---

# Online history

1999 Online is created to sell dedicated servers

Online is mostly known for:

- Dedibox (Baremetal dedicated servers)
- Mutualized hosting
- [Domain name registrar](https://en.wikipedia.org/wiki/Domain_name_registrar)

Money-wise:

- Subscription
- Companies spend as Capital Expenditure (CAPEX)

User-wise:

- Mostly French

---

# Pivot to cloud

---

---

# Scaleway history

2015 Scaleway team is created to offer cloud services

Money:

- On demand usage (Billed by hour)
- Operational Expenditure (OPEX)

User-wise:

- Mostly Foreign

???
Scaleway's goal is to provide a "create server" button: on click, after a few
seconds, users can connect to the server and host applications.
It is possible to move an IP address from a server to another, make snapshots
from volumes, create images to start several servers from the same installation,
set security groups, ...
Servers are billed per hour, so infrastructures can scale up and down depending
on users' needs.

2017 Scaleway brand, la Maison, Go Global

???
On September 2017, we realized it made more sense to gather Online and Scaleway
services behind the brand Scaleway, and that we needed to recruit engineers to
extend our teams, develop our products and go worldwide. We moved from Iliad
headquarters (rue de la Ville l'Évêque, Paris 75008) to "La Maison" in Paris,
and we opened another "Maison" in Lille.

---

---

# Offers

- virtualized vs baremetal
- ARM vs x86
- Start vs Pro

https://www.scaleway.com/pricing/

---

# Offers
## 2015: C1

Pimouss hardware

* 4 cores
* 2Gib RAM
* 18 C1 servers on a blade
* Servers don't have hard drives directly connected. They are connected to a
  NAS, which is a C1 server behind the blade
* 16 blades in a chassis
* two uplinks for 16 blades: one 10Gbit/s for compute traffic, and one 1Gbit/s
for admin traffic

???
In April 2015, [we announce C1 instances](https://blog.online.net/2015/04/02/from-online-labs-to-scaleway/).
C1 is the acronym for "Compute 1st (generation)".

C1 are armv7 servers, internally named "Pimouss" (we have a candy name for each
type of hardware developed by our lab), designed by our hardware lab.

---

# Offers
## 2016: C2

Home made Suchard Hardware

There are three variants of C2 instances:

* C2S: 4 x86-64 cores, 8GiB RAM, one network interface at 2.5Gbits/s
* C2M: 8 x86-64 cores, 16GiB RAM, two network interfaces at 2.5Gbits/s
* C2L: 8 x86-64 cores, 32GiB RAM, two network interfaces at 2.5Gbits/s + one
 SATA disk of 250Gb.  This disk can be used to store anything but it is zeroed
 by a worker when the server is stopped, so it should be considered as a
 temporary disk.

Connected to a (PowerEdge R730xd from Dell). There is one NAS for 6 chassis.

A suchard chassis contains 18 nodes, 9 C2S, 5 C2M and 4 C2L.

???
In March 2016, [we announce C2 instances](https://blog.online.net/2016/03/08/c2-insanely-affordable-x64-servers/).
C2 is the acronym for "Compute 2nd generation".

---

# Offers
## 2016: VPS

Qemu-kvm based

* VC1S: 2 x86-64 cores, 2GB RAM and 50GB SSD
* VC1M: 4 x86-64 cores, 4GB RAM and 100GB SSD
* VC1L  6 x86-64 cores, 8GB RAM and 200GB SSD

The host instance is actually a C2L node with a SATA disk of 1To where are
stored VMs volumes.

???
In April 2016, we announce [VC1 instances](https://blog.online.net/2016/04/05/vc1-new-affordable-variants/).
VC1 is the acronym for "Virtual Cloud 1st generation".

The virtual offer is "over provisioned", we put more client that the maximum
capacity of the host.

As we don't overprovision RAM, the maximum number of client is 60 or 28 on a HV
(suchard or tagada)

---

# Offers
## 2017: Workfload intensive

* X64-60GB (internally called VC2L) 10 cores, 60GB of RAM and 700GB of disk
* X64-120GB (internally called VC2XL) 12 cores, 120GB of RAM and 1To of disk

We realize it becomes difficult to understand all our offers, so we decide to
rename the previous VPS offers:

* VC1S are now also known as X64-2GB
* VC1M are now also known as X64-4GB
* VC1L are now also known as X64-8GB

???
In February 2017, we announce [workload intensive instances](https://blog.online.net/2017/02/07/scaleway-new-cloud-servers-for-intensive-workloads/)

---

# Offers
## 2017: New variants for workload intensive instances

* X64-15GB (internally named VC2S) 6 cores, 15GB of RAM and 200GB of disk
* X64-30GB (internally named VC2M) 8 cores, 30GB of RAM and 400GB of disk

The host machine is a Supermicro server with 12 cores, 128GB of RAM and 2 SSD
disks of 2To

???
In February 2017, we announce [new variants for workload intensive instances](https://blog.online.net/2017/02/23/workload-intensive-cloud-servers-new-variants/)

---

# Offers
## 2017: ARMv8

* ARM64-16GB: 16 cores, 16GB of RAM and 400GB of SSD disk
* ARM64-32GB: 32 cores, 32GB of RAM and 600GB of SSD disk
* ARM64-64GB: 48 cores, 64GB of RAM and 800GB of SSD disk
* ARM64-128GB: 64 cores, 128GB of RAM and 1To of SSD disk

It is running on [Cavium servers](https://www.cavium.com/product-thunderx-arm-processors.html),
which have 96 cores, 256GB of RAM and 3 SATA disks of 1.8To and 1 SATA disk of
1To.

???
In June 2018, we announce [ARMv8 instances](https://blog.online.net/2017/06/22/armv8-scaleway-new-high-core-variants/)

---

# Offers
##  2018: Start family with NVMe disks

* START1-XS (internally named VC3XS) 1 core, 1GB of RAM and 25GB of disk
* START1-S (internally named VC3S) 2 core, 2GB of RAM and 50GB of disk
* START1-M (internally named VC3M) 4 core, 4GB of RAM and 100GB of disk
* START1-L (internally named VC3L) 8 core, 8GB of RAM and 200GB of disk

These instances are run on "tagada" nodes. Tagada is the new server developed by
our hardware lab, with 16 cores, 64GB of RAM and a NVMe disk of 1To.

Tagada nodes are more powerful than C2L instances (where are run VC1
instances) with 64 GB or RAM.
New VC1 instances are now provisioned on tagada nodes, and we plan to stop using
C2L instances for existing servers.

???
In May 2018, we announce [new instances](https://blog.online.net/2018/05/03/introducing-scaleway-nextgen-nvme-cloud-servers-with-hot-snapshots/)

---