Related papers: Development of a Burst Buffer System for Data-Inte…
Serving Large Language Models (LLMs) in production faces significant challenges from highly variable request patterns and severe resource fragmentation in serverless clusters. Current systems rely on static pipeline configurations that…
The continual increase of cores on server-grade CPUs raises demands on memory systems, which are constrained by limited off-chip pin and data transfer rate scalability. As a result, high-end processors typically feature lower memory…
In languages like C, buffer overflows are widespread. A common mitigation technique is to use tools that detect them during execution and abort the program to prevent the leakage of data or the diversion of control flow. However, for server…
To understand and improve DRAM performance, reliability, security and energy efficiency, prior works study characteristics of commodity DRAM chips. Unfortunately, state-of-the-art open source infrastructures capable of conducting such…
As LLMs and foundation models scale, checkpoint/restore has become a critical pattern for training and inference. With 3D parallelism (tensor, pipeline, data), checkpointing involves many processes, each managing numerous tensors of varying…
We describe a novel architecture that combines the simplicity of RESTful architecture with the power of functional programming for delivering web-services. Although, RESTful architecture has been quite useful in simplifying the development…
Autoscaling is a technology that automatically scales resources for applications without human intervention to ensure runtime Quality of Service (QoS) while reducing costs. However, user-facing cloud applications serve dynamic workloads…
Spreadsheets are widely used for data exploration. Since spreadsheet systems have limited capabilities, users often need to load spreadsheets to other data science environments to perform advanced analytics. However, current approaches for…
Due to high accuracy, BERT-like models have been widely adopted by text mining and web searching. However, large BERT-like models suffer from inefficient online inference, facing the following two problems on GPUs: (1) their high accuracy…
Existing methods for training LLMs on long-sequence data, such as Tensor Parallelism and Context Parallelism, exhibit low Model FLOPs Utilization as sequence lengths and number of GPUs increase, especially when sequence lengths exceed 1M…
In today's enterprise storage systems, supported data services such as snapshot delete or drive rebuild can cause tremendous performance interference if executed inline along with heavy foreground IO, often leading to missing SLOs (Service…
The last decades have seen a surge of interests in distributed computing thanks to advances in clustered computing and big data technology. Existing distributed algorithms typically assume {\it all the data are already in one place}, and…
Today, companies and data centers are moving towards cloud and serverless storage systems instead of traditional file systems. As a result of such a transition, allocating sufficient resources to users and parties to satisfy their service…
To accommodate the needs of large-scale distributed P2P systems, scalable data management strategies are required, allowing applications to efficiently cope with continuously growing, highly dis tributed data. This paper addresses the…
This lengthy document often referred to as the "Lustre Book", contains a detailed outline of Lustre file system architecture, as it was created between 2001 and 2005, in accordance with the requirements from various users. Now, in 2019,…
A buffer k-d tree is a k-d tree variant for massively-parallel nearest neighbor search. While providing valuable speed-ups on modern many-core devices in case both a large number of reference and query points are given, buffer k-d trees are…
We discuss the design and ongoing development of the Monitoring Extreme-scale Lustre Toolkit (MELT), a unified Lustre performance monitoring and analysis infrastructure that provides continuous, low-overhead summary information on the…
With the explosive growth of big data, workloads tend to get more complex and computationally demanding. Such applications are processed on distributed interconnected resources that are becoming larger in scale and computational capacity.…
The bittide mechanism enables logically synchronous computation across distributed systems by leveraging the continuous frame transmission inherent to wired networks such as Ethernet. Instead of relying on a global clock, bittide uses a…
Current generation solid-state storage devices are exposing a new bottlenecks in the SCSI and block layers of the Linux kernel, where IO throughput is limited by lock contention, inefficient interrupt handling, and poor memory locality. To…