Method to automatically configure storage cards in a redundant and load-balanced way
Publication Date: 2018-Jun-29
The IP.com Prior Art Database
Mainframe computer systems can be equipped with large numbers of I/O cards, state-
of-the-art mainframe embodiment’s support up to 160 I/O cards. A subset of these
I/O cards, in the following referred to as storage cards, are used to attach enterprise
storage systems via fiber-channel networks. Storage cards offer two modes of
operation (in the remainder of this publication referred to as MODE2 and MODE2),
and have to be configured into one of these modes before they become operational.
In mainframe computing, reliability and performance are of crucial importance. This
includes the I/O connectivity from the mainframe to storage systems. Selecting
storage cards such they best address reliability (by providing redundant
communication paths) and performance (by balancing load across the available I/O
hardware infrastructure) is tedious and error prone. Yet, as of today, the only way to
configure the mode of operation of storage cards is to write configuration files
manually, or to use only very basic tools.
Disclosed is a method to automatically configure storage cards in a redundant and load-
The basic building block of the disclosure is the selection of a redundant pair of storage
cards. This pair will be configured to use the desired mode of operation. Derived
from this building block, the invention also solves the problem of configuring a single
card, as well as configuring an arbitrary number of cards.
The pair selection building block makes use of two concepts to be introduced in the
following: the balance score and the redundancy score.
Based on the plug locations of a storage cards, cards can be grouped into various
categories. Possible categories include: the I/O drawer, the I/O domain, the card's
configuration status and combinations thereof. The balance score is better the more
evenly cards are distributed across all categories.
The redundancy score is defined for a pair of cards and reflects how well one card can
act as fallback of the other card of the pair in case of failure scenarios, taking into
respect physical and architectural properties of the system. A pair of cards plugged
into two different I/O drawers is considered to have the best redundancy score,
followed by a pair of cards plugged into the same drawer, but in redundant I/O
domains each. Next would be a pair of cards plugged into the same drawer, but
arbitrary (different) I/O domains. The least redundancy score has a pair of cards
plugged into a single drawer and into the same I/O domain, which still offers
redundancy against card failures.
The Pair Selection Building Block
Pair selection runs in three phases:
A) Pair Selection
1) Identification of eligible storage cards (see below for details).
2) Selection of a card that maximizes the balance score of the overall configuration.
This is used as the first card of a pair (see below for details).
3) Selection of a second card, by maximizing the redundancy score of the pair of
cards resulting from step 2 and...