SCALE Hardware Guide

From repurposed systems to highly-custom builds, the fundamental freedom of TrueNAS is the ability to run it on almost any x86 computer.

The recommended system requirements to install TrueNAS:

The TrueNAS installer recommends 8 GB of RAM. TrueNAS installs, runs, operates jails, hosts SMB shares, and replicates TBs of data with less. iXsystems recommends the above for better performance and fewer issues.

You do not need an SSD boot device, but we discourage using a spinner or a USB stick for obvious reasons. We do not recommend installing TrueNAS on a single disk or striped pool unless you have a good reason to do so. You can install and run TrueNAS without any data device, but we strongly discourage it.

TrueNAS does not require two cores, as most halfway-modern 64-bit CPUs likely already have at least two.

For help building a system according to your unique performance, storage, and networking requirements, read on!

The heart of any storage system is the symbiotic pairing of its file system and physical storage devices. The ZFS file system in TrueNAS provides the best available data protection of any file system at any cost and makes very effective use of both spinning-disk and all-flash storage or a mix of the two. ZFS is prepared for the eventual failure of storage devices. It is highly configurable to achieve the perfect balance of redundancy and performance to meet any storage goal. A properly-configured TrueNAS system can tolerate the failure of multiple storage devices and even recreate its boot media with a copy of the configuration file.

TrueNAS is capable of managing large quantities of storage devices as part of a single storage array. The community-focused TrueNAS SCALE Angelfish release can manage as many as 400 drives in a single storage array; a significant level of flexibility for home users to larger business deployments. With more Enterprise-level tuning in the mature 13.0 release and similar tuning in the upcoming SCALE Bluefin release, TrueNAS can expand even further and manage as many as 1,250 drives in a single storage array!

Choosing storage media is the first step in designing the storage system to meet immediate objectives and prepare for future capacity expansion.

These storage device media arrange together to create powerful storage solutions.

Zpool layout (the organization of LUNs and volumes, in TrueNAS/ZFS parlance) is outside of the scope of this guide. The availability of double-digit terabyte drives raises a question TrueNAS users now have the luxury of asking: How many drives should I use to achieve my desired capacity? You can mirror two 16TB drives to achieve 16TB of available capacity, but that does not mean you should. Mirroring two large drives offers the advantage of redundancy and balancing reads between the two devices, which could lower power draw, but little else. The write performance of two large drives, at most, is that of a single drive. By contrast, an array of eight 4TB drives offers a wide range of configurations to optimize performance and redundancy at a lower cost. If configured as striped mirrors, eight drives could yield four times greater write performance with a similar total capacity. You might also consider adding a hot-spare drive with any zpool configuration, which lets the zpool automatically rebuild itself if one of its primary drives fails.

Spinning disk hard drives have moving parts that are highly sensitive to shock and vibration and wear out with use. Consider pre-flighting every storage device before putting it into production, paying attention to:

• Start a long HDD self-test (smartctl -t long /dev/), and after the test completes (could take 12+ hrs)
• Check the results (smartctl -a /dev/)
• Check pending sector reallocations (smartctl -a /dev/ | grep Current_Pending_Sector)
• Check reallocated sector count (smartctl -a /dev/ | grep Reallocated_Sector_Ct)
• Check the UDMA CRC errors (smartctl -a /dev/ | grep UDMA_CRC_Error_Count)
• Check HDD and SSD write latency consistency (diskinfo -wS ) Unformatted drives only!
• Check HDD and SSD hours (smartctl -a /dev/ | grep Power_On_Hours)
• Check NVMe percentage used (nvmecontrol logpage -p 2 nvme0 | grep “Percentage used”)

Take time to create a pool before deploying the system. Subject it to as close to a real-world workload as possible to reveal individual drive issues and help determine if an alternative pool layout is better suited to that workload. Be cautious of used drives as vendors may not be honest or informed about their age and health. Check the number of hours on all new drives using smartctl(8) to verify they are not recertified. A drive vendor could also zero the hours of a drive during recertification, masking its true age. iXsystems tests all storage devices it sells for at least 48 hours before shipment.

The uncontested most popular storage controllers used with TrueNAS are the 6 and 12 Gbps (Gigabits per second, sometimes expressed as Gb/s) Broadcom (formerly Avago, formerly LSI) SAS host bus adapters (HBA). Controllers ship embedded on some motherboards but are generally PCIe cards with four or more internal or external SATA/SAS ports. The 6 Gbps LSI 9211 and its rebranded siblings that also use the LSI SAS2008 chip, such as the IBM M1015 and Dell H200, are legendary among TrueNAS users who build systems using parts from the second-hand market. Flash using the latest IT or Target Mode firmware to disable the optional RAID functionality found in the IR firmware on Broadcom controllers. For those with the budget, newer models like the Broadcom 9300/9400 series give 12 Gbps SAS capabilities and even NVMe to SAS translation abilities with the 9400 series. TrueNAS includes the sas2flash, sas3flash, and storcli commands to flash or perform re-flashing operations on 9200, 9300, and 9400 series cards.

Onboard SATA controllers are popular with smaller builds, but motherboard vendors are better at catering to the needs of NAS users by including more than the traditional four SATA interfaces. Be aware that many motherboards ship with a mix of 3 Gbps and 6 Gbps onboard SATA interfaces and that choosing the wrong one could impact performance. If a motherboard includes hardware RAID functionality, do not use or configure it, but note that disabling it in the BIOS might remove some SATA functionality depending on the motherboard. Most SATA compatibility-related issues are immediately apparent.

There are countless warnings against using hardware RAID cards with TrueNAS. ZFS and TrueNAS provide a built-in RAID that protects your data better than any hardware RAID card. You can use a hardware RAID card if it is all you have, but there are limitations. First and most importantly, do not use their RAID facility if your hardware RAID card supports HBA mode, also known as passthrough or JBOD mode (there is one caveat in the bullets below). When used, it allows it to perform indistinguishably from a standard HBA. If your RAID card does not have this mode, you can configure a RAID0 for every single disk in your system. While not the ideal setup, it works in a pinch. If repurposing hardware RAID cards with TrueNAS, be aware that some hardware RAID cards:

• Could mask disk serial number and S.M.A.R.T. health information
• Could perform slower than their HBA equivalents
• Could cause data loss if using a write cache with a dead battery backup unit (BBU))

A direct-attached system, where every disk connects to an interface on the controller card, is optimal but not always possible. A SAS expander (a port multiplier or splitter) enables each SAS port on a controller card to service many disks. You find SAS expanders only on the drive backplane of servers or JBODs with more than twelve drive bays. For example, a TrueNAS JBOD that eclipses 90 drives in only four rack units of space is not possible without SAS expanders. Imagine how many eight-port HBAs you would need to access 90 drives without SAS expanders.

While SAS expanders, designed for SAS disks, can often support SATA disks via the SATA Tunneling Protocol or STP, we still prefer SAS disks for reasons mentioned in the NL-SAS section above (SATA disks function on a SAS-based backplane). Note that the opposite is not true: you cannot use a SAS drive in a port designed for SATA drives.

A much-cited study floating around the Internet asserts that drive temperature has little impact on drive reliability. The study makes for a great headline or conversation starter, but carefully reading the report indicates that the drives were tested under optimal environmental conditions. The average temperature that a well-cooled spinning hard disk reaches in production is around 28 °C, and one study found that disks experience twice the number of failures for every 12 °C increase in temperature. Before adding drive cooling that often comes with added noise (especially on older systems), know that you risk throwing money away by running a server in a data center or closet without noticing that the internal cooling fans are set to their lowest setting. Pay close attention to drive temperature in any chassis that supports 16 or more drives, especially if they are exotic, high-density designs. Every chassis has certain areas that are warmer for whatever reason. Watch for fan failures and the tendency for some models of 8TB drives to run hotter than other drive capacities. In general, try to keep drive temperatures below the drive specification provided by vendor.

TrueNAS has higher memory requirements than many Network Attached Storage solutions for good reason: it shares dynamic random-access memory (DRAM or simply RAM) between sharing services, add-on plugins, jails, and virtual machines, and sophisticated read caching. RAM rarely goes unused on a TrueNAS system and enough RAM is key to maintaining peak performance. You should have at least 8 GB of RAM for basic TrueNAS operations with up to eight drives. Other use cases each have distinct RAM requirements:

• Add 1 GB for each drive added after eight to benefit most use cases.
• Add extra RAM (in general) if more clients will connect to the TrueNAS system. A 20 TB pool backing lots of high-performance VMs over iSCSI might need more RAM than a 200 TB pool storing archival data. If using iSCSI to back VMs, plan to use at least 16 GB of RAM for reasonable performance and 32 GB or more for optimal performance.
• Add 2 GB of RAM for directory services for the winbind internal cache.
• Add more RAM as required for plugins and jails as each has specific application RAM requirements.
• Add more RAM for virtual machines with a guest operating system and application RAM requirements.
• Add the suggested 5 GB per TB of storage for deduplication that depends on an in-RAM deduplication table.
• Add approximately 1 GB of RAM (conservative estimate) for every 50 GB of L2ARC in your pool. Attaching an L2ARC drive to a pool uses some RAM, too. ZFS needs metadata in ARC to know what data is in L2ARC.

Electrical or magnetic interference inside a computer system can cause a spontaneous flip of a single bit of RAM to the opposite state, resulting in a memory error. Memory errors can cause security vulnerabilities, crashes, transcription errors, lost transactions, and corrupted or lost data. So RAM, the temporary data storage location, is one of the most vital areas for preventing data loss.

Error-correcting code or ECC RAM detects and corrects in-memory bit errors as they occur. If errors are severe enough to be uncorrectable, ECC memory causes the system to hang (become unresponsive) rather than continue with errored bits. For ZFS and TrueNAS, this behavior virtually eliminates any chances that RAM errors pass to the drives to cause corruption of the ZFS pools or file errors.

The lengthy, Internet-wide debate on whether to use error-correcting code (ECC) system memory with OpenZFS and TrueNAS summarizes as:

• ECC RAM is strongly recommended as another data integrity defense

However:

• Some CPUs or motherboards support ECC RAM but not all
• Many TrueNAS systems operate every day without ECC RAM
• RAM of any type or grade can fail and cause data loss
• RAM is most likely to fail in the first three months so test all RAM before deployment.

Choosing ECC RAM limits your CPU and motherboard options, but that can be a good thing. Intel^® makes a point of limiting ECC RAM support to their lowest and highest-end CPUs, cutting out the mid-range i5 and i7 models.

Which CPU to choose can come down to a short list of factors:

• An underpowered CPU can create a performance bottleneck because of how OpenZFS does checksums, and compresses and (optional) encrypts data.
• A higher-frequency CPU with fewer cores usually performs best for SMB only workloads because of Samba, the lightly-threaded TrueNAS SMB daemon.
• A higher-core-count CPU is better suited for parallel encryption and virtualization.
• A CPU with AES-NI encryption acceleration support improves the speed of the file system and network encryption.
• A server-class CPU is recommended for its power and ECC memory support.
• A Xeon E5 CPU (or similar) is recommended for software-encrypted pools.
• An Intel Ivy Bridge CPU or later recommended for virtual machine use.

Watch for VT-d/AMD-Vi device virtualization support on the CPU and motherboard to pass PCIe devices to virtual machines. Be aware if a given CPU contains a GPU or requires an external one. Also, note that many server motherboards include a BMC chip with a built-in GPU. See below for more details on BMCs.

AMD CPUs are making a comeback thanks to the Ryzen and EPYC (Naples/Rome) lines. Support for these platforms is limited on FreeBSD and, by extension, TrueNAS CORE. However, Linux has significant support, and TrueNAS SCALE should work with AMD CPUs without issue.

As a courtesy to further limit the motherboard choices, consider the Intelligent Platform Management Interface or IPMI (a.k.a. baseboard management controller, BMC, iLo, iDrac, and other names depending on the vendor) if you need:

• Remote power control and monitoring of remote systems
• Remote console shell access for configuration or data recovery
• Remote virtual media for TrueNAS installation or reinstallation

TrueNAS relies on its web-based user interface (UI), but you might occasionally need console access to make network configuration changes. TrueNAS administration and sharing default to a single network interface, which can be challenging when you need to upgrade features like LACP aggregated networking. The ideal solution is to have a dedicated subnet to access the TrueNAS web UI, but not all users have this luxury. The occasional visit to the hardware console is necessary for global configuration and even for system recovery. The latest TrueNAS Mini and R-Series systems ship with full-featured, HTML5-based IPMI support on a dedicated gigabit network interface.

The top criteria to consider for a power supply unit (or PSU) on a TrueNAS system are its:

• Power capacity (in watts) for the motherboard and number of drives it must support
• Reliability
• Efficiency rating
• Relative noise
• Optional redundancy to keep important systems running if one power supply fails

Select a PSU rated for the initial and a future load placed on it. Have a PSU with adequate power to migrate from a large-capacity chassis to a fully-populated chassis. Also, consider a hot-swappable redundant PSU to help guarantee uptime. Users on a budget can keep a cold spare PSU to limit their potential downtime to hours rather than days. A good, modern PSU is efficient and completely integrates into the IPMI management system to provide real-time fan, temperature, and load information.

Most power supplies carry a certified efficiency rating known as an 80 Plus rating. The 80 plus rating indicates the power drawn from the wall is lost as heat, noise, and vibration, instead of doing useful work like powering your components. If a power supply needs to draw 600 watts from the wall to provide 500 watts of power to your components, it is operating at 500/600 = ~83% efficiency. The other 100 watts get lost as heat, noise, and vibration. Power supplies with higher ratings are more efficient but also far more expensive. Do some return-on-investment calculations if you are unsure what efficiency to buy. For example, if an 80 Plus Platinum PSU costs $50 more than the comparable 80 Plus Gold, it should save you at least $10 per year on your power bill for that investment to pay off over five years. You can read more about 80 Plus ratings in this post.

TrueNAS provides the ability to communicate with a battery-backed, uninterruptible power supply (UPS) over a traditional serial or USB connection to coordinate a graceful shutdown in the case of power loss. TrueNAS works well with APC brand UPSs, followed by CyberPower. Consider budgeting for a UPS with pure sine wave output. Some models of SSD can experience data corruption on power loss. If several SSDs experience simultaneous power loss, it could cause total pool failure, making a UPS a critical investment.

The network in Network Attached Storage is as important as storage, but the topic reduces to a few key points:

• Simplicity - Simplicity is often the secret to reliability with network configurations.
• Individual interfaces - Faster individual interfaces such as 10/25/40/100GbE are preferable to aggregating slower interfaces.
• Interface support - Intel and Chelsio interfaces are the best-supported options.
• Packet fragmentation - Only consider a jumbo frames MTU with dedicated connections such as between servers or video editors and TrueNAS that are unlikely to experience packet fragmentation.
• LRO/LSO offload features - Interfaces with LRO and LSO offload features generally alleviates the need for jumbo frames and their use can result in lower CPU overhead.

Higher band hardware is becoming more accessible as the hardware development pace increases and enterprises upgrade more quickly. Home labs can now deploy and use 40 GB and higher networking components. Home users are now discovering the same issues and problems with these higher speeds found by Enterprise customers.

iXsystems recommends using optical fiber over direct attached copper (DAC) cables for the high speed interconnects listed below:

• 10Gb NICs: SFP+ connectors
• 25Gb NICs: SFP28 connectors
• 40Gb NICs: QSFP+ connectors
• 100Gb NICs: QSFP28 connectors
• 200Gb NICs: QSFP56 connectors
• 400Gb NICs: QSFP-DD connectors

iXsystems also recommends using optical fiber for any transceiver form factors mentioned when using fiber channels. Direct attached copper (DAC) cables could create interoperability issues between the NIC, cable, and switch.

Finally, the ultimate TrueNAS hardware question is whether to use actual hardware or choose a virtualization solution. TrueNAS developers virtualize TrueNAS every day as part of their work, and cloud services are popular among users of all sizes. At the heart of the TrueNAS design is OpenZFS. The design from day one works with physical storage devices. It is aware of their strengths and compensates for their weaknesses. When the need arises to virtualize TrueNAS:

• Pass hardware disks or the entire storage controller to the TrueNAS VM if possible (requires VT-d/AMD-Vi support).
• Disable automatic scrub pools on virtualized storage such as VMFS, and never scrub a pool while also running storage repair tasks on another layer.
• Use a least three vdevs to provide adequate metadata redundancy, even with a striped pool.
• Provide one or more 8 GB or larger boot devices.
• Provide the TrueNAS VM with adequate RAM per its usual requirements.
• Consider jumbo frame networking if all devices support it.
• Understand that the guest tools in FreeBSD might lack features found in other guest operating systems.
• Enable MAC address spoofing on virtual interfaces and enable promiscuous mode to use VNET jail and plugins.