2027 for IT hardware via the EU Right-to-Repair regulation.

Bringing workloads back from the public cloud to your own datacenter — typical when cloud costs become unexpectedly high or compliance requirements shift.

Strategic decision at the EOSL transition: replace hardware with a new generation (replacement) OR overhaul it and keep running with refurb parts (refurbishment). 3–5× cost difference.

Synchronous or asynchronous data mirroring to a second site — basis for disaster recovery.

Pre-phase without contractual intent: vendor market research, capability gathering. Lead-up to RFP.

Formal proposal request with detailed requirements. Usually ends in a comparison matrix of multiple providers. Common volume: from €100k upwards.

Price request for a clearly specified service. Faster than RFP, no requirements document, suitable for standard maintenance contracts.

Authorization to return defective hardware. With TPM contracts, the RMA process runs through the maintenance provider, not the OEM — faster and without warranty disputes.

Ratio of profit to investment cost over a period. With TPM migration: typical payback < 12 months, ROI > 200% over 3 years.

Maximum data loss a business can accept (e.g. RPO 1h = at most 1h data gap after disaster). Determines backup and replication frequency.

Maximum recovery time after an outage. RTO 4h = service must be running again within 4h of the outage. Key driver of SLA tier choice.

Dedicated high-speed network for block storage, usually Fibre Channel. Storage appears to the server as a local disk.

Machine-readable inventory of a software's contents (all dependencies, versions, licenses). Mandatory under NIS2 and the EU Cyber Resilience Act from 2027.

Software-controlled WAN connections over internet, MPLS or LTE. Standard for multi-site companies, replacing classic MPLS-only architectures.

Network configuration centrally from software instead of individual switch CLIs. Simplifies large datacenters, complicates small setups via controller complexity.

Central collection and correlation of security logs from servers, network and endpoints. Examples: Splunk, IBM QRadar, Microsoft Sentinel.

Single-sourcing: one manufacturer for all systems (volume discounts, vendor lock). Multi-sourcing: several OEMs (risk diversification, higher administrative complexity).

Contractually guaranteed response and recovery time per incident — e.g. "4h response 24×7" or "Next Business Day Parts".

Point-in-time image of a data state, without copying the data. Ideal for fast recovery after ransomware or operator error.

Regional warehouse for OEM-original spare parts. Decisive for fast SLA response — TechCare maintains depots across the DACH region.

Documentation requirement (BAIT, KRITIS, NIS2) of who supplied spare parts, serial number, refurbishment status, chain of custody. Verifiable per intervention.

Two-layer datacenter topology: leaf switches (ToR) connect servers, spine switches connect all leaves. Constant latency, linear scaling.

Multiple physical switches managed as one logical device — shared configuration, redundant uplinks, hot-swap of individual switches possible.

Automatic movement of data between performance classes (NVMe → SSD → HDD → tape) based on access frequency.

Total cost of a piece of hardware over its entire lifetime — acquisition + maintenance + power/cooling + personnel + disposal. Calculated per system over 5–10 years.

Datacenter classification by the Uptime Institute. Tier I = simple (99.67% uptime), Tier IV = fully redundant (99.995%, ≤ 26 min downtime/year).

Industry standard of the German automotive industry for information security. Mandatory for suppliers — maintenance contracts in the automotive supply chain need TISAX evidence.

Mandatory annex to the DPA: concrete protective measures (access controls, encryption, pseudonymization, backup strategies). Regularly checked by the auditor.

Switch at the top of the server rack to which all servers in the same rack connect. Saves cabling and is the standard topology in modern datacenters.

Vendor-independent hardware maintenance by a third-party provider — typically 30–70% cheaper than the manufacturer's original maintenance, with identical or better SLAs.

Battery-backed power supply that bridges short outages (seconds to minutes) and allows orderly shutdown during longer outages.

Dependency on a single manufacturer that makes switching expensive/risky. TPM reduces lock-in: one contract covers multiple OEMs without having to buy new hardware.

Multiple logical systems on one physical hardware. Saves power, space and cost — but: hardware failures hit more workloads at once.

Logical separation of a physical network into multiple isolated segments — e.g. to separate office, server and management traffic.

Scheduled time for disruptive maintenance interventions (firmware updates, hardware replacement). Common slots: weekend night, quarterly downtime.

EU directive on disposal of electronic waste. Implemented in Germany via ElektroG. TPM providers with buyback supply WEEE-compliant disposal certificates.

Security architecture that re-verifies every device and every user per access — no implicit trust based on network location.