Shayne WinterSystems Integration

HomeProjectsInfrastructure environment

Project 04 / operating infrastructure

Operating services & technical lab

Eight-server Windows infrastructure environment.

An eight-physical-server environment for client services and a separate private technical lab. Only two servers formed the Hyper-V pair; the remaining physical systems carried dedicated infrastructure, collaboration, storage, and support roles.

Domain: Windows infrastructure, hosting, storage, networking, and virtualizationRole: Architecture, deployment, administration, testing, support, and recoveryScale: Eight physical servers total · two Hyper-V hosts · approximately six to ten VMs

Operational problem

Operate real hosted services while maintaining a separate environment for testing and evaluation.

The environment needed to support client and personal hosted services, identity and network functions, remote access, backup, and changing Microsoft server generations without conflating production licensing with private lab evaluation.

Users and constraints

Multiple physical roles with one two-host virtualization pair.

  • Eight physical servers provided the total environment—not eight virtualization hosts.
  • Two physical servers formed the Hyper-V pair for redundancy, load distribution, and evaluation, carrying approximately six to ten virtual machines.
  • A Comcast Business connection with five static IP addresses and a WatchGuard Firebox edge appliance provided the access boundary.
  • Windows Network Load Balancing supported web and other hosted service paths.
  • One additional physical server carried the centralized storage role; storage resided on that server rather than across a shared storage fabric.
  • A Dell PERC 5/E- or 6/E-generation PCIe RAID controller connected the server to an external SAS/SATA disk enclosure. The storage mix included 10,000- and 15,000-RPM SCSI/SAS enterprise drives, 10,000-RPM SATA Raptor drives, and an early PCIe solid-state device.
  • The clean Hyper-V parent VHD used the fast solid-state tier while changing data used the redundant disk tier.
  • Other physical servers carried dedicated roles, including one system for SharePoint and Office Web Apps.

How I framed the problem

Separate service roles, storage responsibilities, licensing, and recovery paths.

The design had to balance physical isolation with virtualization efficiency. The clean parent image needed fast read access, changing virtual-machine data needed redundancy, public services needed controlled ingress and load distribution, and recovery needed both local storage protection and off-site backup.

Client production systems used customer-owned Microsoft volume licensing. I kept the separate TechNet-based lab for testing and evaluation rather than using evaluation licensing for client production workloads.

How the system worked

A conceptual view of the eight-server physical estate.

Sanitized conceptual reconstruction distinguishing the two Hyper-V hosts from the eight physical servers in the environment.

Business Internet edge

Five static IP addresses, a WatchGuard Firebox appliance, routing, remote access, certificates, and bounded public service paths.

Virtualization pair

Two of the eight physical servers supported redundancy, load distribution, evaluation, and approximately six to ten VMs.

Other physical roles

The other six physical systems carried dedicated infrastructure, application, collaboration, management, storage, and backup responsibilities.

Dedicated storage server

One physical server held the PCIe solid-state parent-image tier and used PERC-controlled external SAS/SATA storage with SCSI/SAS enterprise and SATA Raptor drives.

Hosted services and recovery

Windows application services, dedicated collaboration roles, and off-site remote backup for selected data and services.

Sanitized conceptual reconstruction. It distinguishes the verified physical and virtual counts while omitting historical addresses, hostnames, client identities, exact edge model, and recovered topology.

The environment progressed through Windows NT, 2000, and 2003 Server families plus Windows Small Business Server 2000, 2003, 2008, and 2011. Core infrastructure included Active Directory, DNS, DHCP, Exchange, file and print services, RRAS, VPN, ISA Server, and Certificate Services.

Application and management roles included IIS, Apache HTTP Server on Windows, SQL Server, Terminal Services, Microsoft Systems Management Server 2.0 and 2003, SharePoint, Office Web Apps, web hosting, and off-site remote backup.

What I owned

Designed, assembled, administered, tested, and supported the environment.

Architecture and deploymentAssigned service roles across eight physical systems, including a two-host Hyper-V pair and dedicated application, collaboration, and storage roles.
Edge and storage designIntegrated the business edge, network services, dedicated storage server, PCIe solid-state parent-image tier, and PERC-controlled external-SAS/SATA RAID storage.
Service administrationOperated identity, name services, email, web, database, file, print, remote access, management, and collaboration services.
Licensing and lab boundaryKept customer volume-licensed production systems separate from TechNet-licensed private evaluation workloads.
  • 8 physical servers
  • 2 Hyper-V hosts
  • ~6–10 VMs
  • 1 storage server
  • WatchGuard edge
  • PERC external SAS
  • SCSI / SAS
  • SATA Raptor
  • PCIe solid state
  • Windows NLB
  • Off-site backup

How I tested it

Exercise service paths, storage roles, routing, and recovery—not only host availability.

I checked public-service reachability, load-balanced paths, domain and name services, VPN and remote access, virtual-machine startup, parent and differencing-disk behavior, storage-server availability, RAID health, backup completion, and restore procedures. Version changes were evaluated in the lab before I treated them as client production work.

Result

An operating environment that hosted real services and supported technical evaluation.

The environment provided client and personal web, email, remote-backup, identity, collaboration, and infrastructure services while maintaining a separate testing layer for evaluation and learning.

What I learned

Performance tiers, failure domains, and licensing boundaries are all architecture.

Placing the clean parent image on fast solid-state storage while keeping changing data on redundant SCSI, SAS, and SATA Raptor storage was a deliberate performance and resilience tradeoff. Concentrating storage on one server also created a host-level dependency that disk RAID alone could not remove. A modern iteration would add storage-host redundancy, infrastructure-as-code, immutable image pipelines, centralized observability, automated restore tests, stronger segmentation, and current identity controls.

Exact historical throughput, uptime, workload volume, PERC revision, enclosure model, and network topology are not available. The environment was Windows-based; no Linux claim is made.
Next case studyFrontline workflow productivity tools