Clearing the clouds: a study of emerging scale-out workloads on modern hardware

By Michael Ferdman, Almutaz Adileh, Yusuf Onur Koçberber, Stavros Volos, M. Alisafaee, Djordje Jevdjic, Cansu Kaynak, A. Ailamaki, Babak Falsafi. ASPLOS XVII, 2012

1,015 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Hardware and Architecture, Information Systems

View Paper: https://www.semanticscholar.org/paper/fcd4b0c43fe131cc856b0d20fb1a1d4a2c30cd0d

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

In Hardware and Architecture

In Information Systems

Citations in the corpus

CATalyst: Defeating last-level cache side channel attacks in cloud computing (2016, methodology): “Although we only study the performance of SPEC and multi-threaded PARSEC benchmarks, we would expect that server benchmarks, scale-out workloads and real-world data center applications are even less sensitive to the LLC size than the memory-intensive SPEC and PARSEC benchmarks [13].”
Traffic management: a holistic approach to memory placement on NUMA systems (2013, background): “In summary, we conclude that contrary to the suggestion made in [13], it is too early to reduce the bandwidth of cross-chip interconnects on large multicore systems, especially if they are used for running large data-centric workloads.”
SoftSKU: Optimizing Server Architectures for Microservice Diversity @Scale (2019, methodology): “The microarchitectural trends we discover differ markedly from those of SPEC CPU2006/2017 [19, 20], academic cloud workloads [21, 22], and even some of Google’s major services [1, 23].”
Processing Data Where It Makes Sense: Enabling In-Memory Computation (2019, background): “Across all of these systems, including servers, cloud platforms, and mobile/embedded devices, the data working set sizes of modern applications are rapidly growing, causing the main memory to be a significant bottleneck for these applications [1, 2, 3, 4, 5, 6, 7].”
X-Stream: edge-centric graph processing using streaming partitions (2013, background): “The in-memory engine must deal with a larger number of partitions than the out-of-core engine as CPU caches are small in size with respect to main memory and current architectural trends indicate that they are unlikely to grow further [27].”
Mining on Someone Else's Dime: Mitigating Covert Mining Operations in Clouds and Enterprises (2017, background): “0) [37], SPEC 2006 [20], Rodinia [33] and Parboil [49].”
Optimizing Data-Center TCO with Scale-Out Processors (2012, background): “Functionally, all these applications have similar characteristics— namely, they operate on huge data sets that are split across a large number of nodes into memory-resident shards, the nodes service a large number of completely independent requests that don’t share state, and the internode connectiv”
A Survey on Distributed Machine Learning (2019, background): “The third reason is the increase in aggregate I/O bandwidth compared to a single machine [49].”
Efficient virtual memory for big memory servers (2013, background): “– Charles Thacker, 2010 ACM Turing Award Lecture.”
An Open-Source Benchmark Suite for Microservices and Their Hardware-Software Implications for Cloud & Edge Systems (2019, methodology): “The majority of front-end stalls are due to fetch , while branch mispredic-tions account for a smaller fraction of stalls for microser-vices than for other interactive applications, either cloud or IoT [37, 88].”
A scalable processing-in-memory accelerator for parallel graph processing (2015, methodology): “Our comprehensive evaluations using five state-of-the-art graph processing workloads with large real-world graphs show that the proposed architecture improves average system performance by a factor of ten and achieves 87% average energy reduction over conventional systems.”
Resource Management in Fog/Edge Computing (2018, methodology): “On the cloud, this is performed by running sample micro or macro applications that stress-tests each entity to obtain a snapshot of the performance of a Virtual Machine (VM) at any given point in time [49, 134].”
BigBench V2: The New and Improved BigBench (2017, background): “Comparing and contrasting these benchmarks to BigBench, we identify the uniqueness and superiority of BigBench [5] as follows: • BigBench is technology agnostic, whereas many of the existing benchmarks are technology or component bound (HiBench [16], SparkBench [17], MRBS [19], CloudSuite [20], Link”