beautypg.com

1 sidefile cache for hitachi truecopy asynchronous – Hitachi TAGMASTORE MK-90RD011-25 User Manual

Page 43

background image

Hitachi Command Control Interface (CCI) User and Reference Guide

25

2.4.2.1 Sidefile Cache for Hitachi TrueCopy Asynchronous

The first-in-first-out (FIFO) queue of each CT group is placed in an area of cache called the
sidefile. The sidefile is used for transferring Hitachi TrueCopy Async recordsets to the RCU.
The sidefile is not a fixed area in cache but has variable capacity for write I/Os for the
primary volume. If the host write I/O rate is high and the MCU cannot transfer the Hitachi
TrueCopy Async recordsets to the RCU fast enough, then the sidefile capacity expands
gradually. The sidefile has a threshold to control the quantity of data transfer of host side
write I/O. Host side write I/Os are controlled by delaying response when the sidefile exceeds
the constant quantity limit on cache in the storage system (see Figure 2.10).

Cache
(Total cache minus
FlashAccess)

Sidefile Area

Writing response are delayed.

Writing are waited until under threshold.

High water mark

30% of Cache

For TagmaStore USP

(0 to 70%) of Cache

the default is 40%

Sidefile Threshold =
30%

Æ 70% of Cache

Figure 2.10 Sidefile Quantity Limit

Sidefile area: Sidefile area = 30% to 70% of cache as set on Storage Navigator (or SVP)
(default sidefile = 40% for TagmaStore USP/NSC, 50% for 9900V/9900).

Write I/O control at high-water mark (HWM): When the quantity of data in sidefile reaches
30% of cache, the Hitachi TrueCopy Async pair status is HWM of PAIR state, and the host
write I/Os receive delayed response in the range of 0.5 seconds to 4 seconds. Following is an
arithmetic expression of the HWM at 100% of a sidefile space:

HWM(%) = High water mark(%) / Sidefile threshold (30 to 70) * 100

Write I/O control at sidefile threshold: When the quantity of data in sidefile reaches the
defined sidefile area, host write I/Os are delayed until there is enough sidefile space to
store the next new write data. The copy pending timeout group option is between 1 second
and 255 seconds (600 for seconds for Universal Replicator). The timeout value is defined on
Storage Navigator (or SVP) and specifies the maximum delay between the M-VOL update and
the corresponding R-VOL update. The default timeout value is 90 seconds, 60 seconds for
Universal Replicator. If the timeout occurs during this waiting state, the pair status changes
from PAIR to PSUS (sidefile full), and host write I/Os continue with updates being managed
by the cylinder bitmap. Important: The copy pending timeout value should be less than the
I/O timeout value of the host system.