CROC offers virtual machines (VMs) with flexible scalability (up to 384 GB RAM) and hosting in the cloud based on two distributed data centers. VMs are highly available by default and are re-launched on free platform servers in the case of equipment failures. Moreover, CROC also offers VMs with preinstalled system and application Microsoft, Cisco, SAP HANA, Red Hat, Citrix, and Oracle software provided under respective vendor subscription programs.
CROC Cloud offers customers a choice of two virtual disk types: Universal and Flash, each having its own features and use cases.
Universal type is good for boot disks or apps having no strict demands for disk subsystem performance and latency. Large-size disks feature good linear R/W performance and therefore can be used for log processing, Big Data, and streaming.
- Based on enterprise-class SAS magnetic disks
- Maximum IOPS: 500 (R+W)
- Maximum throughput (MB/s) depends on a disk size and is calculated as follows:
Max MB/s = Size (GB) * 0.25 (MB/s per GB), but no less than 8 MB/s
When disk size increases, its maximum throughput increases automatically.
- Disk size must be at least 32 GB and must be a factor of 8 GB
Flash type guarantees high performance and low latency and is a must for resource-intensive databases and web apps.
- Based on all-flash storages capable of supporting up to 1,000,000 IOPS
- Guaranteed virtual disk performance up to 100,000 IOPS
- IOPS value can be changed online
- Performance does not depend on disk size
Common features of a data storage subsystem:
Data is transmitted via a high-speed 56 Gbps InfiniBand fabric. All virtual disks support:
- Online size change (expansion only)
- Online connection of additional disks to VM
- Online disconnection of disks (except for a boot one)
- Simultaneous connection of different type disks (up to 16 disks in total) to a single VM
- OS loading from any of the connected disks
A backbone of our Cloud is Virtual Private Cloud, or VPC, which is similar to Virtual Routing and Forwarding (VRF) on a conventional networking equipment. VPC ensures network isolation meaning that private IPs in different VPCs are not connected, and also provides various network services as follows.
All subnets in one VPC always have IP connectivity. Using subnets allows breaking VPC’s IP space into multiple L3 segments. You can also create subnets and specify an Availability Zone (AZ) where a subnet is to be located. In such a way, you can allocate services among different AZs to mitigate the impact of incidents in any AZ on the others.
- Elastic IP addresses
You can allocate a random address from a CROC Cloud’s public IP space and assign it to any instance. If necessary, you can reassign this address to any other instance, irrespective of its location in subnets or AZ. Such address associations are enabled by Static NAT technology.
To connect to a remote site or third-party public cloud, you can configure a VPN using IPSec. In this case, you do not need to configure or maintain VPN on instances in the Cloud on your own.
- External networks
You can colocate your equipment in any of CROC’s data centers or lease a third-party communication channel and connect it to a virtual infrastructure in the Cloud. To this end, CROC engineers will connect your hardware to Cloud’s resources. Subsequently, you will be able to independently connect this external network to any Cloud subnet either using a management console, or via API. We support up to 10 Gbps connections.
- Network ACLs (Access Control Lists)
To manage inter-subnet and external access, you can create egress/ingress rules and specify source/destination IPs, protocol, port, and priority. Network ACLs run in a stateless mode: return traffic must be explicitly permitted.
- Security Groups
It is yet another type of a Cloud native firewall. You can create logical containers, to which IP-, protocol-, and port-based permitting rules are added. If there are no rules in a Security Group then all traffic is prohibited. These rules run in a stateful mode: whether you have permitting rules for return traffic or not, it will be permitted. In addition to IP, you can specify other Security Groups in Security Group rules.
Other network services
Each instance in the Cloud automatically receives network configuration via DHCP. You can manage some parameters, such as dns-servers, ntp-servers, domain-name, etc.
Each VPC has a built-in DNS resolver which resolves internal names of instances as well as all public names.
- Route Tables
You can manage traffic routing to external networks, VPNs or networks behind instances, via a VPC routing table. You can also create different routing tables and associate them with different subnets, thus implementing source-based routing.
- Virtual switches
If you want to connect instances in different VPCs or add a network interface to instances, you can create a virtual switch and hot-plug it to instances. It is an L2 network, which has no additional services inside, except for network connectivity between instances in it.
All network entities and services can be managed via API. Virtual switches and external networks can be managed using our fork of python-boto (https://github.com/c2devel/boto), while other entities are compatible with Amazon VPC, thus enabling management through aws cli, terraform, ansible and other AWS compatible automation tools.
Clear and predictable payment for consumed cloud resources, using the system certified by the Russian Ministry of Communications and Mass Media.
CROC’s cloud billing features:
- No minimum requirements regarding bookings for computing resources at the beginning of the reporting period;
- No advance payments;
- Work paid in rubles;
- Servers, disks, and file storage space are accounted for and billed on a per-hour basis;
- Billing is based on the selected configuration of a virtual server (without considering current utilization profile) and the template (image) used for the server launch;
- If a server was created from CROC’s template that included licensed software (Microsoft Windows, Red Hat) then billing will be at a higher rate that includes royalties payable to the respective software vendors;
- There are three modes of server billing based on server status:
- If server is turned on then charges are based on the selected server configuration, disk space, network traffic, and external IP address (if assigned);
- If server is turned off then charges are based on the used disk space only;
- If server is removed then system disk is also removed automatically. In this case, charges will be only based on all additional disks until they are removed manually.