JPS593530A - Input/output control device - Google Patents

Abstract

PURPOSE:To prevent partiality of service due to frequent actuation out of a single system, by providing a counter for counting the number of times of a start on each bus, and reporting a fact that an input/output control device is being used, to actuation out of a bus being in a state that the actuation is generated frequently. CONSTITUTION:When a magnetic disk control device 4 does not receive actuation not of both of CPUs 1a, 1b, a channel switch 5 is kept in a neutral point, and for instance, when actuation is applied from the CPU1a, the switch 5 is set down to the side A, and during that time, even if a actuation is applied from the CPU1b, the device 4 becomes busy. In this case, a counter for counting the number of times of a start is provided on each bus of the switch 5, and in case when a stat of a specified number of times is generated frequently in a prescribed time from the same bus, the device 4 reports that the control device is busy. In this way, frequent generation of actuation out of the same bus is inhibited, and it is prevented to occupy the device 4.

Description

[Detailed Description of the Invention] [Technical Field of the Invention] The present invention 1 relates to a magnetic disk control device shared by a plurality of paths, and is particularly provided with a mechanism to prevent services from being biased towards a single system. Concerning shared device control methods.

[Conventional technology and Makamo point]

Conventionally, when an I10 control device such as a magnetic disk control device is activated by a host, if service processing is possible at that time, the I10 control device accepts the activation unconditionally.

However, when the I10 control device such as the magnetic disk control device is shared by host CPUs with different processing capacities, and startup from the high-speed host CPU occurs frequently, it becomes difficult to start from another system (i.e., low-speed host) CPU. Despite being shared by multiple host CPUs, the service processing is significantly biased toward a single system.

[Purpose of the invention]

SUMMARY OF THE INVENTION An object of the present invention is to provide a startup acceptance control device that can accept service requests from low-speed host CPUs even when startups from high-speed host CPUs occur frequently.

[Structure of the invention]

The present invention provides a counter for counting the number of startup acceptances for each path of a channel switch of an I10 control device such as a magnetic disk control device, and if startup is received from the same path more than a certain number of times within a certain period of time, the said path For new power ups, the I10 controller hardware, such as the magnetic nice controller, automatically reports control unit busy and directs service to specific paths. It is designed to suppress the flow rate and also allow services for other thread paths to be performed.

[Embodiments of the invention]

The figure shows an example of a magnetic disk controller (DKC) shared by a hidden host CPU. When the DKC4 is not activated by either CPU (la, lb), the channel switch 5 is in neutral mode. For example, when the DKC4 is activated from the CPU 1a, the channel switch 5 is switched to the A side. falls down. And this channel
While the switch is on the A side, even if the CPU 1b starts up, it will receive W#Lf and the control unit will be busy.

When the service for the CPU 1a ends, it returns to the neutral point again. - At this time, the CPU 1a is notified of the end of the service,
Then, a report of busy release is sent to the CPU 1b, but there is a time difference of several tens of microseconds between the two. In this case, if both CPUs require service for DKC4, they will try to start up DKC4 immediately after the above report is sent, but due to the time difference, the processing speed of the two CPUs will be If they are the same, or if the CPU 1a is faster, the activation of the CPU 1a will be accepted and the service request of the CPU 1b will become control unit busy. This state will continue as long as CpUla's service requests continue at similar intervals, and the CPU 1b's service requests will not be accepted.

Additionally, the magnetic disk controller typically returns to an idle loop to process termination interrupts from other devices that have reported termination of service to one device, had been abandoned by the seek, set sector, etc. A post-processing time of 100 microseconds is required. Therefore, if the CPU on a certain host is very fast and the time from the service termination report to the next startup is shorter than the post-processing time, t=, Dr<:c means that the high-speed CPU's service continues If the idle loop continues processing, the processing that should be performed in the idle loop cannot be done at all.

In order to prevent these situations, in the present invention, a counter is provided for counting the number of activations for each path of the channel switch multiplier, and a counter is provided for counting the number of activations for each path of the channel switch multiplier, and a counter is provided for counting the number of activations for each path of the channel switch multiplier, and a counter is provided for counting the number of activations from the same path a certain number of times in a certain predetermined unit time. If it starts up too often,
Even if the DKC is in a state where it can accept booting at that time, for a new boot from the relevant path,
Control unit busy is reported, and frequent activation from the same path is suppressed to prevent the DKC from falling into the monopolized state. This operation is performed without going through the DKC firmware and is processed solely by the DKC hardware. The activation suppression continues for a certain period of time, during which time a termination interrupt from one device or activation from another path can be accepted.

E゛Inventive Effects゛According to the present invention, I10 shared by multiple hosts
It is possible to prevent services from becoming unbalanced due to frequent activation of the control device from a single system.

[Brief explanation of drawings]

The figure shows an embodiment of the invention. Further, in the figures, each number indicates the following. la, lb: central processing unit (cpU), 2a, 2b: channel device, 3a, 3b: boot count counter, 4: magnetic disk control unit 1e (nKc), 5: channel switch, 6: magnetic disk Adapter, 7: Magnetic disk device.

Claims (1)

[Claims] In an I10 control device equipped with a channel switch that is shared by a plurality of paths and has a plurality of contacts, means is provided for each path to detect that activation occurs frequently from the path, and the path in the frequent activation state is provided. 17. A 17-inch control device, characterized in that it reports that the I10 control device is currently in use in response to activation from .