Cloud Cost Optimization and Management – Meeting Notes

Overview

Discussion centers on secure access, environment management, and cost optimization across production and non-production (non-prod) environments. Mention of multiple environments (NaviC, SSH, RDP) and access flow: authenticate to network, then to Jambox, then to type server for NaviC/Jambox access.
Production environment has a large server footprint; estimated migration scope is >100 servers, mostly VMs, with past services still in use.
Emphasis on making management traffic come from predictable IPs to enable access lists and tighter control.
Historical context: James Broktar (left) had begun alignment on several items; action points needed to confirm what was flagged, particularly OS versions.
Several cost optimization initiatives discussed: decommissioning, rightsizing, PaaS (Azure) optimization, storage cost optimization, and firewall/IDS/IPS POC with associated networking changes.
Regional coordination required: needs engagement with regional IT teams and TCS (CloudOps) to implement recommendations and refresh reports.
The conversation references a framework for budgeting and forecasting (PXC/PwC reporting), with a focus on aligning reported potential savings with actual actionable items in regional contexts.
Roundtable closes with a push to finalize RI (Reserved Instances) planning, including a move from 1-year to 3-year reservations, to improve savings, and to finalize a model showing regional consumption and reservations.

Key concepts and terms (definitions and context)

IES / IAS: Centrally managed vs. self-managed environments; central management discussed in PWC report figures. The team references IES (likely a specific environment) and the central management model for Azure/PaaS and databases.
EBAT data platform: A large, heavy-user data platform likely sitting in centrally managed environments; suspected as a major driver of 105k Azure PaaS savings if right-sized or terminated.
Samurai: A large environment that’s historically centralized; referenced as a likely high-cost, centralized workload.
Data center licenses vs. Standard: Efforts to remove Enterprise/Datacenter licenses from Windows servers; SQL licenses remain the remaining area to convert to standard licensing.
PaaS (Azure PaaS): Platform as a Service workloads in Azure, including Azure SQL and other Azure PaaS services; targeted for optimization with potential savings stated at $105{,}000$ .
PWC report: A cost-optimization report that estimates savings in centrally managed environments and Azure PaaS workloads; some optimism vs. reality caveats noted.
POC (Proof of Concept): IDS/IPS firewall PoC in the Americas; potential production rollout pending approvals and possibly requiring firewall upgrades.
PXQ model: A cost-tracking or planning model for regional spend and cost optimization; emphasizes reflecting regional changes in future budgets.
Rightsizing: Adjusting instance sizes to better-fit utilization; a major focus area for both compute and database workloads.
Auto-off: Automation to power off non-prod or underutilized instances to reduce cost.
Prism: A planning/optimization approach or tool referenced for reservation planning and cost optimization.
Reservations: Move from shorter to longer-term commitments (e.g., from 12 months to 36 months) to lock in savings.
TCS CloudOps: The outsourcing/partner team responsible for cloud operations and optimization activities.
APAC/EMEA: Regions to be engaged for cost-optimization validation and budgeting.

Major points by topic

1) Security, access, and management posture

Authentication flow: Network authentication -> Jambox -> type server -> NaviC environment or SSH/RDP access.
Goal: Ensure management traffic originates from predictable IP addresses to enable strict access-lists.
Production vs non-prod: Need consistent access-control measures; production requires stricter controls.
OS version flags: Items flagged for OS versions require action; responsibility lies with the team after James Broktar’s departure.

2) Migration and production environment size

Production environment has many servers; >100 servers to migrate.
Mix of VMs and possibly past-services migrating to or being retained in the cloud.
Emphasis on understanding service roles and management traffic to inform firewall rules and access controls.

3) Cost optimization: decommissioning and rightsizing

Decommissioning variance: A cost-saving initiative; Steve acknowledged cost savings. Target savings reported: $2.8 ext{k}$ per month, or $35 ext{k}$ per year.
- 2.8 ext{k} ext{ per month}
  ightarrow 2.8 ext{k} imes 12 = 33.6 ext{k} $per year, approximating the reported$ 35 ext{k}.
Action plan: Hydrate numbers with regional IT teams and with TCS; request TCS to begin implementing the recommendations (e.g., Henrik’s flagged items).
PwC (PWC) report alignment: Potential savings of 195{,}000 $in centrally managed environments; another line shows$ 105{,}000 potential savings for Azure PaaS, though the latter assumes significant terminations and consolidations.
Reality check: Not all savings may be realized; need to review what portion comes from EBAT data platform or Samurai environments, as they are large, centralized workloads.
Azure ADP list and DBs: Adam has pulled Azure ADP lists; DBS list referenced. The team will refresh reports to identify large, quick-win items for Azure PaaS and data-platform workloads.
Data-platform licensing: Consider moving clustered/multi-DB SQL environments from data center licenses to standard licenses; SQL licensing remains the main barrier for cost savings.
Next steps: Adam to refresh Azure PaaS report; identify largest environments for quick wins; engage application/technical owners to confirm requirements or cut if possible.
Bob (Bobbry) involvement: Bob is running in-house tooling to analyze database performance and will advise on right-sizing the database layer.
Governance on how to respond: For each line in the PwC/PaaS data, regions should confirm Yes/No/Needs Clarification; then update the budgeting sheets.

4) PaaS optimization and data platforms

EBAT data platform and Samurai environment are potential major contributors to Azure PaaS spend.
Hypothesis: The 105{,}000 Azure PaaS savings may primarily come from terminating or consolidating large EBAT/Samurai workloads.
Action: Verify the Azure PaaS list; confirm which workloads are centralized (EBAT, Samurai) and which can be terminated or right-sized.
Data-center licensing context: Some data-center workloads migrated to cloud while licensing remained; focus on moving to standard instances for Windows and SQL.
Next steps: Refresh report on Azure PaaS instances; find quick wins; contact owners for validation; if a workload must remain, justify capacity and performance needs.

5) Networking, IDS/IPS, and firewall PoC

IDS/IPS PoC in Americas: POC run done; awaiting approval from James from America to proceed (likely as Americas-first deployment).
Firewall changes: If PoC proceeds, some firewall upgrades may be required; clarify whether new firewall hardware is needed or if upgrades suffice.
PXQ model: Any firewall/network changes should be reflected in the PXQ budgeting model; snapshots of regional networks were taken in July; Americas may incur major changes.
Budget impact: If the PoC goes production in the Americas, budgets should reflect increased costs in that region; costs may be charged directly to sub-subscriptions in each region.
Action: Determine what is required from the team’s side for deployment and upgrade, and align with regional budgets.

6) Storage optimization and backups

Storage cost optimization ongoing with the CloudOps team (TCS): trimming redemption policies and cutting backups in non-prod environments where feasible.
Observed trend: Storage savings reported for July and August; the team is tracking a decrease in reported storage savings due to organic growth and new backups.
Backup growth: Backups add incremental costs annually; as archives and backups mature, costs rise slightly; expect a yearly increase toward year-end.
Auto-off routine: Consider adding automatic power-off for non-prod or unused resources to bolster savings.
Traction tracking: Track auto-off adoption, storage reductions, and rightsizing to measure impact.
Data points noted: In August, RI (Reserved Instances) adjustments show a drop of about 9 per month in the reports; a more detailed modeling is needed to reconcile this with subscription-level changes.
Regional collaboration: Track back with regional teams to ensure backups are trimmed consistently; confirm whether some workloads can be entirely removed from backups.
Prism and long-term reservations: Discuss moving from year-long reservations to 3-year reservations to improve savings; use Prism or a similar model to plan optimal reservation mix.

7) Reservations, RI planning, and regional budgeting

RI refresh work: A person is building a high-level budget view by region/subscription to determine whether to keep, resize, or terminate reservations. The goal is to show a clear path to 26% savings (roughly cited in some plans).
Proposed shift: Move from one-year commitments to three-year commitments to improve savings and align with long-term capacity planning.
Challenges: Complex script logic to map reservation sizes to actual workloads; need to balance risk (underutilized reservations) with potential savings.
Timeline: Begin with a quick, regional, high-level view to guide budgets for 2020/2021 planning; implement earlier rather than later to manage risk.
Next steps: Share initial views with colleagues to review reservation sizing and update the budget sheets accordingly.

8) Regional coordination and responsibilities

Regions involved: APAC, EMEA; need to coordinate with regional IT teams and TCS for actioning recommendations and reporting.
Communication cadence: Weekly calls with regional teams; track traction on auto-off, storage, and rightsizing.
Ownership: Identify application owners/technical owners for Azure PaaS workloads; obtain confirmation on requirements or permission to terminate/scale down.

9) Roundtable: current status and ongoing concerns

RI work in progress: One participant is close to finalizing a view of RI allocations by family size, region, and subscription; aiming to understand 20-30% of 3-year reservations vs. 60% of 1-year reservations.
Forecasting adjustments: Expect a drop in savings when workloads move to power-off or are decommissioned; also expect savings to adjust as new workloads are introduced or re-sized.
The script challenge: Balancing reservations across multiple D-series or compute families; the process is complex due to how reservations are provisioned versus how compute is consumed across regions.
Unified compute across three regions: Acknowledge that compute is now unified across regions, enabling better cost efficiency and easier rightsizing.
Early actions: Request early view of RI planning to begin optimizing reservations sooner rather than later.

Numerical references and formulas (LaTeX)

Monthly decommissioning savings: 2.8 ext{k} ext{ per month}
Annualized decommissioning savings: 2.8 ext{k} imes 12 = 33.6 ext{k} ext{ per year} ext{(approximately }35 ext{k} ext{ per year reported)}
Centralized savings (PwC): 195{,}000
Azure PaaS savings (PwC): 105{,}000
Target reservation savings reference: 26 ext{ ext{%}} (as cited in planning discussions)
Reservation horizon proposal: 12 ext{ months}
ightarrow 36 ext{ months}$$
Storage/funding notes: Regional budgets will reflect costs directly charged to sub-subscriptions in each region; alignment across APAC/EMEA is needed

Action items (summary)

Clarify OS-version flags and ownership for items flagged after James Broktar’s departure; assign responsible owner.
Update Azure PaaS inventory; Adam to refresh Azure PaaS report; identify EBAT/Samurai workloads for potential right-sizing or termination.
Engage with Paprika/Merkel contact (database team) to verify DB license status and potential consolidation to standard licenses; confirm DBA (DBS) list and responsibilities.
Validate POC roadmap for IDS/IPS firewall in Americas; determine deployment scope and required firewall upgrades or replacements.
Hydrate PwC regional figures with regional IT and TCS inputs; update the PWC-based cost-savings estimates by region.
Develop a regional RI plan: build an initial view of family sizes and 3-year reservation opportunities; map to budget lines and share for review.
Implement storage optimization actions: extend auto-off routines; trim backups for non-prod environments where feasible; track monthly savings and update baselines.
Align regional budgeting with PXQ model: reflect expected costs for changes in Americas; ensure regional teams budget for potential spikes.
Schedule follow-up on license reallocation from Data Center to Standard for Windows servers; track SQL licensing separately.
Maintain ongoing right-sizing efforts (compute and databases) with Bob (Bobbry) and regional owners; report progress in weekly calls.
Prepare a consolidated, regional cost-optimization dashboard: RI status, right-sizing, decommissioning, and storage, with Yes/No/Needs Clarification for PwC lines.

Notes on sentiment and risk

Optimism about potential savings exists but requires careful validation against actual workload footprints and regional approvals.
Some initiatives (like POC deployment in Americas) depend on approvals and concrete changes to security infrastructure; delays could shift timelines and budgets.
The team recognizes the complexity of reservations planning across multiple regions and compute families; a rigorous, data-driven approach is essential to avoid over-committing or under-committing resources.

References to people and roles (contextual)

James Broktar: Previously involved in alignment; left the team; action items depend on what he flagged.
James Procter: Mentioned as the correct James in some parts; not to be confused with James Norris.
Henrik: Flagged some recommendations for action.
Adam: Pulled Azure ADP lists; to refresh Azure PaaS report.
Bob/Bobbry: Runs internal tooling to analyze database performance for right-sizing.
Paprika: A contact from Merkel group referenced in discussions about data center and licensing.
Callers/regions: APAC, EMEA, Americas; regional IT teams and TCS CloudOps to be engaged for execution and budgeting.

Summary takeaway

The team is actively pursuing a multi-pronged cost-optimization program spanning decommissioning, rightsizing, storage optimization, and reservation planning, with a strong emphasis on regional alignment and governance. There is a clear pathway to significant savings, but realization depends on timely validation with regional owners, accurate inventory of large workloads (EBAT/Samurai), and concrete changes in regional budgets and subscriptions. The next steps are to refresh key reports, confirm ownership, implement quick wins, and finalize RI planning to enable more aggressive long-term savings.