A YugabyteDB cluster consists of two distributed services - the YB-TServer service and the YB-Master service. Since the YB-Master service serves the role of the cluster metadata manager, it should be brought up first followed by the YB-TServer service. In order to bring up these distributed services, the respective servers (YB-Master or YB-TServer) need to be started across different nodes. Below are some considerations and recommendations on starting these services. The deployment configurations section below has detailed steps on how to setup YugabyteDB clusters.
- YugabyteDB works on a variety of operating systems. For production workloads, the recommended operating systems are CentOS 7.x and RHEL 7.x.
- Set the appropriate system limits using
ulimiton each node running a YugabyteDB server.
- Use ntp to synchronize time among the machines.
YugabyteDB internally replicates data in a strongly consistent manner using Raft consensus protocol in order to survive node failure without compromising data correctness. The number of copies of the data represents the replication factor. Note that this distributed consensus replication is applied at a per-shard (aka tablet) level similar to Google Spanner.
You would first need to choose a Replication Factor (RF). You would need at least as many machines as the RF, which means 1 machine for RF1, 3 machines for RF3 and so on. Below are some recommendations relating to the RF.
- The RF should be an odd number to ensure majority consensus can be established during failures.
- The default replication factor is
- RF of
- RF of
- More generally, if RF is
n, YugabyteDB can survive
(n - 1) / 2failures without compromising correctness or availability of data.
- RF of
- Number of YB-Master servers running in a cluster should match RF. Run each server on a separate machine to prevent losing availability on failures. You have to also specify the RF using the
--replication_factorwhen bringing up the YB-Master servers.
- Number of YB-TServer servers running in the cluster should not be less than the replication factor. Run each server on a separate machine to prevent losing availability on failures.
Note that YugabyteDB works with both hostnames or IP addresses. IP addresses are preferred at this point as they are more extensively tested.
See the yb-master command reference for more information.
YugabyteDB is designed to run well on bare-metal machines, virtual machines (VMs), and containers.
CPU and RAM
You should allocate adequate CPU and RAM. YugabyteDB has good defaults for running on a wide range of machines, and has been tested from 2 core to 64 core machines, and up to 200GB RAM.
- 2 cores
- 2GB RAM
- 16+ cores
- 32GB+ RAM
- Add more CPU (compared to adding more RAM) to improve performance.
For typical OLTP workloads, YugabyteDB performance improves with more aggregate CPU in the cluster. You can achieve this by using larger nodes or adding more nodes to a cluster. Note that if you do not have enough CPUs, this will show up as higher latencies and eventually dropped requests.
Memory depends on your application query pattern. Writes require memory but only up to a certain point (say 4GB, but if you have a write-heavy workload you may need a little more). Beyond that, more memory generally helps improve the read throughput and latencies by caching data in the internal cache. If you do not have enough memory to fit the read working set, then you will typically experience higher read latencies because data has to be read from disk. Having a faster disk could help in some of these cases.
YugabyteDB explicitly manages a block cache, and doesn't need the entire data set to fit in memory. It does not rely on OS to keep data in its buffers. If you give YugabyteDB sufficient memory data accessed and present in block cache will stay in memory.
- SSDs (solid state disks) are required.
- Both local or remote attached storage work with YugabyteDB. Since YugabyteDB internally replicates data for fault tolerance, remote attached storage which does its own additional replication is not a requirement. Local disks often offer better performance at a lower cost.
- Multi-disk nodes
- Do not use RAID across multiple disks. YugabyteDB can natively handle multi-disk nodes (JBOD).
- Create a data directory on each of the data disks and specify a comma separated list of those directories to the yb-master and yb-tserver servers via the
- Mount settings
- XFS is the recommended filesystem.
- Use the
noatimesetting when mounting the data drives.
- ZFS isn't currently supported and is in the roadmap .
- NFS isn't currently supported and is in the roadmap .
YugabyteDB does not require any form of RAID, but runs optimally on a JBOD (just a bunch of disks) setup. It can also leverage multiple disks per node and has been tested beyond 10 TB of storage per node.
Write-heavy applications usually require more disk IOPS (especially if the size of each record is larger), therefore in this case the total IOPS that a disk can support matters. On the read side, if the data does not fit into the cache and data needs to be read from the disk in order to satisfy queries, the disk performance (latency and IOPS) will start to matter.
YugabyteDB uses per-tablet size tiered compaction . Therefore the typical space amplification in YugabyteDB tends to be in the 10-20% range.
Below is a minimal list of default ports (along with the network access required) in order to use YugabyteDB.
Each of the nodes in the YugabyteDB cluster must be able to communicate with each other using TCP/IP on the following ports.
7100 for YB-Master RPC communication 9100 for YB-TServer RPC communication
In order to view the cluster dashboard, you need to be able to navigate to the following ports on the nodes.
7000 for viewing the YB-Master Admin UI
To use the database from the app, the following ports need to be accessible from the app (or CLIs).
9042 for YSQL 9042 for YCQL 6379 for YEDIS
Default ports reference
The above deployment uses the various default ports listed here .
NoteFor Yugabyte Platform, the SSH port is changed for added security.
For YugabyteDB to preserve data consistency, the clock drift and clock skew across different nodes must be bounded. This can be achieved by running clock synchronization software, such as NTP . Below are some recommendations on how to configure clock synchronization.
Set a safe value for the maximum clock skew parameter (
--max_clock_skew_usec) when starting the YugabyteDB servers. We recommend setting this parameter to twice the expected maximum clock skew between any two nodes in your deployment.
For example, if the maximum clock skew across nodes is expected to be no more than 250ms, then set the parameter to 500ms (
The maximum clock drift on any node should be bounded to no more than 500 PPM (or parts per million). This means that the clock on any node should drift by no more than 0.5ms per second. Note that 0.5ms per second is the standard assumption of clock drift in Linux.
NoteIn practice, the clock drift would have to be orders of magnitude higher in order to cause correctness issues.
Running on public clouds
Amazon Web Services (AWS)
- Use the
- Recommended types are
- For the
c5instance family, use
gp2EBS (SSD) disks that are at least 250GB in size, larger if more IOPS are needed. - The number of IOPS are proportional to the size of the disk. - In our testing,
gp2EBS SSDs provide the best performance for a given cost among the various EBS disk options.
- Avoid running on
t2instance types . The
t2instance types are burstable instance types. Their baseline performance and ability to burst are governed by CPU Credits, and makes it hard to get steady performance.
- Use the
n1-highcpuinstance family. As a second choice,
n1-standardinstance family works too.
- Recommended instance types are
- Local SSDs are the preferred storage option. - Each local SSD is 375 GB in size, but you can attach up to eight local SSD devices for 3 TB of total local SSD storage space per instance.
- As a second choice, remote persistent SSDs work well. Make sure the size of these SSDs are at least 250GB in size, larger if more IOPS are needed. - The number of IOPS are proportional to the size of the disk.
- Avoid running on
g1machine families. These are burstable, shared core machines that may not deliver steady performance.