Flash memory offers attractive features for storage of data, such as non-volatility, shock resistance, fast access speed, and low power consumption. However, it requires erasing before it can be overwritten. The erase operations are slow and consume comparatively a great deal of power. Furthermore, flash memory can only be erased a limited number of times. To overcome hardware limitations, we use the non-in-place update mechanism that requires a cleaner to reclaim space occupied by obsolete data. To improve cleaning performance and prolong flash memory lifetime, we propose a new data reorganization method. By this method, data in flash memory are dynamically classified and clustered together according to their accessing frequencies. Experimental results show that this clustering technique significantly improved the cleaning performance for a variety of cleaning algorithms. The number of erase operations performed is greatly reduced and flash memory lifetime is prolonged. Even wearing is ensured as well.

Blockade of the binding between Neonatal Fc receptor (FcRn) and IgG-Fc reduces circulating IgG, and thus emerges as a potential therapy for IgG-mediated autoimmune conditions. This was a double blind, randomized, single ascending dose study to evaluate the safety, pharmacokinetics (PK), and pharmacodynamics (PD) of HBM9161 (a humanized FcR monoclonal antibody) in healthy Chinese volunteers. Subjects were randomized to receive a single SC dose of HBM9161 or placebo in a 3:1 ratio in three dosing cohorts (340 mg, 510 mg, or 680 mg respectively), and then followed up for 85 days. Study endpoints included incidence of adverse event (AE), serum drug concentration, IgG and its subclasses, and anti-drug antibodies (ADA). Twenty-four subjects were randomized. Dose-dependent reduction of total IgG occurred rapidly from baseline to reach nadir at Day 11, then recovered steadily from Day 11 to Day 85. The mean maximum percentage reductions from baseline total IgG were 21.0±9.3%, 39.8±5.13% and 41.2±10.4% for subjects receiving HBM9161 340 mg, 510 mg and 680 mg, respectively. The exposure of HBM9161 (AUCs and Cmax) increased in a more than dose-proportional manner at the dose examined. All reported AEs were mild in severity. The most reported AEs in the HBM9161 groups were influenza-like illness and rash. Two subjects developed ADA during the study period. Accepted Article This article is protected by copyright. All rights reserved A single SC dose of HBM9161 results in sustained and dose-dependent IgG reduction, and was well tolerated at a dose up to 680 mg in Chinese subjects. The data warrant further investigation of its effects in IgG-mediated autoimmune disorders.

Conventional operating systems employ a kernel-controlled caching strategy that cannot properly serve all access-pattern types used by applications. When running under these systems, many memory-intensive applications with mis-matching access patterns cause excessive page faults and page replacements that reduce the application's performance. This paper presents the hipec system, which allows applications to have their own caching strategies for managing page frames with negligible overhead. Since application designers know the access patterns of their applications, the specific caching strategies can be tuned to meet the needs of each application. Empirical results show that the hipec system significantly improves application performance and system throughput.kernel selects victim applications and asks them to surrender page frames. User applications can then decide which page frames to give up in line with their specific access patterns. Previous systems 5,6,7 generally employed domain-crossing approaches to request the user-level replacement decisions. Unfortunately, domain-crossing approaches usually add significant complexity to the kernel, and create significant performance overhead for applications. Problems of domain-crossingExisting systems that employ the domain-crossing approaches generally implement asynchronous communication schemes to delegate the caching decisions to applications. 6,7,8,9 The purpose is to prevent the kernel from synchronously waiting for user-level managers' respondence. The implementation, however, added significant complexity to the kernel. The increased complexity comes from the extra flags added to the kernel-maintained data, the lengthy processing routines and the scattered checking statements. If a system wants to delegate applications, not only the page-replacement decisions but also other finer-grained caching decisions, the kernel complexity will be further increased.Performance overhead is another problem for the domain-crossing approaches. To cross domains needs to do a context switch, which needs to flush the TLB, the cache memory, to allocate user-level stacks 10 and to update many kernel-maintained data, such as the task, thread, the virtual memory management data and the kernel stack. The impact due to flushing the TLB and cache memory is not small and cannot be ignored. 11,12 Moreover, the user-level decision-making managers are subjected to scheduling delay. Other applications compete with the managers for the CPU when the managers want to make their caching decisions. This delay is unbounded, and depends only upon when the managers are scheduled to run.User applications generally need kernel-maintained data in making their specific caching decisions. The information needed includes, for instance, the referenced and modified bits of each page frame and the caching status indicating which regions of virtual memory have been cached. On the one hand, the kernel needs to provide an interface so that applications can invoke the interface to access the necessary informa...

SUMMARYConventional operating systems employ a kernel-controlled caching strategy that cannot properly serve all access-pattern types used by applications. When running under these systems, many memory-intensive applications with mis-matching access patterns cause excessive page faults and page replacements that reduce the application's performance. This paper presents the hipec system, which allows applications to have their own caching strategies for managing page frames with negligible overhead. Since application designers know the access patterns of their applications, the specific caching strategies can be tuned to meet the needs of each application. Empirical results show that the hipec system significantly improves application performance and system throughput.