摘要:
In a digital computer, a circular queue of registers in a register file are allocated as temporary local storage for procedures rather than using the known caller/callee save convention in order to minimize main memory references. A called procedure dynamically allocates local registers as needed without regard to registers used by the caller of the procedure or by any callee of the procedure, whereby register allocation is not restricted by any predetermined window size. Local registers, including parameter passing registers, are allocated in the called procedure, rather than a priori at compile time, by adjusting register stack pointer values. Only the number of registers actually required by the procedure need by allocated. Optionally, rotating registers may be allocated among the local registers. Stack pointer values are stored in one of the parameter passing registers when a procedure is called. Hardware register file access circuitry maps virtual register numbers used by the procedures into the hardware register file. Upon return from a procedure, registers are deallocated by adjusting the register stack pointers to the values stored when the procedure was called.
摘要:
A load balancer allocates requests to a pool of web servers configured to have low queue capacities. If the queue capacity of a web server is reached, the web server responds to an additional request with a rejection notification to the load balancer, which enables the load balancer to quickly send the rejected request to another web server. Each web server self-monitors its rejection rate. If the rejection rate exceeds a threshold, the number of processes concurrently running on the web server is increased. If the rejection rate falls below a threshold, the number of processes concurrently running on the web server is decreased.
摘要:
A method for executing uniprocessor (UP) coded workloads in a computer capable of concurrent thread execution is disclosed. The method identifies threads in the uniprocessor coded workloads (UP-workloads) which can execute concurrently, and identifies threads in the UP-workloads which cannot execute concurrently. First threads which cannot execute concurrently are assigned to a first concurrency group. Second threads which cannot execute concurrently are assigned to a second concurrency group. Any thread in the first concurrency group can execute concurrently with any thread in the second concurrency group. The computer capable of concurrent thread execution then executes the UP-coded workloads in the first concurrency group at substantially the same time as executing the UP-coded workloads in the second concurrency group.
摘要:
A file system layout apportions an underlying physical volume into one or more virtual volumes (vvols) of a storage system. The underlying physical volume is an aggregate comprising one or more groups of disks, such as RAID groups, of the storage system. The aggregate has its own physical volume block number (pvbn) space and maintains metadata, such as block allocation structures, within that pvbn space. Each vvol has its own virtual volume block number (vvbn) space and maintains metadata, such as block allocation structures, within that vvbn space. Notably, the block allocation structures of a vvol are sized to the vvol, and not to the underlying aggregate, to thereby allow operations that manage data served by the storage system (e.g., snapshot operations) to efficiently work over the vvols.
摘要:
A system and method enhances performance of updates to sequential block storage of a storage system. According to an aspect of the invention, a disk array of the storage system is utilized to extend write buffers of the system, thereby rendering a portion of the disk array a disk-based log. To that end, one portion of the disk array is organized into a home location array having a set of sequential home locations for disk blocks. Another portion of the disk array is organized into the disk-based log having a set of log buffers configured to store versions of disk blocks that have yet to be returned to their home locations in the home location array. In addition, non-volatile memory of the storage system is organized as an index configured to provide efficient mappings of disk blocks not yet returned to their home locations. In accordance with another aspect of the invention, a novel disk-based multi-level sort procedure is provided to establish locality among updates (write data) held in the disk-based log, thereby enabling the write data to be efficiently written to home locations on the disk array.
摘要:
The present invention provides a system, method, and computer program product for reducing load imbalance in a storage system having a plurality of storage devices organized in one or more RAIDs for storing data by moving data from heavily-loaded storage devices to less-loaded storage devices during normal data access operations. As a result of moving data to less-loaded storage devices, the service latency of those storage devices decreases, thereby optimizing the system's performance.
摘要:
A technique places content, such as data, of one or more data containers on volumes of a striped volume set (SVS). The placement of data across the volumes of the SVS allows specification of a deterministic pattern of fixed length. That is, the pattern determines a placement of data of a data container that is striped among the volumes of the SVS. The placement pattern is such that the stripes are distributed exactly or nearly equally among the volumes and that, within any local span of a small multiple of the number of volumes, the stripes are distributed nearly equally among the volumes. The placement pattern is also substantially similar for a plurality of SVSs having different numbers of volumes.
摘要:
An improved system and method enhances performance of updates to sequential block storage of a storage system. A disk-based sort procedure is provided to establish locality among updates (write data) held in a disk-based log, thereby enabling the write data to be efficiently written to home locations on a home location array. As the write data is received, a log manager of the storage system temporarily stores the data efficiently on the disk-based log. As more write data arrives, the log manager sorts the data in the log in accordance with the sort procedure, thus increasing the locality of data when stored on the home location array. When the log approaches capacity, the log manager writes the sorted data to their home locations on the array with high locality and performance.
摘要:
A method for transferring data of a hybrid virtual volume of a computer data storage system from a source to a destination is disclosed. The method first translates intermingled virtual and physical volume block numbers of the hybrid virtual volume into a data stream having only virtual volume block numbers. The method then sends the data stream to a destination computer.
摘要:
A system and a method of scheduling a plurality of threads from a multi-threaded program. A shared arena is provided in user memory, wherein the shared arena includes a register save area for each of the plurality of threads. A processor, when allocated to the application, executes the application's user-level scheduler and selects a user-level thread from a plurality of available threads, wherein the step of selecting includes the step of reading register context associated with the selected thread from one of the plurality of register save areas. In multikernel systems, kernels having access to an application's register save areas can execute preempted threads from that application with no kernel-to-kernel communication. Likewise, kernels having access to an application's user-level run queues can execute ready-to-run threads from that application with no kernel-to-kernel communication.