Usage of this storage solution type allows to nearly unlimited scalability of cluster size. The alternative way is to use disk arrays – however their scalability is limited for cloud computing purposes. In the case of object storage, adding another server to the cluster increases all of its parameters: capacity, performance and network throughput.
We use replication to ensure safety of stored data – each data byte is stored in several copies. We are using triple replication for mechanical disk based storage tiers and double replication for SSD disks. In order to exclude simultaneous loss of more than one copy of data, a special software monitors the cluster operation to ensure that one disk or server stores no more than a single copy.
Data replication does not however completely exclude the possibility of damaging the stored data copies. To prevent such situation, the cluster is constantly monitored to verify the correctness of recorded data (scrubbing) and if incorrect data is detected, it is restored from other copies. In case of a loss of one of the servers, the cluster restores missing data copies on other servers. Effective safety level for both disk types is similar, because due to substantially higher performance of SSD disks the restoration of the original safety level is achieved a lot quicker.
- Virtual machines management
- Creating and removing instances
- Start / stop / soft & hard reset
- Instances resize
- Ports management
- Floating IPs management
- Console (KVM)
- Available operating systems
- Debian 7, CentOS 6/7, Ubuntu 14.04 LTS, Fedora Cloud, MS Windows 2012, VyOS, BSDRP
- Storage management
- Creating and removing volumes
- Creating and removing snapshots
- Creating and removing images
- Public IP addresses management (floating IPs)
- Maximum single VM size: 6 vCore + 32 GB RAM
- Tier III+ data center located in Warsaw, Poland
- Direct connections to Tier 1 operators
- Direct connections to traffic exchanges in Poland and Europe
- Peering with main Polish ISPs (UPC, Orange)
- HA for compute - In case of detection of a compute server failure, we verify the cause to avoid false positives. When failure is confirmed, virtual machines from the damaged server are restarted on other operational server from the cluster.
- HA for storage - The use of object storage with a few times replication ensures the continuous access to data also in case of storage server failure. In such a case the data is served from another copy.