The two founders of Veeam, Ratmir Timashev and Andrei Baronov, announced their company Object First and their backup storage product in June. It’s now in beta testing, and we’ve just discovered a host of details, including its flash cache acceleration.
Anthony Cusimano, director of marketing for Object First, told us that the system delivers over 1 GB/s of throughput and uses S3-compatible object storage, with object locking for immutability, and the public cloud. for longer term retention.
We found out that it is a 2RU case with two models. 96 TB capacity with 10 x 12 TB x 7200 RPM SAS hard drives; eight for data and two for parity in a RAID-6 scheme. This can support the loss of two disks. There is a hot spare in the chassis.
The larger capacity 128TB variant has ten 16TB x 7200 rpm SAS hard drives and the same RAID-6 scheme with two pins reserved for parity plus an additional spare.
In both products, the drives are front-end by a 1.6 TB NVMe SSD. All writes arrive via 2 10GbitE ports, land on the SSD, and the data is then sent to the hard drives. It can be retrieved quickly from there, but obviously less quickly from a public cloud retention tier. Supported back-end clouds are AWS, Azure, Backblaze, and Wasabi.
As shown above, the data drives are installed flat on the front of the chassis in three rows and four columns. There are also 2 x 240GB SATA SSDs for the Linux operating system which power the system through a pair of Xeon processors. These SSDs are in a RAID-1 configuration. The object storage software was written by Object First; it is not MinIO software.
There can be four Object First nodes in a cluster, with each node adding approximately 1 GB/s of performance. Four is an arbitrary cluster limit, put in place while the product is in beta testing. Cusimano expects the limit to be increased, possibly significantly in the future.
We note that there is also disk capacity headroom with 16TB, 18TB and 20TB drives already available from vendors. 20 TB suggests a maximum node capacity of 200 TB raw and, at 20% overload, a usable capacity of 160 TB.
Pricing hasn’t been set, but we’re told the system will be affordable. It’s also simple to use with four basic commands; configure the S3 bucket, write to it, read from it, and configure an S3 key.
This Object First box appears to be simple to use, simple to deploy, quick to use, and hopefully affordable to purchase. Deduplication and compression are obvious development additions, increasing the effective capacity two, three times or more, and it might require more NVMe SSD capacity and CPU power. Any reduction in data would lower the price per terabyte.
The Object First system is designed to be a Veeam backup target in this iteration. Another software development product could be to expand the range of supported backup software. We will follow with great interest the evolution of all this.