Versions and configurations
Versions
Kafka version 3.5 is supported.
Node configurations
At creating a Kafka cloud database cluster you can select the node configuration — number of vCPUs, RAM and disk size.
The available node configurations are summarized in rulers.
When selecting a configuration, consider that about 5 GB of disk space in all configurations is reserved for the operating system, service components, and log storage. The rest of the disk space is available for databases.
Once the cluster is created, you can change the configuration of the nodes — cluster scale.
Configuration lines
The available node configurations are summarized in rulers. The selected line of configurations depends on the configurations used processors и disks.
The ratio of resources in line configurations can be:
To see the availability of the configuration lines in the regions, see the availability matrix Cloud databases.
Standard
A line of fixed configurations with local disk and balanced vCPU:RAM ratio, suitable for most DBMSs. We recommend using this lineup if you do not know the load profile.
Intel® Xeon® Scalable or AMD EPYC™ 2.2-2.4 GHz processors are used. To see the processor frequency in different configurations, see the table below Processors.
To see the performance and throughput of disks in different lines, see the table below Performance and throughput.
CPU
A line of fixed configurations with local disk and balanced vCPU:RAM ratio. Suitable for workload profiles that are computationally demanding. For example, if the database is running analytic queries, multiple nested queries, or data encryption. One way to determine such a load profile is to track metrics Load Averagewhich shows the average system utilization over one, five or 15 minutes.
Intel® Xeon® Scalable or AMD EPYC™ 2.2-2.4 GHz processors are used. To see the processor frequency in different configurations, see the table below Processors.
To see the performance and throughput of disks in different lines, see the table below Performance and throughput.
Memory
A line of fixed configurations with local disk and balanced vCPU:RAM ratio. Suitable for workload profiles that are cache-intensive. For example, if the database runs multiple, infrequent queries on different parts of tables. One way to determine such a load profile is to monitor the metric Попадание в кэш (Cash_hit_ratio)which shows the percentage of data in the query that is read from the cache.
Intel® Xeon® Scalable or AMD EPYC™ 2.2-2.4 GHz processors are used. To see the processor frequency in different configurations, see the table below Processors.
To see the performance and throughput of disks in different lines, see the table below Performance and throughput.
HighFreq
A line of fixed configurations with a balanced vCPU:RAM ratio.
High-performance Enterprise-level equipment is used:
- Intel® Xeon® Gold 6354 processors with Turbo Boost frequency up to 3.6 GHz. To see the processor frequency in different configurations, see the table below Processors;
- RAM ECC Reg 3.2 GHz;
- SSD NVMe disks with higher performance. See the table below to see the performance and throughput of the drives in the different product lines Performance and throughput.
Dedicated
A line of fixed configurations with cluster nodes on separate cloud servers. Each cloud server occupies the entire dedicated host (physical server). Suitable for users who require physical isolation of databases from other clients, maximum performance and maximum size of available resources.
High-performance Enterprise-level equipment is used:
- one Intel® Xeon® Gold 6240 processor with Turbo Boost speed up to 3.9 GHz. To see the processor frequency in different configurations, please refer to the table below Processors;
- RAM 64 GB DDR4 ECC Reg;
- two SSD NVMe disks in RAID 1. To see the performance and throughput of the disks in different ranges, see the table below Performance and throughput;
- two 2 × 25 GE NICs for the main network + MC-LAG with 25 Gbps connection speed for the service network (for backup, monitoring, data replication in the cluster).
* To improve the performance of the DBMS cluster, Hyper-Threading Technology is used. This technology allows you to use 34 vCPUs based on the physical 18 CPUs. This performance is suitable for highly loaded systems or analytical load profile.
** One slate of RAM is reserved for the services that maintain the physical server.
*** To provide additional fault tolerance, the disks are placed in RAID 1. RAID 1 uses mirroring, so 50% of the disk space is available for the database. Part of the disk space is also reserved for services that maintain the physical server.
Flex
A lineup of arbitrary configurations with local disk, in which you can choose the resource ratio.
To see the performance and throughput of the drives in the different lines, see the table below Performance and throughput.
Intel® Xeon® Scalable or AMD EPYC™ 2.2-2.4 GHz processors are used. To see the processor frequency in different configurations, see the table below Processors.
Configuration limits depend on pool. To see the availability of configurations in the regions, see the availability matrix Cloud databases.
* If there are more than 8 vCPUs in a configuration, the vCPU:RAM ratio must be at least 1:4. For example, 10 vCPUs require at least 40 GB of RAM.
** If the configuration has more than 8 vCPUs, the vCPU:Local Disk ratio must be at least 1:32. For example, 10 vCPUs require a disk size of at least 320 GB.
If the configurations are not suitable, you can order your own configuration. Create a ticket and indicate the ratio of resources:
- vCPU:RAM — at least 1:4;
- vCPU:RAM:Local disk — at least 1:2:16.
SDS
A line of random configurations with a network disk in which you can select the resource ratio. The network disk is managed by Software-Defined Storage (SDS) — Ceph. Ceph provides three-fold replication of disk volumes.
Network latency with network disks is higher than with local disks, so it is not recommended to use the SDS line for systems sensitive to network latency and latency. Suitable for archival storage.
When changing disk size in SDS line configurations, there is no down time regardless of the number of nodes in the cluster. When changing the number of vCPUs and RAM, the down time depends on the number of nodes in the cluster: in single-node clusters, the down time is equal to the time to reboot a node (one to two minutes), and in multi-node clusters, the down time is equal to the time to switch to another node (less than one minute).
To see the performance and throughput of the drives in the different lines, see the table below Performance and throughput.
Intel® Xeon® Scalable or AMD EPYC™ 2.2-2.4 GHz processors are used. To see the processor frequency in different configuration lines, see the table below Processors.
Configuration limits depend on pool. To see the availability of configurations in the regions, see the availability matrix Cloud databases.
Processors
Configuration lines differ in the available processors. The processor frequency affects the speed of processing user requests, executing complex algorithms and data operations.
To see the availability of configuration lines in the regions, see the availability matrix Cloud databases.
* With a cloud server load of 100%, the processor is running with Turbo Boost technology and a maximum frequency of 3.6 GHz for the HighFreq lineup and 3.9 GHz for the Dedicated lineup. Since the processor is emulated, the test will display 3.00 GHz for the HighFreq lineup and 2.6 GHz for the Dedicated lineup.