Notice

This document is for a development version of Ceph.

hardware recommendations

Ceph is designed to run on commodity hardware, which makes building and maintaining petabyte-scale data clusters flexible and economically feasible. When planning your cluster’s hardware, you will need to balance a number of considerations, including failure domains, cost, and performance. Hardware planning should include distributing Ceph daemons and other processes that use Ceph across many hosts. Generally, we recommend running Ceph daemons of a specific type on a host configured for that type of daemon. We recommend using separate hosts for processes that utilize your data cluster (e.g., OpenStack, OpenNebula, CloudStack, Kubernetes, etc).

The requirements of one Ceph cluster are not the same as the requirements of another, but below are some general guidelines.

Tip

check out the ceph blog too.

CPU

CephFS Metadata Servers (MDS) are CPU-intensive. They are are single-threaded and perform best with CPUs with a high clock rate (GHz). MDS servers do not need a large number of CPU cores unless they are also hosting other services, such as SSD OSDs for the CephFS metadata pool. OSD nodes need enough processing power to run the RADOS service, to calculate data placement with CRUSH, to replicate data, and to maintain their own copies of the cluster map.

With earlier releases of Ceph, we would make hardware recommendations based on the number of cores per OSD, but this cores-per-osd metric is no longer as useful a metric as the number of cycles per IOP and the number of IOPS per OSD. For example, with NVMe OSD drives, Ceph can easily utilize five or six cores on real clusters and up to about fourteen cores on single OSDs in isolation. So cores per OSD are no longer as pressing a concern as they were. When selecting hardware, select for IOPS per core.

Tip

When we speak of CPU cores, we mean threads when hyperthreading is enabled. Hyperthreading is usually beneficial for Ceph servers.

Monitor nodes and Manager nodes do not have heavy CPU demands and require only modest processors. if your hosts will run CPU-intensive processes in addition to Ceph daemons, make sure that you have enough processing power to run both the CPU-intensive processes and the Ceph daemons. (OpenStack Nova is one example of a CPU-intensive process.) We recommend that you run non-Ceph CPU-intensive processes on separate hosts (that is, on hosts that are not your Monitor and Manager nodes) in order to avoid resource contention. If your cluster deployes the Ceph Object Gateway, RGW daemons may co-reside with your Mon and Manager services if the nodes have sufficient resources.

RAM

Generally, more RAM is better. Monitor / Manager nodes for a modest cluster might do fine with 64GB; for a larger cluster with hundreds of OSDs 128GB is advised.

Tip

when we speak of RAM and storage requirements, we often describe the needs of a single daemon of a given type. A given server as a whole will thus need at least the sum of the needs of the daemons that it hosts as well as resources for logs and other operating system components. Keep in mind that a server’s need for RAM and storage will be greater at startup and when components fail or are added and the cluster rebalances. In other words, allow headroom past what you might see used during a calm period on a small initial cluster footprint.

There is an osd_memory_target setting for BlueStore OSDs that defaults to 4GB. Factor in a prudent margin for the operating system and administrative tasks (like monitoring and metrics) as well as increased consumption during recovery: provisioning ~8GB per BlueStore OSD is thus advised.

Monitors and managers (ceph-mon and ceph-mgr)

Monitor and manager daemon memory usage scales with the size of the cluster. Note that at boot-time and during topology changes and recovery these daemons will need more RAM than they do during steady-state operation, so plan for peak usage. For very small clusters, 32 GB suffices. For clusters of up to, say, 300 OSDs go with 64GB. For clusters built with (or which will grow to) even more OSDs you should provision 128GB. You may also want to consider tuning the following settings:

Metadata servers (ceph-mds)

CephFS metadata daemon memory utilization depends on the configured size of its cache. We recommend 1 GB as a minimum for most systems. See mds_cache_memory_limit.

Memory

Bluestore uses its own memory to cache data rather than relying on the operating system’s page cache. In Bluestore you can adjust the amount of memory that the OSD attempts to consume by changing the osd_memory_target configuration option.

  • Setting the osd_memory_target below 2GB is not recommended. Ceph may fail to keep the memory consumption under 2GB and extremely slow performance is likely.

  • Setting the memory target between 2GB and 4GB typically works but may result in degraded performance: metadata may need to be read from disk during IO unless the active data set is relatively small.

  • 4GB is the current default value for osd_memory_target This default was chosen for typical use cases, and is intended to balance RAM cost and OSD performance.

  • Setting the osd_memory_target higher than 4GB can improve performance when there many (small) objects or when large (256GB/OSD or more) data sets are processed. This is especially true with fast NVMe OSDs.

Important

OSD memory management is “best effort”. Although the OSD may unmap memory to allow the kernel to reclaim it, there is no guarantee that the kernel will actually reclaim freed memory within a specific time frame. This applies especially in older versions of Ceph, where transparent huge pages can prevent the kernel from reclaiming memory that was freed from fragmented huge pages. Modern versions of Ceph disable transparent huge pages at the application level to avoid this, but that does not guarantee that the kernel will immediately reclaim unmapped memory. The OSD may still at times exceed its memory target. We recommend budgeting at least 20% extra memory on your system to prevent OSDs from going OOM (Out Of Memory) during temporary spikes or due to delay in the kernel reclaiming freed pages. That 20% value might be more or less than needed, depending on the exact configuration of the system.

Tip

Configuring the operating system with swap to provide additional virtual memory for daemons is not advised for modern systems. Doing may result in lower performance, and your Ceph cluster may well be happier with a daemon that crashes vs one that slows to a crawl.

When using the legacy FileStore back end, the OS page cache was used for caching data, so tuning was not normally needed. When using the legacy FileStore backend, the OSD memory consumption was related to the number of PGs per daemon in the system.

Data Storage

Plan your data storage configuration carefully. There are significant cost and performance tradeoffs to consider when planning for data storage. Simultaneous OS operations and simultaneous requests from multiple daemons for read and write operations against a single drive can impact performance.

OSDs require substantial storage drive space for RADOS data. We recommend a minimum drive size of 1 terabyte. OSD drives much smaller than one terabyte use a significant fraction of their capacity for metadata, and drives smaller than 100 gigabytes will not be effective at all.

It is strongly suggested that (enterprise-class) SSDs are provisioned for, at a minimum, Ceph Monitor and Ceph Manager hosts, as well as CephFS Metadata Server metadata pools and Ceph Object Gateway (RGW) index pools, even if HDDs are to be provisioned for bulk OSD data.

To get the best performance out of Ceph, provision the following on separate drives:

  • The operating systems

  • OSD data

  • BlueStore WAL+DB

For more information on how to effectively use a mix of fast drives and slow drives in your Ceph cluster, see the block and block.db section of the Bluestore Configuration Reference.

Hard Disk Drives

Consider carefully the cost-per-gigabyte advantage of larger disks. We recommend dividing the price of the disk drive by the number of gigabytes to arrive at a cost per gigabyte, because larger drives may have a significant impact on the cost-per-gigabyte. For example, a 1 terabyte hard disk priced at $75.00 has a cost of $0.07 per gigabyte (i.e., $75 / 1024 = 0.0732). By contrast, a 3 terabyte disk priced at $150.00 has a cost of $0.05 per gigabyte (i.e., $150 / 3072 = 0.0488). In the foregoing example, using the 1 terabyte disks would generally increase the cost per gigabyte by 40%--rendering your cluster substantially less cost efficient.

Tip

Hosting multiple OSDs on a single SAS / SATA HDD is NOT a good idea.

Tip

Hosting an OSD with monitor, manager, or MDS data on a single drive is also NOT a good idea.

Tip

With spinning disks, the SATA and SAS interface increasingly becomes a bottleneck at larger capacities. See also the Storage Networking Industry Association’s Total Cost of Ownership calculator.

Storage drives are subject to limitations on seek time, access time, read and write times, as well as total throughput. These physical limitations affect overall system performance--especially during recovery. We recommend using a dedicated (ideally mirrored) drive for the operating system and software, and one drive for each Ceph OSD Daemon you run on the host. Many “slow OSD” issues (when they are not attributable to hardware failure) arise from running an operating system and multiple OSDs on the same drive. Also be aware that today’s 22TB HDD uses the same SATA interface as a 3TB HDD from ten years ago: more than seven times the data to squeeze through the same interface. For this reason, when using HDDs for OSDs, drives larger than 8TB may be best suited for storage of large files / objects that are not at all performance-sensitive.

Solid State Drives

Ceph performance is much improved when using solid-state drives (SSDs). This reduces random access time and reduces latency while increasing throughput.

SSDs cost more per gigabyte than do HDDs but SSDs often offer access times that are, at a minimum, 100 times faster than HDDs. SSDs avoid hotspot issues and bottleneck issues within busy clusters, and they may offer better economics when TCO is evaluated holistically. Notably, the amortized drive cost for a given number of IOPS is much lower with SSDs than with HDDs. SSDs do not suffer rotational or seek latency and in addition to improved client performance, they substantially improve the speed and client impact of cluster changes including rebalancing when OSDs or Monitors are added, removed, or fail.

SSDs do not have moving mechanical parts, so they are not subject to many of the limitations of HDDs. SSDs do have significant limitations though. When evaluating SSDs, it is important to consider the performance of sequential and random reads and writes.

Important

We recommend exploring the use of SSDs to improve performance. However, before making a significant investment in SSDs, we strongly recommend reviewing the performance metrics of an SSD and testing the SSD in a test configuration in order to gauge performance.

Relatively inexpensive SSDs may appeal to your sense of economy. Use caution. Acceptable IOPS are not the only factor to consider when selecting SSDs for use with Ceph. Bargain SSDs are often a false economy: they may experience “cliffing”, which means that after an initial burst, sustained performance once a limited cache is filled declines considerably. Consider also durability: a drive rated for 0.3 Drive Writes Per Day (DWPD or equivalent) may be fine for OSDs dedicated to certain types of sequentially-written read-mostly data, but are not a good choice for Ceph Monitor duty. Enterprise-class SSDs are best for Ceph: they almost always feature power loss protection (PLP) and do not suffer the dramatic cliffing that client (desktop) models may experience.

When using a single (or mirrored pair) SSD for both operating system boot and Ceph Monitor / Manager purposes, a minimum capacity of 256GB is advised and at least 480GB is recommended. A drive model rated at 1+ DWPD (or the equivalent in TBW (TeraBytes Written) is suggested. However, for a given write workload, a larger drive than technically required will provide more endurance because it effectively has greater overprovsioning. We stress that enterprise-class drives are best for production use, as they feature power loss protection and increased durability compared to client (desktop) SKUs that are intended for much lighter and intermittent duty cycles.

SSDs have historically been cost prohibitive for object storage, but QLC SSDs are closing the gap, offering greater density with lower power consumption and less power spent on cooling. Also, HDD OSDs may see a significant write latency improvement by offloading WAL+DB onto an SSD. Many Ceph OSD deployments do not require an SSD with greater endurance than 1 DWPD (aka “read-optimized”). “Mixed-use” SSDs in the 3 DWPD class are often overkill for this purpose and cost signficantly more.

To get a better sense of the factors that determine the total cost of storage, you might use the Storage Networking Industry Association’s Total Cost of Ownership calculator

Partition Alignment

When using SSDs with Ceph, make sure that your partitions are properly aligned. Improperly aligned partitions suffer slower data transfer speeds than do properly aligned partitions. For more information about proper partition alignment and example commands that show how to align partitions properly, see Werner Fischer’s blog post on partition alignment.

CephFS Metadata Segregation

One way that Ceph accelerates CephFS file system performance is by separating the storage of CephFS metadata from the storage of the CephFS file contents. Ceph provides a default metadata pool for CephFS metadata. You will never have to manually create a pool for CephFS metadata, but you can create a CRUSH map hierarchy for your CephFS metadata pool that includes only SSD storage media. See CRUSH Device Class for details.

Controllers

Disk controllers (HBAs) can have a significant impact on write throughput. Carefully consider your selection of HBAs to ensure that they do not create a performance bottleneck. Notably, RAID-mode (IR) HBAs may exhibit higher latency than simpler “JBOD” (IT) mode HBAs. The RAID SoC, write cache, and battery backup can substantially increase hardware and maintenance costs. Many RAID HBAs can be configured with an IT-mode “personality” or “JBOD mode” for streamlined operation.

You do not need an RoC (RAID-capable) HBA. ZFS or Linux MD software mirroring serve well for boot volume durability. When using SAS or SATA data drives, forgoing HBA RAID capabilities can reduce the gap between HDD and SSD media cost. Moreover, when using NVMe SSDs, you do not need any HBA. This additionally reduces the HDD vs SSD cost gap when the system as a whole is considered. The initial cost of a fancy RAID HBA plus onboard cache plus battery backup (BBU or supercapacitor) can easily exceed more than 1000 US dollars even after discounts - a sum that goes a log way toward SSD cost parity. An HBA-free system may also cost hundreds of US dollars less every year if one purchases an annual maintenance contract or extended warranty.

Tip

The Ceph blog is often an excellent source of information on Ceph performance issues. See Ceph Write Throughput 1 and Ceph Write Throughput 2 for additional details.

Benchmarking

BlueStore opens storage devices with O_DIRECT and issues fsync() frequently to ensure that data is safely persisted to media. You can evaluate a drive’s low-level write performance using fio. For example, 4kB random write performance is measured as follows:

# fio --name=/dev/sdX --ioengine=libaio --direct=1 --fsync=1 --readwrite=randwrite --blocksize=4k --runtime=300

Write Caches

Enterprise SSDs and HDDs normally include power loss protection features which ensure data durability when power is lost while operating, and use multi-level caches to speed up direct or synchronous writes. These devices can be toggled between two caching modes -- a volatile cache flushed to persistent media with fsync, or a non-volatile cache written synchronously.

These two modes are selected by either “enabling” or “disabling” the write (volatile) cache. When the volatile cache is enabled, Linux uses a device in “write back” mode, and when disabled, it uses “write through”.

The default configuration (usually: caching is enabled) may not be optimal, and OSD performance may be dramatically increased in terms of increased IOPS and decreased commit latency by disabling this write cache.

Users are therefore encouraged to benchmark their devices with fio as described earlier and persist the optimal cache configuration for their devices.

The cache configuration can be queried with hdparm, sdparm, smartctl or by reading the values in /sys/class/scsi_disk/*/cache_type, for example:

# hdparm -W /dev/sda

/dev/sda:
 write-caching =  1 (on)

# sdparm --get WCE /dev/sda
    /dev/sda: ATA       TOSHIBA MG07ACA1  0101
WCE           1  [cha: y]
# smartctl -g wcache /dev/sda
smartctl 7.1 2020-04-05 r5049 [x86_64-linux-4.18.0-305.19.1.el8_4.x86_64] (local build)
Copyright (C) 2002-19, Bruce Allen, Christian Franke, www.smartmontools.org

Write cache is:   Enabled

# cat /sys/class/scsi_disk/0\:0\:0\:0/cache_type
write back

The write cache can be disabled with those same tools:

# hdparm -W0 /dev/sda

/dev/sda:
 setting drive write-caching to 0 (off)
 write-caching =  0 (off)

# sdparm --clear WCE /dev/sda
    /dev/sda: ATA       TOSHIBA MG07ACA1  0101
# smartctl -s wcache,off /dev/sda
smartctl 7.1 2020-04-05 r5049 [x86_64-linux-4.18.0-305.19.1.el8_4.x86_64] (local build)
Copyright (C) 2002-19, Bruce Allen, Christian Franke, www.smartmontools.org

=== START OF ENABLE/DISABLE COMMANDS SECTION ===
Write cache disabled

In most cases, disabling this cache using hdparm, sdparm, or smartctl results in the cache_type changing automatically to “write through”. If this is not the case, you can try setting it directly as follows. (Users should ensure that setting cache_type also correctly persists the caching mode of the device until the next reboot as some drives require this to be repeated at every boot):

# echo "write through" > /sys/class/scsi_disk/0\:0\:0\:0/cache_type

# hdparm -W /dev/sda

/dev/sda:
 write-caching =  0 (off)

Tip

This udev rule (tested on CentOS 8) will set all SATA/SAS device cache_types to “write through”:

# cat /etc/udev/rules.d/99-ceph-write-through.rules
ACTION=="add", SUBSYSTEM=="scsi_disk", ATTR{cache_type}:="write through"

Tip

This udev rule (tested on CentOS 7) will set all SATA/SAS device cache_types to “write through”:

# cat /etc/udev/rules.d/99-ceph-write-through-el7.rules
ACTION=="add", SUBSYSTEM=="scsi_disk", RUN+="/bin/sh -c 'echo write through > /sys/class/scsi_disk/$kernel/cache_type'"

Tip

The sdparm utility can be used to view/change the volatile write cache on several devices at once:

# sdparm --get WCE /dev/sd*
    /dev/sda: ATA       TOSHIBA MG07ACA1  0101
WCE           0  [cha: y]
    /dev/sdb: ATA       TOSHIBA MG07ACA1  0101
WCE           0  [cha: y]
# sdparm --clear WCE /dev/sd*
    /dev/sda: ATA       TOSHIBA MG07ACA1  0101
    /dev/sdb: ATA       TOSHIBA MG07ACA1  0101

Additional Considerations

Ceph operators typically provision multiple OSDs per host, but you should ensure that the aggregate throughput of your OSD drives doesn’t exceed the network bandwidth required to service a client’s read and write operations. You should also consider each host’s percentage of the cluster’s overall capacity. If the percentage located on a particular host is large and the host fails, it can lead to problems such as recovery causing OSDs to exceed the full ratio, which in turn causes Ceph to halt operations to prevent data loss.

When you run multiple OSDs per host, you also need to ensure that the kernel is up to date. See OS Recommendations for notes on glibc and syncfs(2) to ensure that your hardware performs as expected when running multiple OSDs per host.

Networks

Provision at least 10 Gb/s networking in your datacenter, both among Ceph hosts and between clients and your Ceph cluster. Network link active/active bonding across separate network switches is strongly recommended both for increased throughput and for tolerance of network failures and maintenance. Take care that your bonding hash policy distributes traffic across links.

Speed

It takes three hours to replicate 1 TB of data across a 1 Gb/s network and it takes thirty hours to replicate 10 TB across a 1 Gb/s network. But it takes only twenty minutes to replicate 1 TB across a 10 Gb/s network, and it takes only one hour to replicate 10 TB across a 10 Gb/s network.

Note that a 40 Gb/s network link is effectively four 10 Gb/s channels in parallel, and that a 100Gb/s network link is effectively four 25 Gb/s channels in parallel. Thus, and perhaps somewhat counterintuitively, an individual packet on a 25 Gb/s network has slightly lower latency compared to a 40 Gb/s network.

Cost

The larger the Ceph cluster, the more common OSD failures will be. The faster that a placement group (PG) can recover from a degraded state to an active + clean state, the better. Notably, fast recovery minimizes the likelihood of multiple, overlapping failures that can cause data to become temporarily unavailable or even lost. Of course, when provisioning your network, you will have to balance price against performance.

Some deployment tools employ VLANs to make hardware and network cabling more manageable. VLANs that use the 802.1q protocol require VLAN-capable NICs and switches. The added expense of this hardware may be offset by the operational cost savings on network setup and maintenance. When using VLANs to handle VM traffic between the cluster and compute stacks (e.g., OpenStack, CloudStack, etc.), there is additional value in using 10 Gb/s Ethernet or better; 40 Gb/s or increasingly 25/50/100 Gb/s networking as of 2022 is common for production clusters.

Top-of-rack (TOR) switches also need fast and redundant uplinks to core / spine network switches or routers, often at least 40 Gb/s.

Baseboard Management Controller (BMC)

Your server chassis should have a Baseboard Management Controller (BMC). Well-known examples are iDRAC (Dell), CIMC (Cisco UCS), and iLO (HPE). Administration and deployment tools may also use BMCs extensively, especially via IPMI or Redfish, so consider the cost/benefit tradeoff of an out-of-band network for security and administration. Hypervisor SSH access, VM image uploads, OS image installs, management sockets, etc. can impose significant loads on a network. Running multiple networks may seem like overkill, but each traffic path represents a potential capacity, throughput and/or performance bottleneck that you should carefully consider before deploying a large scale data cluster.

Additionally BMCs as of 2023 rarely sport network connections faster than 1 Gb/s, so dedicated and inexpensive 1 Gb/s switches for BMC administrative traffic may reduce costs by wasting fewer expenive ports on faster host switches.

Failure Domains

A failure domain can be thought of as any component loss that prevents access to one or more OSDs or other Ceph daemons. These could be a stopped daemon on a host; a storage drive failure, an OS crash, a malfunctioning NIC, a failed power supply, a network outage, a power outage, and so forth. When planning your hardware deployment, you must balance the risk of reducing costs by placing too many responsibilities into too few failure domains against the added costs of isolating every potential failure domain.

Minimum Hardware Recommendations

Ceph can run on inexpensive commodity hardware. Small production clusters and development clusters can run successfully with modest hardware. As we noted above: when we speak of CPU cores, we mean threads when hyperthreading (HT) is enabled. Each modern physical x64 CPU core typically provides two logical CPU threads; other CPU architectures may vary.

Take care that there are many factors that influence resource choices. The minimum resources that suffice for one purpose will not necessarily suffice for another. A sandbox cluster with one OSD built on a laptop with VirtualBox or on a trio of Raspberry PIs will get by with fewer resources than a production deployment with a thousand OSDs serving five thousand of RBD clients. The classic Fisher Price PXL 2000 captures video, as does an IMAX or RED camera. One would not expect the former to do the job of the latter. We especially cannot stress enough the criticality of using enterprise-quality storage media for production workloads.

Additional insights into resource planning for production clusters are found above and elsewhere within this documentation.

Process

Criteria

Bare Minimum and Recommended

ceph-osd

Processor

  • 1 core minimum, 2 recommended

  • 1 core per 200-500 MB/s throughput

  • 1 core per 1000-3000 IOPS

  • Results are before replication.

  • Results may vary across CPU and drive models and Ceph configuration: (erasure coding, compression, etc)

  • ARM processors specifically may require more cores for performance.

  • SSD OSDs, especially NVMe, will benefit from additional cores per OSD.

  • Actual performance depends on many factors including drives, net, and client throughput and latency. Benchmarking is highly recommended.

RAM

  • 4GB+ per daemon (more is better)

  • 2-4GB may function but may be slow

  • Less than 2GB is not recommended

Storage Drives

1x storage drive per OSD

DB/WAL (optional)

1x SSD partion per HDD OSD 4-5x HDD OSDs per DB/WAL SATA SSD <= 10 HDD OSDss per DB/WAL NVMe SSD

Network

1x 1Gb/s (bonded 10+ Gb/s recommended)

ceph-mon

Processor

  • 2 cores minimum

RAM

5GB+ per daemon (large / production clusters need more)

Storage

100 GB per daemon, SSD is recommended

Network

1x 1Gb/s (10+ Gb/s recommended)

ceph-mds

Processor

  • 2 cores minimum

RAM

2GB+ per daemon (more for production)

Disk Space

1 GB per daemon

Network

1x 1Gb/s (10+ Gb/s recommended)

Tip

If you are running an OSD node with a single storage drive, create a partition for your OSD that is separate from the partition containing the OS. We recommend separate drives for the OS and for OSD storage.

Brought to you by the Ceph Foundation

The Ceph Documentation is a community resource funded and hosted by the non-profit Ceph Foundation. If you would like to support this and our other efforts, please consider joining now.