Sarthak Chakraborty

Shubham Agarwal, Sarthak Chakraborty, Shaddy Garg et al.

Cloud services are omnipresent and critical cloud service failure is a fact of life. In order to retain customers and prevent revenue loss, it is important to provide high reliability guarantees for these services. One way to do this is by predicting outages in advance, which can help in reducing the severity as well as time to recovery. It is difficult to forecast critical failures due to the rarity of these events. Moreover, critical failures are ill-defined in terms of observable data. Our proposed method, Outage-Watch, defines critical service outages as deteriorations in the Quality of Service (QoS) captured by a set of metrics. Outage-Watch detects such outages in advance by using current system state to predict whether the QoS metrics will cross a threshold and initiate an extreme event. A mixture of Gaussian is used to model the distribution of the QoS metrics for flexibility and an extreme event regularizer helps in improving learning in tail of the distribution. An outage is predicted if the probability of any one of the QoS metrics crossing threshold changes significantly. Our evaluation on a real-world SaaS company dataset shows that Outage-Watch significantly outperforms traditional methods with an average AUC of 0.98. Additionally, Outage-Watch detects all the outages exhibiting a change in service metrics and reduces the Mean Time To Detection (MTTD) of outages by up to 88% when deployed in an enterprise cloud-service system, demonstrating efficacy of our proposed method.

Shubham Agarwal, Subrata Mitra, Sarthak Chakraborty et al.

Text-to-image generation using diffusion models has seen explosive popularity owing to their ability in producing high quality images adhering to text prompts. However, production-grade diffusion model serving is a resource intensive task that not only require high-end GPUs which are expensive but also incurs considerable latency. In this paper, we introduce a technique called approximate-caching that can reduce such iterative denoising steps for an image generation based on a prompt by reusing intermediate noise states created during a prior image generation for similar prompts. Based on this idea, we present an end to end text-to-image system, Nirvana, that uses the approximate-caching with a novel cache management-policy Least Computationally Beneficial and Frequently Used (LCBFU) to provide % GPU compute savings, 19.8% end-to-end latency reduction and 19% dollar savings, on average, on two real production workloads. We further present an extensive characterization of real production text-to-image prompts from the perspective of caching, popularity and reuse of intermediate states in a large production environment.

Lovish Chopra, Sarthak Chakraborty, Abhijit Mondal et al.

With increasing advancements in technologies for capturing 360° videos, advances in streaming such videos have become a popular research topic. However, streaming 360° videos require high bandwidth, thus escalating the need for developing optimized streaming algorithms. Researchers have proposed various methods to tackle the problem, considering the network bandwidth or attempt to predict future viewports in advance. However, most of the existing works either (1) do not consider video contents to predict user viewport, or (2) do not adapt to user preferences dynamically, or (3) require a lot of training data for new videos, thus making them potentially unfit for video streaming purposes. We develop PARIMA, a fast and efficient online viewport prediction model that uses past viewports of users along with the trajectories of prime objects as a representative of video content to predict future viewports. We claim that the head movement of a user majorly depends upon the trajectories of the prime objects in the video. We employ a pyramid-based bitrate allocation scheme and perform a comprehensive evaluation of the performance of PARIMA. In our evaluation, we show that PARIMA outperforms state-of-the-art approaches, improving the Quality of Experience by over 30\% while maintaining a short response time.