iXsystems – most well-known as a supplier of network-attached storage (NAS) software and entry-level products – has taken its hyper-converged TrueNAS Scale into general availability.
iXsystems uses open source operating systems and commodity hardware, delivering hyper-converged products that can be half the price of those from Dell EMC or HPE. However, the key aim for iXsystems has been to become competitive by adding scale-out capability.
“TrueNAS Scale brings the ability to run applications in the form of KVM virtual machines or in Kubernetes containers,” said iXsystems director general Brett Davis in an interview with Computer Weekly sister publication LeMagIT.
“But we consider these use cases to be at the margins. Scale is above all about being able to extend our storage capacity at will. Our real goal is to compete with super-elastic scale-out NAS products such as Dell EMC’s Isilon,” he said.
“Our high-end TrueNAS M60 supports a maximum capacity of 20PB [petabytes] when you equip it with 12 TrueNAS ES102 shelves with 102 disks each. Which is fine, but our customers want to go further,” said Davis.
“Replacing TrueNAS in the M60 with TrueNAS Scale means you can now deploy it in clusters. We’ve tested a cluster of 100 nodes with 2,000PB capacity.”
Davis is optimistic. During the pandemic, iXsystems sales grew by 70%, notably in Europe and Asia. Meanwhile, in New York, one bank customer has replaced its Pure Storage FlashBlade NAS with TrueNAS M60 arrays because it offers 20PB of capacity for half the cost.
TrueNAS Scale sees a shift to FreeBSD
Historically, iXsystems arrays have been built on x86 servers and used a version of the Linux FreeBSD operating system with the storage file system OpenZFS. TrueNAS Scale replaces FreeBSD with a Debian Linux operating system (OS).
“We have gained the ability to run virtual machines under the KVM hypervisor and to automatically group OpenZFS volumes in a global pool using GlusterFS as a distributed file system,” said Davis.
Brett Davis, iXsystems
Gluster is also open source. Sold commercially for a long time by Red Hat, it was recently dropped as the supplier favoured Ceph. “No one, not even Oracle, supports ZFS now,” said Davis. “That doesn’t scare us, though. It’s a solid storage system, still actively developed by the community, and we are convinced it is right for the challenges that we want to address.”
TrueNAS, historically, exists as a freely installable software product for any x86 hardware (now called TrueNAS Core, previously FreeNAS). It’s probable that TrueNAS Scale will soon be also available as a free download, but the iXsystems team isn’t very talkative about this.
TrueNAS Core has the same storage functionality (SMB, NFS, S3 connectivity with iSCSI to additional disk shelves) from the same TrueCommand console and the same security functionality as the commercial TrueNAS Enterprise and the new TrueNAS Scale. But it lacks 24/7 support, said Davis, and customers tend to use it to test in small configurations.
iXsystems envisages following the model of Nutanix, in which it provides the system software and third-party suppliers sell products based on their hardware selection.
M series top of the range
TrueNAS Scale can potentially work with all iXsystems arrays. But the M range, with lots of network connectors, is the most well-suited to large storage clusters.
Products in iXsystems’ M range are all 4U and can hold 28 solid-state drives (SSDs) or 24 hard-disk drives (HDDs). The differences between models are in the controller. The M60, for example, has 64 CPU cores, 1.5TB (terabytes) of RAM, eight 100Gbps Ethernet ports, and supports up to 12 disk shelves with 102 drives each.
Meanwhile, the M50 runs 40 cores, has 768GB of RAM and four Ethernet ports, with eight ES102 add-on shelves supported.
By the end of the year, iXsystems will launch the entry-level M30, with 16 cores, 128GB of RAM, four 25Gbps Ethernet ports and support for two ES102 disk shelves. Performance attained includes throughput of 3GBps and 25,000 IOPS.
The idea behind the M30 is that while having little impact on capacity it can put more controller nodes into the cluster than disk shelves. That could be interesting in hyper-converged infrastructure (HCI) use cases where there is a need for more processing power rather than capacity. For example, when numerous containers run an application, it might still be from the same dataset.
Entry-level R series
Of more modest performance is the R range, with fewer cores and limited numbers of 100Gbps ports. They present the advantage of being less costly and perhaps being well-suited to deployment at smaller sites. There are three R range models – the R10, R40 and R50 – which go from 90TB capacity with up to 30 cores and 288GB of RAM to 8.6PB, which includes some NVMe flash.
Even smaller, TrueNAS Mini arrays are intended for deployment in offices and come in four models, from the Mini E to XL, with up to eight HDDs.
Coming next: Active-active nodes and standalone TrueNAS Scale
iXsystems has also previewed enhancements for its hyper-converged infrastructure that will be enabled in the coming year.
From this week, machines equipped with TrueNAS Scale can be used in a cluster of at least three nodes, but with only one of them managing access.
Then, towards the middle of 2022, iXsystems will add the ability to function in active-active mode in which each node will be able to manage access from the network. In this configuration, the two controllers of an M series array will each manage their own OpenZFS volumes to maximise speed of access.
At the start of 2022, iXsystems will launch TrueNAS Scale standalone nodes, with the capacity to run containers and virtual machines without the need to deploy an entire cluster. In this configuration the two controllers provide redundancy.
Read more on hyper-converged infrastructure
- Five reasons to look at hyper-converged infrastructure. HCI promises easy setup, purchasing flexibility, low procurement and operation costs, reliability and all in one box – but there are challenges too.
- Five hyper-converged infrastructure use cases to watch. We look at workloads ripe for hyper-converged infrastructure deployment. These include virtual servers and desktops, SMEs/remote offices, Kubernetes, and analytics and machine learning.