Exoscale vs Leafcloud Object Storage

High-Performance Compute Instances

Exoscale's compute instances use SSD local storage with options for standard, CPU-optimized, memory-optimized and storage-optimized flavors. Sizes range from Micro (512MB RAM, 1 core) to Jumbo (225GB RAM, 24 cores). GPU instances with NVIDIA cards (A30, V100, A40, A5000, RTX 6000 Pro) suit AI tasks. Anti-affinity ensures instances spread across hosts for resilience. Snapshots and custom templates aid backups and deployments. Billing prorates to the second, with powered-off instances charged only for storage.

Scalable Kubernetes Service (SKS)

Managed Kubernetes deploys clusters in 120 seconds. Starter clusters are free (no SLA), Pro offers 99.95% uptime. Users scale worker nodes, manage control planes via portal, CLI, API or Terraform. Integrates with network load balancers. Supports cloud-native apps across zones.

Managed DBaaS and Storage

DBaaS covers PostgreSQL, MySQL, Kafka, OpenSearch, Valkey and Grafana, with plans from Hobbyist (2GB RAM) to Premium (225GB RAM, multiple replicas). Object Storage is S3-compatible with multi-redundancy and replication. NVMe Block Storage attaches multiple volumes per instance for low-latency needs. Traffic: inbound free, intra-zone free, 1TB outbound free per instance monthly.

S3 API at leafcloud.store (europe-nl-ams1)

Path-style S3 endpoint https://leafcloud.store, region europe-nl-ams1. Works with AWS CLI, boto3, rclone, Cyberduck, MinIO mc, and other S3 SDKs. Documented features include public or private containers, bucket policies and ACLs, multipart uploads, and presigned URLs. Max object size is 5 TB via multipart. Authentication uses OpenStack EC2 credentials (openstack ec2 credentials create), not the dashboard password.

OpenStack Swift and Horizon dashboard

The same store is reachable as OpenStack object storage (Swift). Create and browse containers at create.leaf.cloud under Object Store. Native CLI uses project-scoped OpenStack auth (openstack container create / object create). Container names must be unique across all Leafcloud users. Docs say there is a limit on how many containers you can create (the exact quota is not published).

Ceph cluster with 3x replication in Amsterdam

The product page describes a Ceph backend (Apache 2.0 software) with triple replication across separate physical nodes in Amsterdam. Persistent objects sit at the Core facility. Compute Leaf sites are a different path and are not where object data is stored, according to the security pages. This is a single-region store, not a multi-region S3 deployment.

SSE-C plus TLS in transit

Official docs show S3 server-side encryption with customer-provided keys (SSE-C, AES256). Leafcloud uses the key on each request and does not store it. Lose the key and you cannot read the object. The product table also lists encryption at rest and TLS in transit as supported. DPA language is TLS 1.2+. LUKS-by-default messaging on the security pages refers to block volumes, not this object API.

Terraform state, Velero, and Nextcloud recipes

Leafcloud publishes working recipes for Terraform’s S3 backend (path-style, skip checksum/region checks, endpoint leafcloud.store), Velero Kubernetes backups (s3ForcePathStyle plus s3Url), and Nextcloud primary objectstore pointing at leafcloud.store:443. Useful if you already run those tools. Object versioning is not available, so state and backup designs must version keys themselves or copy out.