Sending Enough Concurrent Mount and Dismount Requests
Some client applications are designed for simple SCSI Media Changer libraries, which only process a single request at a time. However, ACSLS and SL8500 and SL3000 libraries, are designed to process many requests at the same time. Each library can have at least 40-50 concurrent requests in its queue, and the ACSLS queue is essentially unlimited.
A client can send ACSLS hundreds of concurrent mount and dismount requests. To maximize library performance, the client(s) should send ACSLS enough requests to keep all of the robots, and so forth in the library, busy.
If a client can only send ACSLS a limited number of concurrent requests, you can attach multiple clients to ACSLS. For example, if a particular client application only sends ACSLS a single mount or dismount request at a time, by connecting 12-16 client applications per SL8500 to ACSLS, you can send ACSLS enough concurrent mount and dismount requests to keep all of the robots in an SL8500 busy.
Here are the details to understand how many concurrent requests may be needed to keep of the library's robots busy:
-
The library does not return the mount response to ACSLS until the drive is loaded and ready. In this way, ACSLS knows whether the mount was successful.
-
In a smaller SL8500, with about 3000 storage cells, it takes the robot about 10-15 seconds to move a cartridge to a drive.
-
However, it takes an LTO drive about 19 seconds to load the cartridge and come ready.
-
To keep even one robot on a rail busy, you need at least three overlapping mount requests on that rail.
This lets the robot(s) process other mounts while the first drive receiving a cartridge is loading and coming ready. With two robots per rail, you may want to have at least four concurrent mounts or dismounts per rail.
-
An SL8500 has four rails, so you need 12-16 concurrent mount and dismount requests per SL8500 to keep all of the robots in an SL8500 busy.
All of these concurrent requests could come from a single ACSAPI client, or they could come from 12-16 different clients, each of which only issues a single request at a time, and waits for the response before issuing the next request.