SCv2020 - Architecture

The SCv is a scaled down version of the full SC series storage. SC storage uses virtualized storage pools to allocate blocks for LUNS. Within those LUNS multiple RAID levels are used simultaneously to provide the desired mix of performance and storage efficicency. For example, by default, writes will be processed as RAID10 (fastest) and then converted to RAID5/6 during data progression (most storage efficicent and no read penalty). SC systems speed is derived from the type and number of disks in the system. A system with 11 disks is going to generate 600 - 1600 IOPS depending on the type of disk installed.

There is a specific limit to the number of mappings in the iSCSI world, but I cant find anything concrete on a limit of server entities or paths in the FC world. Anyone else know?  (With 11 disks your performance will be exhausted far before the number of servers)

The SCOS version release notes will have the scalabililty guidelines for each system. The SCv2xxx maximum code version today is 6.6.x. In that guide it has the following for the "maximum recommended design guidelines" for servers:

"100 servers (Total number of servers includes physical servers, virtual

servers, and server clusters.)

• 100 server folders

• 200 server HBA initiator ports (Total number of HBA initiator ports

includes all physical and virtual ports that are visible to the SAN.)"

Other than that, I echo Michael's response.

Given the nature of SSD and how it reacts to different workloads, I use between 3K and 5K IOPS per disk to estimate  with the system max being the limit. Remember the spare doesnt count in I/O.