Volumes in a Kafka cluster

In a Kafka Managed Database cluster, you can use local and network volumes. The type of disk used depends on the selected configuration line.

You can select a configuration line with the required disk type and disk size when creating a Managed Database cluster. After the cluster is created, you can change the line to another one with the same disk type and increase the disk size — scale the cluster.

Disk performance and throughput for different configuration lines can be viewed in the Performance and throughput table.

Local disks

Configuration lines	Standard; CPU; Memory; HighFreq; Dedicated; Flex
Disk type	SSD NVMe disk
Technical specifications	The disk subsystem is configured in RAID 1 for the Dedicated lineup and in RAID 10 for the Standard, CPU, Memory, HighFreq, and Flex lineups
Advantages	Low data access latency
Usage	Suitable for tasks sensitive to read and write performance
Number of disks in the DBMS	1
Changing configuration	in all configuration lines with a local disk, except for Flex, you cannot increase the local disk size alone; you must change the entire node configuration — scale the cluster; a configuration line with a local disk can only be changed to a line with a network volume during cluster recovery
Detaching a disk	✗
Connecting to another database	✗
Creating another disk, snapshot, or backup	✗
Deletion	Deleted along with all data when the cluster is deleted

Network volumes

Configuration lines	Flex
Disk type	SSD NVMe disk
Technical specifications	Managed by a software-defined storage (SDS) system — Ceph
Advantages	triple disk volume replication; network volume size is independent of the number of vCPUs
Usage	Not recommended for systems sensitive to network latency and response time. For example, suitable for data archiving
Number of disks in the DBMS	1
Changing configuration	no downtime when changing disk size; when changing the number of vCPUs and RAM, downtime depends on the number of nodes in the cluster: in single-node clusters, downtime is equal to the node reboot time (one to two minutes), and in multi-node clusters — the time taken to switch to another node (less than one minute); a configuration line with a network volume can only be changed to a line with a local disk during cluster recovery
Detaching a disk	✗
Connecting to another database	✗
Creating another disk, snapshot, or backup	✗
Deletion	Deleted along with all data when the cluster is deleted

Performance and throughput

	Volumes up to 256 GB	Volumes 512 GB	Volumes 768 GB	Volumes 1 TB	Volumes 1.5 TB	Volumes 2 TB
Throughput	400 MB/s	500 MB/s	600 MB/s	700 MB/s	800 MB/s	900 MB/s
Performance (read)	25600 IOPS	40000 IOPS	50000 IOPS	60000 IOPS	70000 IOPS	80000 IOPS
Performance (write)	12800 IOPS	20000 IOPS	25000 IOPS	30000 IOPS	35000 IOPS	40000 IOPS

	Micro, Extra Small	Small, Medium, Large, Extra Large, Mega
Throughput	7000 MB/s	7000 MB/s
Performance (read)	700000 IOPS	1500000 IOPS
Performance (write)	180000 IOPS	750000 IOPS

	Local disk	Network volume
Throughput	300 MB/s	500 MB/s
Performance (read)	12800 IOPS	25000 IOPS
Performance (write)	6400 IOPS	15000 IOPS