Search for: All records

Serverless (or Function-as-a-Service) compute model enables new applications with dynamic scaling. However, all current Serverless systems are best-effort, and as we prove this means they cannot guarantee hard real-time deadlines, rendering them unsuitable for such real-time applications. We analyze a proposed extension of the Serverless model that adds a guaranteed invocation rate to the serverless model called Real-time Serverless. This approach aims to meet real-time deadlines with dynamically allocated function invocations. We first prove that the Serverless model does not support real-time guarantees. Next, we analyze Real-time Serverless, showing it can guarantee application real-time deadlines for rate-monotonic real-time workloads. Further, we derive bounds on the required invocation rate to meet any set of workload runtimes and periods. Subsequently, we explore an application technique, pre-invocation, and show that it can reduce the required guaranteed invocation rate. We derive bounds for the feasible rate guarantee reduction, and corresponding overhead in wasted compute resources. Finally, we apply the theoretical results to improve the experience quality of a distributed virtual reality/ augmented reality application as well as simplify the application design and resource management. 

Incorporating inductive biases is a promising approach for tackling challenging robot learning domains with sample-efficient solutions. This paper identifies partially observable domains where symmetries can be a useful inductive bias for efficient learning. Specifically, by encoding the equivariance regarding specific group symmetries into the neural networks, our actor-critic reinforcement learning agents can reuse solutions in the past for related scenarios. Consequently, our equivariant agents outperform non-equivariant approaches significantly in terms of sample efficiency and final performance, demonstrated through experiments on a range of robotic tasks in simulation and real hardware. 

We demonstrate a single-component hydrophilic photocrosslinkable copolymer system that incorporates all critical functionalities into one chain. This design allows for the creation of uniform functional organic coatings on a variety of substrates. The copolymers were composed of a poly(ethylene oxide)-containing monomer, a monomer that can release a primary amine upon UV light, and a monomer with reactive epoxide or cyclic dithiocarbonate with a primary amine. These copolymers are easily incorporated into the solution-casting process using polar solvents. Furthermore, the resulting coating can be readily stabilized through UV light-induced crosslinking, providing an advantage for controlling the surface properties of various substrates. The photocrosslinking capability further enables us to photolithographically define stable polymer domains in a desirable region. The resulting copolymer coatings were chemically versatile in immobilizing complex molecules by (i) post-crosslinking functionalization with the reactive groups on the surface and (ii) the formation of a composite coating by mixing varying amounts of a protein of interest, i.e., fish skin gelatin, which can form a uniform dual crosslinked network. The number of functionalization sites in a thin film could be controlled by tuning the composition of the copolymers. In photocrosslinking and subsequent functionalizations, we assessed the reactivity of the epoxide and cyclic dithiocarbonate with the generated primary amine. Moreover, the orthogonality of the possible reactions of the presented reactive functionalities in the crosslinked thin films with complex molecules is assessed. The resulting copolymer coatings were further utilized to define a hydrophobic surface or an active surface for the adhesion of biological objects.

 

Stream Processing Engines (SPEs) traditionally de-ploy applications on a set of shared workers (e.g., threads, processes, or containers) requiring complex performance man-agement by SPEs and application developers. We explore a new approach that replaces workers with Rate-based Abstract Ma-chines (RBAMs). This allows SPEs to translate stream operations into FaaS invocations, and exploit guaranteed invocation rates to manage performance. This approach enables SPE applications to achieve transparent and predictable performance. We realize the approach in the Storm-RTS system. Exploring 36 stream processing scenarios over 5 different hardware config-urations, we demonstrate several key advantages. First, Storm-RTS provides stable application performance and can enable flexible reconfiguration across cloud resource configurations. Sec-ond, SPEs built on RBAM can be resource-efficient and scalable. Finally, Storm-RTS allows the stream-processing paradigm to be extended from the cloud to the edge, using its performance stability to hide edge heterogeneity and resource competition. An experiment with 4 cloud and edge sites over 300 cores shows how Storm-RTS can support flexible reconfiguration and simple high-level declarative policies that optimize resource cost or other criteria. 

Generative AI, exemplified in ChatGPT, Dall-E 2, and Stable Diffusion, are exciting new applications consuming growing quantities of computing. We study the compute, energy, and carbon impacts of generative AI inference. Using ChatGPT as an exemplar, we create a workload model and compare request direction approaches (Local, Balance, CarbonMin), assessing their power use and carbon impacts. Our workload model shows that for ChatGPT-like services, in- ference dominates emissions, in one year producing 25x the carbon-emissions of training GPT-3. The workload model characterizes user experience, and experiments show that carbon emissions-aware algorithms (CarbonMin) can both maintain user experience and reduce carbon emissions dramatically (35%). We also consider a future scenario (2035 workload and power grids), and show that CarbonMin can reduce emissions by 56%. In both cases, the key is intelligent direction of requests to locations with low-carbon power. Combined with hardware technology advances, CarbonMin can keep emissions increase to only 20% compared to 2022 levels for 55x greater workload. Finally we consider datacenter headroom to increase effectiveness of shifting. With headroom, CarbonMin reduces 2035 emissions by 71%. 

This work unifies the analysis of various randomized methods for solving linear and nonlinear inverse problems with Gaussian priors by framing the problem in a stochastic optimization setting. By doing so, we show that many randomized methods are variants of a sample average approximation (SAA). More importantly, we are able to prove a single theoretical result that guarantees the asymptotic convergence for a variety of randomized methods. Additionally, viewing randomized methods as an SAA enables us to prove, for the first time, a single non-asymptotic error result that holds for randomized methods under consideration. Another important consequence of our unified framework is that it allows us to discover new randomization methods. We present various numerical results for linear, nonlinear, algebraic, and PDE-constrained inverse problems that verify the theoretical convergence results and provide a discussion on the apparently different convergence rates and the behavior for various randomized methods.

 

Reinforcement learning in partially observable domains is challenging due to the lack of observable state information. Thankfully, learning offline in a simulator with such state information is often possible. In particular, we propose a method for partially observable reinforcement learning that uses a fully observable policy (which we call a \emph{state expert}) during training to improve performance. Based on Soft Actor-Critic (SAC), our agent balances performing actions similar to the state expert and getting high returns under partial observability. Our approach can leverage the fully-observable policy for exploration and parts of the domain that are fully observable while still being able to learn under partial observability. On six robotics domains, our method outperforms pure imitation, pure reinforcement learning, the sequential or parallel combination of both types, and a recent state-of-the-art method in the same setting. A successful policy transfer to a physical robot in a manipulation task from pixels shows our approach's practicality in learning interesting policies under partial observability.