Cloud & infrastructure · Events & hospitality
This event-venue management platform was hitting public-cloud costs and reliability limits at the same time — booking spikes strained a shared-tenant environment that was already the most expensive option on the table. intSignal moved the ERP onto dedicated infrastructure it operates, cutting the run-rate by roughly 78% while making it faster where it mattered.
The problem
Running production on Azure/AWS with a bill that grew every month and no ceiling in sight.
What we did
Migrated off public cloud onto Proxmox virtual machines running on intSignal-operated dedicated hardware across three regions.
The outcome
Roughly 78% lower monthly infrastructure cost than the equivalent Azure/AWS footprint.
The challenge
Running production on Azure/AWS with a bill that grew every month and no ceiling in sight. Public-cloud vCPU is a shared hyperthread — noisy neighbours and CPU steal made performance unpredictable under load. Network-attached block storage capped IOPS and added latency to every database call. Egress metered per gigabyte, so growth in traffic was punished with a bigger invoice. Backups existed on paper but had never been proven with a real restore, and nothing watched the perimeter.
Public cloud = 100. The same ERP workload on dedicated infrastructure runs near a fifth of the price, against list pricing for an equivalent footprint.
The thinking
An ERP under booking-peak load is exactly the kind of workload public cloud handles expensively and unevenly — provisioned IOPS to buy database performance, and egress metered on every customer interaction. We moved it to dedicated infrastructure because the demand was predictable enough that renting elasticity made no sense, and because local NVMe gives an ERP the storage headroom it needs at peak without a per-IOPS bill. Public cloud wasn't wrong in principle; it was the wrong fit for this workload, and choosing deliberately is what made it both cheaper and faster.
The approach
Migrated off public cloud onto Proxmox virtual machines running on intSignal-operated dedicated hardware across three regions. Dual-CPU, 32-core hosts with DDR5 ECC memory and enterprise NVMe — full physical cores, no shared-tenant CPU steal, ~10× lower storage latency than network block storage. 10 Gbps symmetrical unmetered connectivity, with VM-to-VM traffic running at host speed. Redundant hosts so a single machine isn't the whole business. Delta and incremental backup with tested restores, plus Cloud Firewall (FWaaS) and continuous monitoring. Microsoft 365 administration and day-to-day MSP support.
Why it works
Underneath the cost saving is a simpler benefit: one team owns the platform, its security, and the support around it. Split hosting, firewall, backup and helpdesk across separate vendors and every outage turns into finger-pointing while the business waits. Here there is one provider accountable end to end — infrastructure, protection and people — so problems get solved instead of triaged between suppliers. And because the engagement is substantial, it's a relationship the provider is genuinely invested in protecting, not a low-margin contract to be serviced at arm's length.
Built to hold up
Sized the hosts against real booking-peak load, not average utilisation. Proved backups with restore tests before the old environment was retired. Kept the ERP database on local NVMe so peak booking traffic stops being an IOPS problem. Fronted the platform with FWaaS and monitoring so exposure didn't grow with traffic.
Outcome
The ERP database now sits on local NVMe that shrugs off booking-peak load instead of throttling on provisioned IOPS, the platform is fronted by a managed firewall and monitoring, and egress is no longer a per-gigabyte tax on serving customers. Reliability and cost improved together, which rarely happens in public cloud.
Egress stopped being a line item: 10 Gbps symmetrical unmetered instead of paying per gigabyte to talk to your own customers. Local enterprise NVMe replaced provisioned-IOPS block storage — the most expensive way there is to buy performance in public cloud. Delta and incremental backup cut storage consumption and backup windows. Firewall delivered as a service removed appliance purchases and refresh cycles. One accountable provider instead of separate hosting, security, and support vendors.
How these figures were derived
Effort and outcomes are reconstructed from production systems of record — ticket histories, call detail records, dispatch and SOC/NOC telemetry — read end to end rather than sampled. Cost comparisons are indexed to a baseline, never a fee or rate. Where a figure is modelled rather than measured it is labelled as such, and the model is documented and available to defend on request.
Customer
Shown on request. Some engagements remain confidential by contract.
Professional services
A consulting firm was paying public-cloud prices for a platform that still buckled under load. We moved them onto dedicated hardware we operate, cut the bill by about 82%, and made it measurably faster.
Read ➔Technology & hosting
An analytics provider is only as good as the platform under it. They needed hosting that survives failure, a network that stays up, and someone watching it at 3 a.m.
Read ➔Nonprofit
A nonprofit was running its operations on a local ERP that tied them to one office, one server, and one person who understood it. We migrated them to cloud Business Central.
Read ➔Tell us your environment and priorities — we return scope, ownership, and a plan.