Cloud – ScalarX Limited

Insourcing, Autonomy, and Independence

Published on December 12, 2025 in Cloud

Hello everyone,

As some of you already know, ScalarX operates with very few external technical dependencies (SaaS, and so on).

Recent events around the world (recent incidents involving AWS, Cloudflare, and others), combined with the current political environment, seem to bring us a little closer every day to a loss of individual freedoms and privacy, both personal and professional.

For more than 25 years, I have always worked to build my companies while preserving security and privacy as much as possible, along with the autonomy and independence of both my organizations and those of our clients.

Where We Stand Today

ScalarX is no exception to this rule—quite the opposite: legal and technical structures across several continents, no external shareholders, no exclusive supplier contracts, and a decentralized, multipoint infrastructure operations and maintenance model. Everything was designed with this objective in mind.

Services that directly handle critical data, such as technical support, have always been 100% in-house.

However, a few external services remain. Most are noncritical and depend on third-party providers or vendors (for example, Trello, our accounting system, CRM, SolarWinds, and others).

We will retain some of these services, such as Pingdom, because they provide an independent source of evidence when ScalarX and a client need to assess an incident.

We are not dependent on these services: not only can they be moved to other providers, but we also have internal interactive tools that allow us to surface this information without “breaking” the decentralized nature of monitoring and operational oversight.

What Comes Next

For others, such as Trello or the CRM, we have begun bringing these functions in-house to remove these dependencies. For us, this is the only way to guarantee technically, and not just legally, that no one is “playing” with data that could potentially concern them.

We are using this work to test and certify several of these tools with StackX. In a second phase, we will be able to offer solutions for those who want to deploy them.

At the same time, we are also working on a hardware solution that will allow you to create independent, automated local backups of your hosted sites and applications, in addition to the included backups.

Have an excellent weekend, everyone!

🔴 RAM/NVMe Market Alert: Impacts and Opportunities for ScalarCloud Clients

Published on December 10, 2025 in Cloud

Hello everyone,

As some of you have already noticed, RAM prices and NVMe storage prices have been skyrocketing for some time.

The figures speak for themselves: the same RAM model (G.Skill Ripjaws M5 RGB 96 GB) from the same supplier (LDLC Pro) that I bought in September has gone from €374.96 excluding VAT to €1,609.96 excluding VAT. That is an increase of more than 400%. At the same time, the price of the NVMe drive (Corsair MP700 PRO SE 4 TB) rose from €499.96 to €583.29.

The cause? Explosive demand for “AI compute.” It has become far more profitable for manufacturers to prioritize GPU-specific RAM, drastically reducing production capacity for the rest of the market.

The situation accelerated on December 3 with the announcement that Micron was withdrawing from the consumer market (the Crucial product line) to focus on data centers. The result: pressure on inventory is intensifying. The indicators point to a potential increase of 500% over the next 6 months, with a peak expected in summer 2026.

What Does This Mean for You?

1. Current ScalarCloud clients: Your prices remain unchanged. We secured our supplier contracts for several years when your infrastructure was deployed. Your current resources are protected.

2. The current inventory dilemma: We have infrastructure in stock. Financial logic would suggest using it to attract new clients at a high price. I made a different decision.

Starting today, we are prioritizing our partners and existing clients. Throughout December, you can purchase RAM at the current price, before the increase, while supplies last.

If you expect to need additional resources next year, now is the time to secure them. If any inventory remains after this operation, it will be offered at the standard price through December 31.

– Billing starts, even if allocation takes place earlier: January 1, 2026. – Commitment: None. You can return this RAM at any time, at which point it will return to inventory and be offered at the new price.

Looking Ahead

For new infrastructure, excluding current inventory, we are still considering the fairest model: passing through the actual cost of RAM, with a significant impact for some, or spreading the increase across all services. Simulations are underway, and a hybrid approach may emerge.

If you expect to need an expansion or new machines, contact us quickly. We estimate that the current inventory will be depleted by the end of December, excluding the “spare” safety stock that we must retain for maintenance.

Have an excellent day, everyone.

Backup Infrastructure: Some News from the Front!

Published on April 27, 2025 in Cloud

Following the launch of the new ScalarCloud production infrastructure, now distributed across 4 datacenters in Europe (and 2 in the United States), it is time for the backup infrastructure to evolve.

Objective: extend the security-through-distribution model already applied to production to the backup infrastructure.

In practical terms, instead of having one datacenter dedicated to backups and the others to production, we now have backup infrastructure distributed across 4 separate datacenters.

To ensure the lowest latency between each production site and its backup site, we conducted several rounds of measurements. The result:

– Gravelines will be backed up to Roubaix
– Roubaix will be backed up to Gravelines
– Limburg will be backed up to Strasbourg
– Strasbourg will be backed up to Limburg

Finally, SBG remains the standby site for production for now, while Limburg becomes the standby site for the backup infrastructure.

ScalarCloud Europe 2025 Infrastructure: Updates

Published on November 27, 2024 in Cloud

Hello everyone, we have just approved the new ScalarCloud Europe 2025 infrastructure, and the hardware order was placed today.

Former ScalarCloud infrastructure (2021):

– Processor: 2 × Intel Xeon Gold 6242R – 20c/40t
– RAM: 768 GB ECC 2933 MHz
– Storage: 12 × 3.84 TB NVMe SSDs (2 arrays made up of 6 NVMe SSDs in RAID 10)
Redundant connection: 2 x 10 Gb/s

Points of presence: UK (ERI), DE (LIM), FR (RBX, GRA)

New ScalarCloud infrastructure (2025):

– Processor: 2 x Dual AMD EPYC 9354 – 64c/128t
– RAM: 1.5 TB DDR5 ECC 4800 MHz
– Storage: 18 x 3.84 TB NVMe SSDs (3 arrays made up of 6 NVMe SSDs in RAID 10)
Redundant connection: 2 x 25 Gb/s

Points of presence: FR (RBX, GRA, SBG), DE (LIM).

HDS healthcare data hosting certification.

In short, we are closing the ERI site (UK) and replacing it with Strasbourg, as Brexit has created too many certification issues for some European clients.

HDS (healthcare certification) is becoming the default for our infrastructure.

We are keeping the same hardware policy: no compromise on performance. The new AMD EPYC 9354 processors deliver better performance in every area, along with memory bandwidth increasing from 140.8 to 460.8 GB/s.

On the network side, our standard redundant connections are increasing from 10 to 25 Gb/s.

Finally, on the storage side, the standard is increasing from 2 RAID arrays to 3.

The current 2 separate arrays provide clients with ultra-high-performance, readily available local backup space without requiring network services.

The 3rd array strengthens the overall system by serving as a spare array, enabling a more efficient, safer, and faster switch in the event of a hardware incident:

1) Local migration to the spare array to secure the data immediately.
2) Migration of the VMs to another node once the data is secure
3) Hardware maintenance on the original node
4) Tests and validation
5) Return to the original node.

Not a client yet? Discover our managed Linux, PHP (5.6 to 8.3), MariaDB / MySQL, and Node.js solution starting at €145 excl. VAT/month: StackX

ScalarCloud / PBS (Proxmox Backup Server) Backup Infrastructure Upgrade

Published on October 25, 2024 in Cloud

Hello everyone, we are currently deploying new backup infrastructure for our hosting clients (ScalarCloud).

To address overlapping backup jobs, which can sometimes occur, particularly during the nighttime slowdowns on the OVH network currently in use, the new infrastructure will no longer be shared.

From now on, each node will have its own independent PBS (Proxmox Backup Server), along with a dedicated 25 Gb/s interconnection.

Have an excellent weekend, everyone!

Not a client yet? Discover our managed Linux, PHP (5.6 to 8.3), MariaDB / MySQL, and Node.js solution starting at €145 excl. VAT/month: StackX

Hypervisor Update

Published on December 23, 2023 in Cloud

This is the final year-end update for all Proxmox VE hypervisors supporting the Public Cloud infrastructure of ScalarCloud.

Based on the tests we have conducted for more than a month, we will keep the Proxmox VE 7.4 branch on the Public Cloud production clusters for now.

We will schedule an update once all our tests are 100% conclusive. (In any event, the version has no client-side impact because this layer is not exposed.)

No service interruption is expected. The machines are live-migrated from node to node as the updates are carried out, then returned to their original location.

Increasing ScalarCloud Local NVMe SSD RAID Storage Capacity

Published on September 20, 2023 in Cloud

Hello everyone, to meet the demand for space, which has accelerated in recent months, we will soon increase the NVMe SSD storage capacity of our ScalarCloud platform.

These capacity increases will be carried out in stages, one node at a time. No service interruption is expected.

We will follow the same process used for hypervisor updates.

Your machines will be live-migrated and distributed across other standby nodes for the duration of the operation, then returned to their original location.

We will then perform load balancing through redistribution: some virtual machines will keep their virtual NVMe storage on the first volume, while others will be moved live to the second volume.

These operations will begin on October 3 and aim to double our current RAID 10 storage capacity by adding a new array, taking us from 24 TB of NVMe storage per node to 48 TB.

We do not yet know when these operations will be completed, as the hardware upgrades require coordination with OVHcloud, one node at a time. We will let you know once the work is complete.

Have an excellent rest of your weekend.

Category: Cloud