Claro can be configured to process one, two, three or four images simultaneously. We call this concurrent processes. The requirements are for 4 concurrent image processes.
Processors: 6 processor cores
Clock speed: >= 3 Ghz
Memory: 20 GB RAM
(Disk I/O encompasses the input/output operations on a physical disk)
Not usually found in system requirements, I/O performance can nevertheless be very important for the performance of applications, and indeed it is for Claro. On physical, dedicated machines, this typically is not an issue, although there is a considerable advantage in using SSD drives.
Note: Claro works fine in virtual environments. The performance is comparable with running Claro on physical machines IF the I/O performance is good. Which is not always the case, and also difficult to asses in a virtual environment.
Throughput / performance
Lets start with a some processing times on a macbook pro with a 2.5 Ghz Intel i7 and 16GB RAM, and an SSD harddrive, using one concurrent image process, and all typical features enabled. Keep in mind that you can process up to 4 images simultaneously in a properly dimensioned server. That means that to calculate throughput you can divide the processing time per image by approximately a factor of 3 (since there is a little overhead in processing 4 images simultaneously).
8 million-pixel image: 14 sec
12 million-pixel image: 20 sec
28 million-pixel image: 61 sec
Note: As you can see, a big factor in performance is the number of pixels of images. Claro can also resample images. It is the first step in the processing sequence. If you can downsample images based on a known output size and required resolution, the processing will typically be much faster.