QoS Policies and Architecture for Cache/Memory in CMP Platforms

Authors:   Ravi Iyer, Li Zhao, Fei Guo, Steve Reinhardt (Intel, NC State, U Michigan)
Venue:      SIGMETRICS 2007

First and foremost, the biggest thing this paper has going for it is the evaluation methodology. While the paper does not have adequate space to display a lot of test cases (few workloads are shown), they provide performance numbers for a purely trace-based methodology as well as a full-system mock up. What makes this particularly impressive is that the full-system is done with a modified Linux kernel which adds priority level to processes as well as new system calls to change QoS bits in platform registers.

Unlike other QoS papers, this one focuses on app prioritization. They do so by adjusting two key shared resources: memory bandwidth and cache allocation. The ideas present don't seem to provide any unsurprising insight. Essentially, giving more resources to a high-priority app (more than it's fair share) boost that app's performance at the cost of other low priority apps. The trade off is, not surprisingly, workload depending. A high priority streaming app doesn't need a lot of cache space, a low-priority cache-friendly workload does. While my read through was a bit light, I didn't see any specific algorithm notes that prevent this case from happening. Additionally, the workloads studied (at least for intervals tested) don't display any phase behavior, making the static allocations nearly identical (or in some cases better) than the dynamic allocations.

TL;DR: Resource allocation via application priority. Good test methodology. App (process) priority exists today in current Linux kernels.

Comments

Post a Comment

Popular posts from this blog

Fundamental Latency Trade-offs in Architecting DRAM Caches (Alloy Cache)

Authors: Moinuddin K. Qureshi and Gabriel H. Loh Venue:    MICRO 2012 The authors present a new architecture for DRAM cache which targets optimizing latency rather than hit rate. The authors show that for every optimization made to improve hit rate, one must also consider the added cache latency, thus resulting in a new "break even hit rate". The authors compare to three key baselines: an infeasibly large SRAM-tag array, the prior "LH-Cache", and a ideal latency-optimized cache. They find that "unoptimizations" improve latency, while minimally impacting hit-rate. Their final result is a cache which boosts performance 35% improvement over the baseline, and nearly 20% better than the previous LH-cache. The specific contributions are: Moving to a direct mapped cache provides significant improvement (by reducing latency) Combining tag-store and data-store into a single entity improves performance 21% Implementing a small per-core predictor enhances per...
Read more

ZCOMP: Reducing DNN Cross-Layer Memory Footprint Using Vector Extensions

Authors: Berkin Akin, Zeshan A. Christi, and Alaa R. Alameldeen Venue: MICRO, 2019 With accelerators dominating the deep learning space in architecture conferences, this paper stands out as it focuses on reducing DNN inference/training overhead while using a CPU. An obvious question is, with the wide-spread use of GPUs/TPUs for deep learning, why should we focus on optimizing CPUs for deep learning. In a recent paper published at HPCA 2019, Facebook claims that CPUs are preferred for applications where a tight integration is required between DNN and non-DNN tasks. Also, Intel's recent AVX512 has specialized support for DNNs in the form of new instructions called Vector Neural Network Instructions (VNNI). Broadly, optimizing DNNs can be viewed from two different perspectives; computation, and communication. This paper targets reducing communication overhead, more specifically, activation or feature-map communication overhead by compressing them. Compressing activations/weights...
Read more

AutoFDO: Automatic Feedback-Directed Optimization for Warehouse-Scale Applications

Authors: Dehao Chen, David Xinliang Li, Tipp Moseley Venue:    CGO 2016 This paper presents AutoFDO, a system used which profiles warehouse-scale applications, and applies feedback to the compilation for the next release. AutoFDO works by profiling and storing profiles in an aggregate database, annotating the profiles via an intermediate representation, and finally providing feedback. On average, this technique boosts performance by around 10% and works well even with stale releases. For profiles to be useful, they must be in an intermediate representation. This is built from binary-level profile, which uses LBR to map instruction frequencies. By using these frequencies, program counters, and the source, a source profile can be generated. The source profile can be used to build an annotated call-graph with edge frequencies, with some inaccuracy. These annotated call-graphs can then be used as feedback to the compiler. Because of the size of the applications and scale of d...
Read more