IBM System Storage TS3500, Introduction And Planning Manual

Page: 112 / 290

balancing of the tiers, which stores cartridges across all HD slots in the library
string. In other words, all HD slots are filled to a minimum tier level until that tier
is full across the library. This reduces the need to access cartridges in higher tiers.
The mount performance for a TS3500 Tape Library that includes HD frames is
dependent on the library configuration and cartridge usage. For most
configurations, the impact is not noticeable. For configurations with high accessor
utilization (as measured in mounts per hour), the overall reduction in mount
performance compared to a TS3500 Tape Library without HD frames can range
from no impact to a worst-case 50% reduction in mounts per hour (for a 99% full
library in which all expansion frames are HD frames). This reduction in mount
performance is determined by the distribution of tiers that cartridges are mounted
from and the need to destage least-recently-used (LRU) cartridges from the
cartridge cache. The library configuration and cartridge usage influences the
reduction in mount performance in the following ways:
v
Each non-HD frame in the configuration increases the mounts from Tier 0
v
Unlicensed capacity decreases the mounts from the highest tiers and decreases
the time required for shuffle operations (load balancing ensures the highest tiers
are the last ones utilized)
v Lower capacity utilization decreases the mounts from the highest tiers and
decreases the time required for shuffle operations (load balancing ensures the
highest tiers are the last ones utilized)
v
A higher ratio of cartridges that are mounted more than once ( cartridge cache hits )
increases the mounts from Tiers 0 and 1
v
Cartridge eject operations prior to mount operations reduce the need to destage
LRU cartridges (this is because eject operations empty some Tier 0 slots)
Eject and insert performance can also be impacted for a TS3500 Tape Library that
includes HD frames. The impact to eject performance is determined by factors
similar to mount performance (distribution of tiers from which cartridges are
ejected). The impact to insert performance is different because inserts tend to go to
HD frames since Tier 0 slots are nominally kept full. This impact to insert
performance is determined by the following factors:
v A shorter time delay between eject and insert operations increases the inserts to
non-HD frames (this is because eject operations will empty some number of Tier
0 slots)
v
A shorter distance between the I/O stations used for inserts and the HD frames
decreases the insert move time to those frames
For more information about HD frames, see “High density technology” on page 9.
Shuttle complex performance considerations
In order to move a cartridge from one library string to another, the cartridge is first
placed in a shuttle car by the cartridge accessor. The cartridge is then moved
across the shuttle connection and is removed by the cartridge accessor at the
destination library. The time for the shuttle car to move from one library to the
next varies based on installation distances, but is as low as five seconds in optimal
configurations.
Mount performance and cartridge move time in a shuttle complex depends on the
overall configuration of the complex, including the length of the shuttle
connections and the location of the available drive. In order to maximize
performance, the application should only use remote tape drives if all drives are
88
IBM System Storage TS3500 Tape Library with ALMS: Introduction and Planning Guide