When one of our devices crashes or shuts down because of memory issues, one of the last words we would ever use to describe the circumstances is "graceful."
Yes, but graceful is indeed part of the title of a just-published BlackBerry Patent application entitled Detection Of Out-Of-Memory and Graceful Shutdown.
Interestingly, if you read through the Patent app's Abstract and literature, this invention seems to be more oriented toward stand-alone and networked PCs rather than BlackBerrys.
The Abstract describes:
A method, system and computer readable medium for managing low memory in a first computing device are provided. The system is configured to cause part of the memory allocated to a specialized application to be held in reserve so that it can be used to support the specialized application during an occurrence of low memory, thus providing time for data backup or remedial steps to be carried out before the affected application crashes.
Reading further in this application's literature, we note language that directly implicates PC operating systems as vulnerable to lockups that could be forewarned by the technology that is being described here:
As known in the art, operating systems such Microsoft Windows and UNIX based systems have a physical limit on the amount of addressable memory that can be used by an application. If the application exceeds this limit, it will often crash, leading to a loss of data that has not been stored to persistent memory, and perhaps also to crashes or slowdowns in any other applications that rely on the crashed application.
In distributed computing networks in which separate computers connected to a common network perform work in parallel, one computer may be configured to monitor its own memory usage or the memory usage of other computers and take precautionary backup measures, either automatically or after alerting a human operator, in the event that the monitored conditions of one of the computers indicates that an application crash has occurred or is about to occur.
However, such a system requires almost near constant polling of the monitored computers in order to have timely detection of a problem, and even with timely detection, the affected computer application may crash before critical data can be saved.
Accordingly, a method and system for detecting and managing low memory situations that provides time for rectifying or otherwise responding to the situation is desirable.
OK getting serious with you again. I've read through the remainder of this literature. Having done so, I think Figure 4, and its accompanying documentation, draws us closest to an understanding of what is being proposed here.
FIG. 4 illustrates, according to embodiments of the present disclosure, a process indicated generally by reference 400, taken by low memory manager 146 to manage low memory conditions of a specialized application 144i (where 1.ltoreq.i.ltoreq.N).
As indicated in step 402, in embodiments where the low memory manager is a stand alone application or integrated into the operating system 142, it monitors all specialized applications that are started by the operating system 142 and determines if the specialized application being started (in this example, application 144i) has been pre-selected for low memory management, and if so performs the memory management functions set out below.
In some embodiments, all specialized applications may be selected for low memory management, in which case step 402 may be omitted. In some embodiments, the low memory manager 146 may be a module of specialized application 144i, and be configured to automatically perform its memory management steps only in respect of its associated specialized application 144i.
If low memory management is to be performed in respect of specialized application 144i, then, as indicated in step 404, the low memory manager 146 instructs the operating system 142 to allocate reserve memory blocks for the specialized application 144i.
As noted above, when the operating system 142 first starts specialized application 144i, it books or allocates a block of memory 152i (FIG. 3) for the application.
In a Windows operating system embodiment, the memory block may be 2 GB, which is the maximum amount of memory that Windows can presently allocate to an application.
According to embodiments of the present disclosure, in step 404 the low memory manager 146 requests the operating system 142 to book or allocate two reserve sub-blocks within memory block 152i such that it is broken up into three sub-blocks, namely a main block 158, and two reserve buffers B1 154 and B2 156.
The main block 158 is made fully available for the normal operating requirements of specialized application 144i, but, from the perspective of both the operating system 142 and the specialized application 144i, the two reserve buffers B1 154 and B2 156 are booked up and unavailable until freed by the low memory manager 146 (as described below).
As known in the art, the operating system 142 will typically divide memory into "pages" of predetermined size. Although low memory manager 146 books the reserve buffers B1 154 and B2 154 upon start up of the specialized application 144i, the reserve buffers are, in one example embodiment, not actually physically divided up into memory pages by the operating system until data is actually written to them, thereby reducing unnecessary processor activity.
In one embodiment, the reserve buffers B1 154 and B2 156 are relatively small compared to main block 158, with the second reserve buffer B2 156 being smaller than the first buffer B1 154. By way of non-limiting example, in one embodiment, the main block 158 may be approximately 88% of the memory block 152i that has been allocated for application 144i, the first reserve buffer B1 154 approximately 10% of the memory block 152i, and the second reserve buffer B22 156 approximately 2% of the memory block 152i. However, in various embodiments different relative buffer sizes are used, and in some embodiments the first and second reserve buffers are identical in size.
As specialized application 144i runs, it will use the main memory block 158 for its memory requirements.
Application 144i and operating system 142 may employ various memory management techniques known in the art to manage usage of the memory within main block 158, such as caching data to hard drive 134.
As known in the art, operating system 142 monitors memory usage of specialized application 144i, and is configured to issue an out-of-memory exception in the event that the available memory for specialized application 144i falls below a predetermined threshold (which may be zero available memory). In many specialized applications, the out-of memory exception results in an immediate or almost immediate crash of the specialized application, without sufficient warning for corrective or remedial action to be executed.
The low memory manager 146 of the present disclosure provides a method by which warning of an impending out-of-memory problem can be provided sufficiently early to allow corrective or remedial action to be taken before the crash of an application.
More particularly, as indicated in step 406, the low memory manager 146 is configured to detect, as long as the specialized application 144i is running, any out-of-memory exceptions issued by the operating system 142 in respect of specialized application 144i. An out-of-memory exception will result when the available memory in main block 158 drops below the predetermined threshold.
As indicated at step 408, upon detecting an out-of-memory exception issued by the operating system in respect of specialized application 144i, the low memory manager 146 instructs the operating system 142 to release the first reserve buffer B1 154, effectively integrating the memory of buffer B1 with the main block 158, thereby increasing, from the perspective of the specialized application 144i and the operating system 142, the amount of free memory available for use by the specialized application.
The newly freed memory will in many cases delay, if not prevent altogether, an impending crash.
In addition to freeing the memory the low memory manager may, as indicated in step 410, also issue a first warning alert so that corrective or remedial action can be taken. In some embodiments, the first warning alert could be an audio and/or visual signal intended to alert a human operator of the condition.
Additionally or alternatively, the first warning alert could be an electronic signal or a variable passed to a management application that is running on the affected computer. The management application could be a discrete specialized application, or could be a module of the affected specialized application, or a module of the low memory manager 146.
In some embodiments, the first warning signal could be sent to a management application running on a remote computer to which the affected computer is connected by a network. For example, in an implementation where the low memory manager 146 is running on a wireless transport interface 128.sub.k of distributed gateway 100, the first warning signal could be sent to a management application running on controller 122, which in turn may issue an alert for a human operator to investigate the problem.
The management application on controller 122 could be configured to take corrective action, either automatically, or with the intervention of a human operator. For example, the management application on the controller 122 is in some embodiments configured to cause electronic traffic to be rerouted to other wireless transport interfaces 128 to relieve some of the demand on the affected wireless transport interface.
In some embodiments, steps may be taken after the first warning signal to store information in memory 152i to a persistent storage.
Turning again to process 400, once the reserve buffer B1 154 is released, the low memory manager 146 waits to see if the operating system 142 issues a second out-of-memory exception in respect of specialized application 144i (step 412).
While waiting for a second out-of-memory exception, the specialized application periodically attempts to re-reserve the first reserve buffer B1 154 (step 414)--if the memory usage level by the specialized application 144i drops down to a level that does not require reserve buffer B1 154 anymore, then the operating system 142 will allow the memory to be re-reserved by the low memory manager 146, otherwise the operating system 142 will not allow the memory to be released to the low memory manager 146.
The frequency of the attempts by low memory manager 146 to re-reserve first buffer B1 154 is configurable in one embodiment of the present disclosure. In the event that the reserve buffer B1 154 is no longer needed and is re-reserved by the low memory manager 146 in step 414, the low memory manager 146 resets and returns to step 406 to wait for a new first occurrence of an out-of-memory exception.
Turning again to step 412, the operating system 142 will issue a second out-of-memory exception when the additional memory that came available due to release of the reserve buffer B1 154 is full.
The occurrence of a second out-of-memory exception will typically indicate that any corrective action that was taken in response to the first warning alert has failed, or that the first warning alert has gone unheeded.
As indicated in step 416, in response to a second out-of-memory exception, the low memory manager 146 releases the second reserve buffer B2 156, effectively integrating the memory of buffer B2 with the main block 158 and the previously released first reserve buffer B1 154, thereby increasing, from the perspective of the specialized application 144i and the operating system 142, the amount of free memory available for use by the specialized application 144i.
The newly released memory B2 156 preferably buys sufficient time for recovery supporting activity to take place prior to crashing of the affected application 144i--for example, time for selected information in the memory block 152i to be stored to a persistent storage location so that it can be retrieved at a later time.
As indicated in step 418, upon releasing the second reserve buffer, the low memory manager preferably issues a second warning alert to signal that a crash is pending. In some embodiments, the second warning alert is effectively a command to commence an automated shutdown of the specialized application 144i.
Preferably, the second warning alert triggers recovery supporting activities to take place. In some embodiments, the second warning alert could be a signal or variable passed to another application or within an application located on the same computer as the affected application. Alternatively, or additionally, the second warning alert could be sent to a remotely located management application on another computer.
By way of non-limiting example, the affected specialized application 144i may be running on wireless transport interface 128.sub.k and responsible for sending outgoing messages to and receiving incoming messages from mobile devices 106.
The specialized application maintains a pending message buffer within memory block 152i that includes pending outgoing messages that are waiting to be sent out to mobile devices 106, and pending incoming messages that have been received from mobile devices 106 but not yet processed by the gateway 100.
The pending messages may include, among other things, status messages such as error codes and acknowledgements, and conventional email messages.
Delivery of some of the pending messages in the pending message buffer may be critical in order to avoid "lost" messages, namely messages in which an originating device such as a mobile device 106 is left with incorrect or ambiguous information as to the status of an email message that was sent from the device.
Accordingly, in the event of an impeding crash of the specialized application 144i, it is desirable to serialize to a persistent storage at least the critical pending messages in memory block 152i that are required to ultimately be sent to avoid lost messages.
In the presently described example, the specialized application 144i includes a shutdown manager 160 that is configured to receive the second warning alert issued in step 418, and subsequently begin a shutdown routine that includes, among other things, serializing to persistent storage any critical messages that are stored in memory block 152i, thereby allowing the critical pending messages to be retrieved and sent at a later time.
The messages could be serialized to a local persistent storage of the computer on which the affected application is running, or to database 124. The second warning alert issued in step 418 is preferably also provided, directly or indirectly, to the controller 132 so that it can cause all traffic to be rerouted away from the wireless transport interface 128.sub.k that is running the crashing application 144i, and designate one or more other wireless transport interfaces 128 to take over the traffic and also to retrieve and send the serialized critical messages.
It will thus be appreciated that the low memory management process of the present disclosure provides advanced warning of impending crash of an application due to an out-of-memory situation, thus allowing corrective actions to be taken to try and prevent the crash and also recovery supporting actions to be taken so that critical data is not lost if a crash occurs.
In the example embodiment described above, two reserve buffers are reserved from the memory available to the application. When an out-of-memory exception is issued, the first reserve buffer is released, and a warning issued so that either or both automated and human controlled investigations and corrective actions can occur.
If the warning does not result in a successful fix, then a second out-of-memory exception is received, causing the second buffer to be released, and the commencement of a controlled shutdown of the application during which critical data is stored to persistent memory.
The low memory manager 146 relies on out-of-memory exceptions issued by the operating system 142, rather than attempting to directly and continuously monitor memory usage itself. As a result, low memory manager 146 uses relatively few processing resources of the computer upon which it is implemented.
In some embodiments, more or less than two reserve buffers may be used.
In one embodiment, there is provided a computer program product having a computer-readable medium tangibly embodying computer executable instructions for implementing embodiments of the present disclosure described above. The computer readable medium could, among other things be a storage medium such as a magnetic medium or an optical medium, or could be a communications medium such as an electrical or optical signal onto which the computer executable instructions have been modulated.