Cloud Deployment Options
Overview
To account for a variety of customer requirements, we offer multiple deployment variants.
We use a combination of the HPE blade server platform (for servers and partly for storages), HPE proliant (for large 4 socket servers), Supermicro Servers for Cluster Storage, Mellanox and Arista switches.
-
All units provide separate ILO ports and management functionality
-
All units provide at least 2 times redunant power supply
-
We monitor all hardware in real-time and manage it through a separate ILO network connected via two distinct VPN channels to our operations center
-
We service and manage converged hypervisors (combined compute and network nodes), storage nodes and the cloud controllers through a separate management network connected via two distinct VPN channels to our operations center
-
Cloud controllers are deployed three-times redundant
-
We perform real-time monitoring of cloud controllers and all required house keeping tasks; Regular Cloud controller backups are kept offsite at our operations center
-
Tenant Virtual Networks use distributed virtual routing; For performance reasons, they are deployed through VLAN instead of VXLAN/GRE
Basic Single Site Deployment
This scenario fits for needs that require medium performance, limited scaleability and standard availability.
It is based on the HPE blade server platform and combined Compute, Network and Storage Nodes, using up to 2 x 18 Core Intel XEON E5 CPUs per node.
Storages Blades in the Chassis are attached via SAS3 RAID controllers and are used to supply the image repository and virtual disks as well as virtual file shares. They may be configured with either 12 x 1.2 TB 10K HDDs or 12 x 1960 GB SAS3 SSDs.
The total capacity per enclosure is thus up to 576 physical CPU cores (1.152 CPU threads) or 376.320 GB of SSD Storage or any combination of it.
A cloud can be deployed across multiple (many) enclosures, but storage cannot be shared across enclosures and thus the images, instances, virtual disks and file shares deployed to hypervisors within one enclosure are local to this enclosure (limited scaleability).
For backups, a separate nearline storage cluster can be optionally deployed.
Advanced Single Site Deployment
In this scenario, the HPE blade platform is used for compute resources and network virtualization, storages are deployed separately and are shared across enclosures. This improves cross-enclosure scaleability for larger environments.
Storages are deployed via our storage solutions and provide nearly unlimited scaleability.
For backups, a separate nearline storage cluster may be optionally deployed either on the primary or a secondary site or both. Storage traffic (virtual machines, virtual disks, virtual file shares) flows over SAN or GbE switch fabrics between the blade systems and the storage controllers / clusters.
In addition, large 4 socket HP proliant rack servers can be deployed to meet the demand for large virtual machines with up to 80 physical CPU cores.
Tenant traffic from and to virtual machines is switched over GbE switch fabrics to/from the core routing capabilities of the customer or directly to receiving L2 switches, if deployed in the same subnet.
Cloud Management traffic (tenants) also goes through other IT infrastructure of the customer, including VPN, or directly to/from the Internet.
Cloud and other management traffic for cloud system administration and ILO traffic (hypervisors, storage nodes, network switches) goes through the switching fabric and VPN routing is routed to / from our operations center.
Primary and Secondary Site Deployment
n this scenario, we are able to run ciritical virtualized resources critical, on both sites.
Cloud controllers are initially set up on both sites with the mirrored versions and configurations. VLANs and Routing Tables are configured as mirrors of each other.
A replication mechanism for the controller database is set up so that all cloud management configurations are also applied to the controller databases on the secondary site.
Further on, backups are usually either directly performed in a cluster at the secondary site or they are replicated between two backup clusters. File shares may be replicated across sites either.
Replication bandwidth is usually very limited and only the really critical data is selected.
In case of SAN storage, volumes, which contain the data of critical virtual machines and virtual disks, are configured for cross-site SAN to SAN replication.
HPE BL2500 Storage Blade
HPE C7000 Enclosure (16 Blades)