Mobile cloud computing (MCC) enables the mobile devices to offload their applications and thus greatly enriches types of on enhances quality service applications. Under various circumstances, researchers have put forward several MCC architectures. However, how reduce response latency while efficiently utilizing idle capacities still remains a challenge. In this paper, we firstly give definition divide recently proposed architectures into four categories. Secondly, present Hybrid Local Cloud...

Our extensive experiments reveal that existing key-value stores (KVSs) achieve high performance at the expense of a huge memory footprint is often impractical or unacceptable. Even with emerging ultra-fast byte-addressable persistent (PM), KVSs fall far short delivering promised by PM's superior I/O bandwidth. To find root causes and bridge performance/memory-footprint gap, we revisit architectural features two representative indexing mechanisms (single-stage multi-stage) propose three-stage...

Hybrid storage servers combining solid-state drives (SSDs) and hard-drive disks (HDDs) provide cost-effectiveness μs-level responsiveness for applications. However, observations from cloud system Pangu manifest that HDDs are often underutilized while SSDs overused, especially under intensive writes. It leads to fast wear-out high tail latency SSDs. On the other hand, our experimental study reveals a series of sequential continuous writes exhibit periodic, staircase-shaped pattern write...

Cloud providers like the Alibaba cloud routinely and widely employ hybrid storage nodes composed of solid-state drives (SSDs) hard disk (HDDs), reaping their respective benefits: performance from SSD capacity HDD. These generally write incoming data to its SSDs then flush them HDD counterparts, referred as Write Back (SWB) mode, thereby ensuring low latency. When comprehensively analyzing real production workloads Pangu, a large-scale platform underlying cloud, we find that (1) there exist...

Mobile Cloud Computing (MCC) enables the mobile devices to offload their applications cloud, and thus greatly enriches types of on enhances quality service applications. Under various circumstances, researchers have put forward several MCC architectures. However, how reduce response latency while efficiently utilizing idle capacities is still a challenge. This paper firstly gives definition MCC. Secondly, by extending Cloudlet architecture, Hybrid Local Model (HLMCM) presented. Thirdly, from...

Attempts to improve the performance of key-value stores (KVS) by replacing slow Hard Disk Drives (HDDs) with much faster Solid-State (SSDs) have consistently fallen short gains implied large speed gap between SSDs and HDDs, especially for small KV items. We experimentally holistically explore root causes inefficiency existing LSM-tree based KVSs running on powerful modern hardware multicore processors fast SSDs. Our findings reveal that global write-ahead-logging (WAL) index-updating...

Parity-based RAID suffering from partial-stripe write-penalty has to introduce write buffer fast absorb and merge incoming writes, then flush them array in batch. However, we experimentally observe that the popular buffering mechanism as Linux journal partial parity logging (PPL) becomes a bottleneck for ultra-fast SSD-based RAID, further uncover centralized log-buffer model is prime cause.

Following a conventional design principle that pays more fast-CPU-cycles for fewer slow-I/Os, popular software storage architecture Linux Multiple-Disk (MD) parity-based RAID (e.g., RAID5 and RAID6) assigns one or centralized worker threads to efficiently process all user requests based on multi-stage asynchronous control global data structures, successfully exploiting characteristics of slow devices, e.g., Hard Disk Drives (HDDs). However, we observe that, with high-performance NVMe-based...

The application of product-level persistent memory (PM) presents a great opportunity for key-value stores. However, PM devices differ significantly from traditional block-based storage such as HDD and SSD in terms IO characteristics approaches. To reveal the adaptability existing store on to explore potential optimization space PM-based stores, we migrate one most widely used store, RocksDB, device evaluated its performance. results show that performance RocksDB is limited by stacks...

Key-value (KV) stores based on multi-stage structures are widely deployed to organize massive amounts of easily searchable user data. However, current KV storage systems inevitably sacrifice at least one the performance objectives, such as write, read, space efficiency etc., for optimization others. To understand root cause and ultimately remove disparities among representative existing stores, we analyze their enabling mechanisms classify them into two fundamental models data facilitating...

Emerging persistent memory (PM) with fast per-sistence and byte-addressability physically shares the channel DRAM-based main memory. We experimentally uncover that throughput of application accessing DRAM collapses when multiple threads access PM due to head-of-line blockage in controller within CPU. To address this problem, we design a PM-Accessing Thread Scheduling (PATS) mechanism is guided by contention model, adaptively tune maximum number contention-free concurrent PM-threads....

Netflix offered special service that allows members to watch TV movies and shows on internet-connected devices, while & online or stream right your smart TV, game console. This paper revealed participants' gender identities were kept confidential through case study. found is expected put out the same shows, with it ensuring available for people in UK, may need be Individuals USA increase number of viewership.

Existing key-value stores (KVSs) based on log-structured merge-tree (LSM-tree) have been broadly deployed in practice to leverage characteristics of conventional block storage via file system, but lack effective exploitation for emerging byte-addressed persistent memory (PM). We reveal that these KVSs running upon existing PM-aware File systems cause inefficient PM I/O behaviors, including 1) numerous page faults, 2) misaligned with cacheline, and 3) bandwidth wastage concurrent threads. To...