Initial situation In a small call center (18 operators), after expanding the infrastructure, instability began: for some users, RDP sessions periodically freeze for 5-30 seconds, sometimes the sound in the headset drops out, packet loss and voice robotization appear in the softphone. The problem is floating: more often during peak hours, sometimes only at specific workplaces. A regular reboot helps for a short time; there is no obvious error in one place. Technical environment and limitations - Windows Server 2019: AD DS, GPO, RDS Session Host (2 nodes), file server - Clients: Windows 10/11, some thin clients on Windows Embedded - Network: 1G, managed switches, VLAN (Voice/Data), Wi-Fi for some workstations - VoIP: SIP via local Asterisk/FreePBX, USB headsets (several models) - There is administrator access to servers and network equipment, but the office cannot be stopped during the day; windows for active interventions - 20:00-08:00 Moscow time - It is prohibited to upload personal data and call recordings outside the perimeter Task Conduct professional diagnostics and eliminate the root causes of RDP instability and voice quality, confirming the result with measurements. The work should include telemetry collection, analysis, hypotheses, verifications, implementation of changes and retests. Stages and expected results 1) Source data collection and playback - Inventory: OS versions, audio/USB drivers, RDP settings, GPO, SIP/codec parameters, VLAN/QoS topology - Fixation of symptoms: when, from whom, on what devices, correlation with load - Diagnostic plan with a list of measurement points 2) Instrumental diagnostics - Analysis of Windows logs (RDS, TermService, audio, USB, Hyper-V/drivers if available) and events on servers - Network analysis: traffic captures (Wireshark), measurement of jitter/latency/packet loss (iperf3, ping, pathping), checking DSCP/QoS, queues on switches, port errors - RDP check: UDP/TCP parameters, UDP usage policy, MTU/MSS, impact of encryption and audio/USB redirection - VoIP check: RTP statistics, codecs, PLC/VAD, jitter buffer settings, NAT/ALG influence (if any) 3) Implementation of corrections - Preparing a set of changes with minimal risk and the possibility of rollback - Spot edits: GPO/RDP parameters, RDS settings, audio/USB redirection, QoS network policies, VLAN/tagging adjustments, troubleshooting errors on ports/cable/Wi-Fi channel, setting up SIP/RTP if necessary - Documentation: what was changed, where, why, how to roll back 4) Testing after changes - Load tests in the evening window: emulation of parallel RDP sessions and VoIP calls (within the available environment) - Repeated measurements of jitter/packet loss, comparison with the baseline Artifacts that must be transferred - Investigation report: identified causes (at least 2-3), supporting facts and logs/metrics, decisions made - List of changes made with exact paths/parameters (GPO, registry, Asterisk/FreePBX settings, switches) - Checklist for IT administrator: how to quickly diagnose recurrence and what metrics to collect - Recommendations for prevention (no equipment purchases, only setup and regulations) Acceptance criteria (measurable) - In 5 working days after implementation there are no massive RDP freezes: no more than 1 incident with freezes of more than 3 seconds for 50 sessions during peak hours - For a test group of 8 operators: packet loss RTP no higher than 0.5% according to softphone/FreePBX statistics and jitter no higher than 30 ms in the 95th percentile during peak hours - Basic and final metrics before/after are recorded, screenshots/logs are attached - All changes are reversible, there is a rollback plan Work format Remotely via VPN/AnyDesk/Jump host, with coordination of evening windows. Communication via chat, 1-2 calls for 30 minutes at the start and at the end.