In persistent B+ -Tree, random updates of small key-value (KV) pairs will cause severe XPBuffer-induced write amplification (XBI-amplification) because CPU cacheline size is smaller than media access granularity in persistent memory (PM). We observe that XBI-amplification directly determines the application performance when the PM bandwidth is exhausted in multi-thread scenarios. However, none of the existing work can efficiently address the XBI-amplification issue while maintaining superior range query performance.

Persistent memory allocation is a fundamental building block for developing high-performance and in-memory applications. Existing persistent memory allocators suffer from many performance issues. First, they may introduce repeated cache line flushes and small random accesses in persistent memory for their poor heap metadata management. Second, they use static slab segregation resulting in a dramatic increase in memory consumption when allocation request size is changed. Third, they are not aware of NUMA effect, leading to remote persistent memory accesses in memory allocation and deallocation processes. In this article, we design a novel allocator, named PMAlloc, to solve the above issues simultaneously. (1) PMAlloc eliminates cache line reflushes by mapping contiguous data blocks in slabs to interleaved metadata entries stored in different cache lines. (2) It writes small metadata units to a persistent bookkeeping log in a sequential pattern to remove random heap metadata accesses in persistent memory. (3) Instead of using static slab segregation, it supports slab morphing, which allows slabs to be transformed between size classes to significantly improve slab usage. (4) It uses a local-first allocation policy to avoid allocating remote memory blocks. And it supports a two-phase deallocation mechanism including recording and synchronization to minimize the number of remote memory access in the deallocation. PMAlloc is complementary to the existing consistency models. Results on six benchmarks demonstrate that PMAlloc improves the performance of state-of-the-art persistent memory allocators by up to 6.4× and 57× for small and large allocations, respectively. PMAlloc with NUMA optimizations brings a 2.9× speedup in multi-socket evaluation and is up to 36× faster than other persistent memory allocators. Using PMAlloc reduces memory usage by up to 57.8%. Besides, we integrate PMAlloc in a persistent FPTree. Compared to the state-of-the-art allocators, PMAlloc improves the performance of this application by up to 3.1×.

Persistent memory allocation is a fundamental building block for developing high-performance and in-memory applications. Existing persistent memory allocators suffer from suboptimal heap organizations that introduce repeated cache line flushes and small random accesses in persistent memory. Worse, many allocators use static slab segregation resulting in a dramatic increase in memory consumption when allocation request size is changed. In this paper, we design a novel allocator, named NVAlloc, to solve the above issues simultaneously. First, NVAlloc eliminates cache line reflushes by mapping contiguous data blocks in slabs to interleaved metadata entries stored in different cache lines. Second, it writes small metadata units to a persistent bookkeeping log in a sequential pattern to remove random heap metadata accesses in persistent memory. Third, instead of using static slab segregation, it supports slab morphing, which allows slabs to be transformed between size classes to significantly improve slab usage. NVAlloc is complementary to the existing consistency models. Results on 6 benchmarks demonstrate that NVAlloc improves the performance of state-of-the-art persistent memory allocators by up to 6.4x and 57x for small and large allocations, respectively. Using NVAlloc reduces memory usage by up to 57.8%. Besides, we integrate NVAlloc in a persistent FPTree. Compared to the state-of-the-art allocators, NVAlloc improves the performance of this application by up to 3.1x.

Data swapping between CPUs and GPUs is widely used to address the GPU memory shortage issue when training deep neural networks (DNNs) requiring a larger amount of memory than that a GPU may have. Data swapping may become a bottleneck when its latency is longer than the latency of DNN computations. Tensor compression in GPUs can reduce the data swapping time. However, existing works on compressing tensors in the virtual memory of GPUs have three major issues: lack of portability because its implementation requires additional (de)compression units in memory controllers, sub-optimal compression performance for varying tensor compression ratios and sizes, and poor adaptation to dense tensors because they only focus on sparse tensors. We propose a self-tuning tensor compression framework, namedCSwap+, for improving the virtual memory management of GPUs. It uses GPUs for (de)compression directly and thus has high portability and is minimally dependent on GPU architecture features. Furthermore, it only applies compression on tensors that are deemed to be cost-effective considering their compression ratio, size, and the characteristics of compression algorithms at runtime. Finally, to adapt to DNN models with dense tensors, it also supports cost-effective lossy compression for dense tensors with nearly no model training accuracy degradation. We conduct the experiments through six representative memory-intensive DNN models. Compared to vDNN,CSwap+reduces tensor swapping latency by up to 50.9% and 46.1% with NVIDIA V100 GPU, for DNN models with sparse and dense tensors, respectively.

Graphic Processing Units (GPUs) have limited memory capacity. Training popular deep neural networks (DNNs) often requires a larger amount of memory than that a GPU may have. Consequently, training data needs to be swapped between CPUs and GPUs. Data swapping may become a bottleneck when its latency is longer than the latency of DNN computations. Tensor compression in GPUs can reduce the data swapping time. However, existing works on compressing tensors in the virtual memory of GPUs have two major issues: sub-optimal compression performance for varying tensor sparsity and sizes and lack of portability because its implementation requires additional (de)compression units in memory controllers. We propose a self-tuning tensor compression framework, named CSWAP, for improving the virtual memory management of GPUs. It has high portability and is minimally dependent on GPU architecture features. Furthermore, its runtime only applies compression on tensors that are deemed to be cost-effective considering their sparsity and size and the characteristics of compression algorithms. Finally, our framework is fully automated and can customize the compression policy for different neural network architectures and GPU architectures. Our experimental results using six representative memory-intensive DNN models show that CSWAP reduces tensor swapping latency by up to 50.9% and reduces the DNN training time by 20.7% on average with NVIDIA V100 GPUs compared to vDNN.