You ship the app, we run the stack
Infrastructure operated by engineers, on our own hardware in Paris. High-availability Postgres, verified backups, 24/7 monitoring, maintained security — no SRE hiring required, no opaque cloud billing.
Today, the choice boils down to three trade-offs
A dedicated server, self-managed
You pay for the hardware, but you carry the operations: Postgres, backups, security, updates, after-hours incidents. You need an in-house SRE — or you accept the risk.
AWS, GCP or Azure
Billing in dollars, costs that climb with load, expensive egress fees. Your data leaves the country, and you end up tied to proprietary services (RDS, ElastiCache, Lambda…) that are hard to repatriate.
A low-cost managed contract
An ops contract on paper, but ticket-based scripted support behind it. No one really knows your workload, the high availability is theoretical, and 24/7 on-call lacks expertise.
Our angle
We operate your infrastructure layer like our own — a single technical contact who designs the architecture with you, runs your database, handles incidents and writes the post-mortems. Pricing adapts to the workload, not a catalog.
- An identified engineer as your contact, not a ticket system
- Architecture co-designed, operations co-piloted with your team
- Quarterly failover drills, written and shared post-mortems
- Workload-based pricing, no hidden overbilling
The scope, in plain terms
No gray area. What we operate, what we don't touch.
What we manage
- OS, kernel, security patches, updates
- Postgres: configuration, tuning, backups, restores, failover
- Reverse proxy, TLS, certificates, renewals
- Infrastructure and business monitoring (RED / USE), on-call alerting
- Security: SSH hardening, fail2ban, audits, CVE patching
- Network, anti-DDoS, IP addressing, BGP, transit (our AS)
- Encrypted backups and monthly restore tests
- Written post-mortems within 5 business days after major incidents
What we don't (unless specifically agreed)
- Application code (no access to your repository)
- Database schema migrations (executed when handed over)
- Business logic, feature flags, application deploys
- End-user support for your product
Three tiers, calibrated to workload
The right tier depends on incident tolerance, not a checkbox list. Each case is studied before any proposal.
Managed Core
depending on workload
Application and database on a single node, disk redundancy, off-site backups.
Early-stage startups, scaling B2B SaaS, critical internal apps — for teams without an in-house SRE who want professionalized operations.
- Single-node Postgres operated: configuration, tuning, backups
- Continuous monitoring with engineer on-call
- Encrypted off-site backups, monthly restore test
- OS security: patches, hardening, fail2ban
- Managed reverse proxy and TLS
- 1 priority ticket per week included
- 99.9% contractual SLA
Managed HA
depending on data size and traffic
Regional high availability: Patroni Postgres cluster spread across our two Paris datacenters (PAR3 + PAR5), tested automatic failover. Equivalent to RDS Multi-AZ.
Applications where downtime has a measurable cost. Paid SaaS, e-commerce, B2B platforms with contractual SLAs. Migration from RDS Multi-AZ.
- HA Patroni Postgres cluster + etcd, 3-node inter-DC Paris quorum
- Failover tested in real conditions (quarterly drills)
- RPO ≤ 5 min, RTO ≤ 5 min on hardware or site incident
- Monitoring, alerting and engineer on-call 24/7
- Written post-mortems with shared action plans
- Unlimited priority tickets, dedicated Slack or Discord channel
- Biannual security audit included
- 99.95% contractual SLA
Multi-PoP / Geo-Redundant
architecture on quote
Redundancy beyond the Paris region and multi-site architectures. Extends high availability outside Paris (Roubaix, Frankfurt, or international PoPs). Equivalent to Cross-Region DR.
Regulated sectors (health, finance, public sector), international audiences requiring local latency, 99.99% SLA requirements or strict business-continuity constraints.
- Architecture co-designed during a dedicated included workshop
- Geo-redundant Postgres replication and steered multi-region failover
- Contractual RPO and RTO matched to business constraints
- Compliance: ISO 27001, HDS, advanced GDPR on request
- Annual pen-test and business-continuity drills included
- Dedicated on-call architect and custom SLA up to 99.99%
Why not another solution?
The same questions keep coming up. Here are honest answers.
Data in France, transparent pricing, a team that knows your schema, co-piloted operations
Costs that climb with load, egress fees, scripted tier-1 support, strong regional dependency
Hardware, operations and engineer on-call — no hiring required
Either an in-house SRE (high salary cost), or accepting incidents without fast response
A technical contact who knows your architecture, HA tested in real conditions, written post-mortems
Standardized tickets and scripts, theoretical HA, 24/7 on-call without deep expertise
Plain Postgres, no lock-in, data with us, migration possible at any time
Proprietary layer on top of Postgres, dependency on a third party and its pricing policy
What makes the promise credible
Our hardware, our racks
We operate on our own infrastructure hosted at Digital Realty PAR5 and DC2Scale PAR3. No IaaS reselling, no middleman.
Our AS, our transit
AS199415 — we announce our own prefixes and set up our peerings directly. StormWall and GCORE anti-DDoS integrated into transit.
HA Postgres in production
Our 3-DC etcd cluster has handled 29 failovers without losing a transaction. High availability is proven, not just claimed.
Security at the core of operations
In-house security expertise, CVE patches applied within 48h, hardening by default, biannual audits included from the HA tier.
Effective sovereignty
Data stored and processed in France, under French law. No extraterritorial reach (CLOUD Act, FISA), no transfer outside the EU.
A contact person, not a ticket
The engineer operating your stack is identified and reachable. They built the architecture, know your database, write the post-mortems.
Let's talk about your workload
A 30-minute conversation to understand your infrastructure, identify friction points and assess whether a partnership makes sense. No sales pitch.
Frequently asked questions
RDS Multi-AZ offers managed high availability but stays opaque: billing in dollars, no identified contact on your database, costs that scale with load and high egress fees. With YorkHost, you run on plain Postgres hosted in France, with transparent pricing and an engineer who knows your architecture.
Yes. Storage, processing and backups stay in our racks in Paris (PAR3, PAR5). No transfer outside the EU, no extraterritorial reach (CLOUD Act, FISA), no US subsidiary. If a secondary PoP is required, it is placed in Europe unless explicitly agreed otherwise.
We notify you before it becomes a problem. No surprise overbilling: when your load approaches the tier's limits, we share a clear summary (metrics and recommendations) and we agree together on the next tier. The transition is downtime-free on the HA and Multi-PoP tiers.
Yes, it's a case we handle regularly. We audit the existing setup (RDS, S3, managed services), propose a migration plan with dual-run and execute the cutover in a short window. No big-bang approach: the migration is co-piloted with your team.
Everything within the tier's scope is included, including incidents, patch application, restores and post-mortems. Anything outside scope (architecture rework, custom development, out-of-tier pen-test audits) is quoted in advance. No billing surprises.
After the first conversation: 1 to 2 weeks for the Core tier (hardware delivery and setup), 2 to 4 weeks for HA (cluster provisioning and failover tests), 4 to 8 weeks for Multi-PoP (custom architecture and dedicated workshop).
99.9% on Core, 99.95% on HA, up to 99.99% on Multi-PoP — with contractual penalties on non-compliance. Our post-mortems are written and shared for every major incident.
The stack is built on open standards (plain Postgres, standard Linux, interoperable formats), with no proprietary lock-in. We assist with data export and orientation to another infrastructure if that's the chosen path. We aim for retention by value, not by trapping customers.
Discuss your workload, not a catalog
30 minutes with an engineer who actually operates infrastructure. Case review, honest take on partnership relevance and guidance elsewhere when needed.