SupermicroAS-2015CS-TNR-01-G2

Eight 24 TB SATA drives, two 3.8 TB NVMe and a 48-core EPYC 9455P with 768 GB of memory in 2U, with enough processor headroom for the same node to store data and run virtual machines.

Supermicro AS-2015CS-TNR-01-G2Rack 2U · Gold Series G2

The exact configuration

Official datasheet

Front view diagram: hover over each component.
Processor1× AMD EPYC 9455P48 cores · 3.15 GHz
Memory768 GBDDR5-6400
Storage2× 240 GB 2.5" SATA SSD8× 24 TB 3.5" SATA HDD + 2× 3.8 TB U.2 NVMe
Network1× dual 25 GbE SFP28
Form factorRack 2U
Warranty
3 years parts and labour
On-site engineer
Optional: next business day, for 3 years

Storage and virtualisation on the same node

Supermicro positions it as storage hardware for private clouds, and the parts list explains why. The 8 × 24 TB give 192 TB raw in a chassis with twelve 3.5-inch bays, and the two 3.8 TB U.2 NVMe drives (the 2.5-inch server form factor) do the thing that most improves a spinning-disk Ceph cluster: they hold each drive's internal database (what Ceph calls DB/WAL) on flash, four drives per NVMe, so small writes do not wait for the disk head. With 48 cores and 768 GB there is plenty of machine for the eight Ceph daemons, and whatever is left over can run virtual machines on the same node with Proxmox.

With 3-way replication each node contributes 64 TB usable and a three-node cluster ends up at 192 TB; with 4+2 erasure coding and six nodes you reach 768 TB usable out of 1,152 raw. Networking is 2 × 25 GbE, which keeps up with eight spinning disks but falls short if the cluster moves to all-NVMe, and for that day the chassis has four PCIe 5.0 x16 slots.

What we do

We set it up as a Proxmox VE node with built-in Ceph, the most direct way to get virtualisation and shared storage without a separate array, or as a node in a dedicated Ceph cluster if the virtual machines run on other servers. The OS is mirrored across the two 240 GB SSDs, and we hand it over with the BMC (the remote management controller), firmware and the split between VM traffic and replication traffic already in place. It fits as a node in a VMware to Proxmox migration.

Compared with the rest of the family

ModelWhat it hasWhen to choose it
ASG-2015S-E1CR24L-01-G21x AMD EPYC 9334 · 512 GBThe most capacity per U, for archive, backup and Ceph.

See all storage servers

What the delivery includes

From the factory

  • Configuration assembled and tested by Supermicro
  • Warranty: 3 years parts and labour

From us

  • Rack mounting
  • Remote management (BMC) and up-to-date firmware
  • Operating system and the stack that runs on top
  • Testing with your workload before it goes into production

If you need it

  • 24×7 on-call support with SLA
  • On-site engineer next business day for the full 3 years

How we build it

What runs on top of these servers, explained in detail:

Support and training for your team

Questions about the AS-2015CS-TNR-01-G2

What are the two NVMe drives for if there are already 24 TB disks?

In Ceph, they keep the database and write-ahead log of each spinning disk (DB/WAL) on flash, at four disks per NVMe. It is the change that most improves latency on a spinning-disk cluster. Without Ceph, they work as a separate fast volume for virtual machine disks.

Why 48 cores in a storage server?

Because it is designed for the node to do two jobs. Eight Ceph daemons take up a small share of an EPYC 9455P and 768 GB, and the rest is available for virtual machines in a hyperconverged cluster, where every server stores and runs workloads at the same time. If you only want capacity, the ASG-2015S-E1CR24L-01-G2 gives almost three times the terabytes per chassis.

How many do I need for a Ceph cluster?

Three with 3-way replication, leaving 192 TB usable out of 576 raw, or six with 4+2 erasure coding, leaving 768 TB usable out of 1,152. In both cases we recommend one node above the minimum, so the cluster can rebuild onto the others when one goes down.

Is the AS-2015CS-TNR-01-G2 right for what you're building?

Tell us about the workload and the data centre it's going into, and we'll come back with the price, a confirmed lead time and what it would take to get it running.