Physical resilience
Man-traps, badge + biometric layers, CCTV retention, cage-level locks—you should expect evidence-friendly access logging if your compliance questionnaires ask for SOC2-style controls downstream of the hyperscaler SLA chain.
Facilities
Datacenter design engineered for repeatable performance
Power diversity, disciplined change management, measurable storage classes, and support that understands kernel updates vs application faults. Pair this briefing with our routing & network architecture page for traceroute-ready context.
Growing network of live sites, stores & WordPress installs on NVMe-backed plans.
- Tier-1 facilities & monitored rollouts
- Ryzen-class nodes & NVMe tiers on qualifying SKUs
- Live uptime & incident history on status
On this page
Reliability framing
Service snapshot for procurement decks
Marketing numbers help short-list vendors; contractual credits and exclusions live in executed agreements—always reconcile this overview with invoices, DPAs where applicable, and acceptable use policy guardrails.
| Domain | Declared posture |
|---|---|
| Facility class | Industry-standard hardened datacentres with audited physical access workflows—exact tier nomenclature depends on leased hall. |
| Planned uptime target | 99.9% monthly trajectory on covered subsystems referenced in SLA addenda—not a blanket promise on every bespoke integration. |
| Change communication | Announcements via status; emergency mitigations prioritise restoration over ceremonial notice. |
| Data residency cues | SKU-level region labels in checkout—consult legal stakeholders when sovereign placement is compulsory. |
| Concern | Primary reference |
|---|---|
| BGP, transit, jitter | Network architecture & traceroute playbook |
| CPU socket family | SKU detail here + provisioning receipts post-order |
| Historical uptime | Status incident archive & your own monitors |
| Incident response RACI | Support tiers & ticket templates |
Facility systems
Power, cooling & physical controls
HostStack leases footprints inside facilities operated by specialised datacentre providers. While each hall differs slightly, deployments target environments with diversified utility feeds, on-site rotary or static UPS bridges, standby generation for sustained outages where economically justified, and hot/cold containment discipline that preserves stable inlet temperatures.
Environmental stewardship
Facility partners publish PUE targets and renewable energy fractions on their dashboards; HostStack aligns expansion with operators who publish transparent efficiency metrics—we can forward those artifacts when procurement requests them explicitly.
Compute & storage
Node classes, virtualization & I/O stewardship
We standardise around high-clock AMD Ryzen and EPYC-class processors for performance SKUs—with DDR4/DDR5 memory speeds matched to motherboard validation lists. SSD tiers use enterprise NVMe where durability and sustained write profiles warrant it.
| SKU family | Compute profile | Storage class | Virtualisation stance |
|---|---|---|---|
| Shared LiteSpeed stacks | Burst-friendly vCPU pooling with CloudLinux-style isolation where licensed | Distributed NVMe with inode & IOPS quotas per plan | Multi-tenant; noisy-neighbour policing via LVE & policy |
| Ryzen KVM VPS | Dedicated vCPU pinning per package table | Locally attached NVMe or distributed block per region | Guests run full virtio stack; hypervisor patching coordinated |
| Managed WordPress tiers | Opcode-friendly CPU pools + object cache integration | SSD tuned for concurrent small-file reads | Platform guardrails enforced (plugins, outbound mail) |
Integrity note
Synthetic benchmark screenshots age quickly—we prefer sharing steady-state utilization graphs from production-adjacent test tenants over one-off Geekbench marketing shots. Invite your QA team to run controlled load generators after onboarding.
Platform pillars
What “enterprise infrastructure” means in practice
Hardware
Ryzen / EPYC-class tiers
Nodes sized for predictable PHP, database, and automation workloads—with NVMe storage classes matched to plan limits.
Facilities
Redundant power paths
Facility-grade redundancy paired with disciplined change windows—paired with backups and failover you configure for your apps.
Software stack
LiteSpeed, cPanel-ready
Panels and optimisation layers deployed where licensed—exact bundle is spelled out per product in the client area.
Integrity
No phantom regions
Regions and uplinks tie back to selectable footprints on eligible lines—avoid hosts that bury geography in fine print.
Stack partners
Powered by software you already trust
Partners and components vary by SKU; every order page lists exactly what ships on that line.
AMD EPYC / Ryzen-class
NVMe SSD
LiteSpeed
CloudLinux
cPanel
DirectAdmin
Imunify360
OVHcloud
Hetzner
Up to 17 TB DDoS mitigation
Premium uplinks
Stack components vary by product type—see checkout for your exact SKU.
Runbooks
Change management & customer-visible maintenance
Hosting reliability is partly hardware and partly discipline: kernel CVEs must ship, RAID arrays need predictive disk swaps, TLS roots rotate, firmware mitigations arrive for transient execution bugs. HostStack publishes intent through status calendars when customer traffic may shift.
Class A — transparent
Scheduled hypervisor lifts, upstream fiber relocations with controlled reroutes, certificate authority root bundle updates that require graceful reload sequences.
Class B — reactive
Thermal excursions, PSU faults, latent disk SMART warnings—we page on-call engineers and merge telemetry into your existing HostStack ticket stream.
Backup & DR ownership
Snapshots help but are not theatrical DR unless you routinely restore. Drill quarterly; store immutable copies cross-region where compliance demands it—even if HostStack facilitates transport.
Evidence
Speed tests, traceroute & looking glass etiquette
We intentionally do not publish a single static marketing test IP—hardware refreshes churn addresses, footprints differ across SKUs, and procurement teams extrapolate falsely from one traceroute screenshot.
Throughput tests
Exercise both single-flow and parallel TCP sessions; watch bufferbloat during Wi‑Fi legs; rerun during Indian evening peak.
Traceroute etiquette
Share MTR traces that cover at least 500 packets; annotate ASNs where your security team insists on BGP forensics prep.
Operational truth
Historical uptime snapshots sit on our status stream—mirror them against your Grafana or Better Stack dashboards after cut-over.
| Evidence item | Owner | Acceptance heuristic |
|---|---|---|
| Facility attestation bundle | Legal / infra | Operator SOC / ISO dossier matches invoice POP code |
| Disk endurance spec sheet | Buyer storage lead | DWPD or TBW within planned DB growth curve |
| Backup restore rehearsal log | Your DevOps squad | Quarterly table-top with checksum verification |
FAQ
Common questions
Can I use a HostStack-hosted looking glass?
Dedicated looking-glass tooling is rolled out sparingly—if you need BGP path visibility before procurement, cite your ASN and workloads in sales and we will advise what is feasible.
Are datacenters physically in India?
HostStack is operated from Kolkata. Shared optimisation plans prioritise fast delivery to visitors in India. KVM footprints may expose non-IN regions at checkout—select what matches latency, sovereignty, and back-up workflows.
How does this relate to /network/?
This page is about cages, disks, CPUs, and change discipline. Routing, peering posture, and DDoS layering are spelled out alongside measurement guidance on our network page.
Who owns backups and disaster recovery rehearsals?
You retain accountability for restores, credential custody, ransomware scenarios, and off-site copies—even when HostStack provides snapshots or couriered backup SKUs.
Can procurement receive facility certification bundles?
Yes—open a ticket referencing your SKU, auditor name, and data categories. Sensitive PDFs ship under confidentiality terms because operators restrict redistribution.
Infrastructure that survives scrutiny
Bring your charts, traceroutes, and TTFB numbers—we would rather onboard teams who validate than cancel after week one surprises.