Dmitrii Ustiugov

Official source

About this result

This page is automatically generated and may contain information that is not correct, complete, up-to-date, or relevant to your search query. The same applies to every other page on this website. Please make sure to verify the information with EPFL's official sources.

Related units

Courses taught by this person

Related research domains

Related publications

People doing similar research

Benchmarking, Analysis, and Optimization of Serverless Function Snapshots

Edouard Bugnion, Boris Robert Grot, Evangelos Marios Kogias, Dmitrii Ustiugov

Serverless computing has seen rapid adoption due to its high scalability and flexible, pay-as-you-go billing model. In serverless, developers structure their services as a collection of functions, sporadically invoked by various events like clicks. High inter-arrival time variability of function invocations motivates the providers to start new function instances upon each invocation, leading to significant cold-start delays that degrade user experience. To reduce cold-start latency, the industry has turned to snapshotting, whereby an image of a fully-booted function is stored on disk, enabling a faster invocation compared to booting a function from scratch. This work introduces vHive, an open-source framework for serverless experimentation with the goal of enabling researchers to study and innovate across the entire serverless stack. Using vHive, we characterize a state-of-the-art snapshot-based serverless infrastructure, based on industry-leading Containerd orchestration framework and Firecracker hypervisor technologies. We find that the execution time of a function started from a snapshot is 95% higher, on average, than when the same function is memory- resident. We show that the high latency is attributable to frequent page faults as the function’s state is brought from disk into guest memory one page at a time. Our analysis further reveals that funccloud computing, datacenters, serverless, virtualization, snapshotstions access the same stable working set of pages across different invocations of the same function. By leveraging this insight, we build REAP, a light-weight software mechanism for serverless hosts that records functions’ stable working set of guest memory pages and proactively prefetches it from disk into memory. Compared to baseline snapshotting, REAP slashes the cold-start delays by 3.7×, on average.

2021

, , , , ,

To provide low-latency and high-throughput guarantees, most large key-value stores keep the data in the memory of many servers. Despite the natural parallelism across lookups, the load imbalance, introduced by heavy skew in the popularity distribution of keys, limits performance. To avoid violating tail latency service-level objectives, systems tend to keep server utilization low and organize the data in micro-shards, which provides units of migration and replication for the purpose of load balancing. These techniques reduce the skew but incur additional monitoring, data replication, and consistency maintenance overheads. In this work, we introduce RackOut, a memory pooling technique that leverages the one-sided remote read primitive of emerging rack-scale systems to mitigate load imbalance while respecting service-level objectives. In RackOut, the data are aggregated at rack-scale granularity, with all of the participating servers in the rack jointly servicing all of the rack’s micro-shards. We develop a queuing model to evaluate the impact of RackOut at the datacenter scale. In addition, we implement a RackOut proof-of-concept key value store, evaluate it on two experimental platforms based on RDMA and Scale-Out NUMA, and use these results to validate the model. We devise two distinct approaches to load balancing within a RackOut unit, one based on random selection of nodes—RackOut_static—and another one based on an adaptive load balancing mechanism— RackOut_adaptive. Our results show that RackOut_static increases throughput by up to 6× for RDMA and 8.6× for Scale-Out NUMA compared to a scale-out deployment, while respecting tight tail latency service-level objectives. RackOut_adaptive improves the throughput by 30% for workloads with 20% of writes over RackOut_static.

2019

Prefetched Address Translation

Edouard Bugnion, Boris Robert Grot, Dmitrii Ustiugov

With explosive growth in dataset sizes and increasing machine memory capacities, per-application memory footprints are commonly reaching into hundreds of GBs. Such huge datasets pressure the TLB, resulting in frequent misses that must be resolved through a page walk - a long-latency pointer chase through multiple levels of the in-memory radix tree-based page table. Anticipating further growth in dataset sizes and their adverse affect on TLB hit rates, this work seeks to accelerate page walks while fully preserving existing virtual memory abstractions and mechanisms - a must for software compatibility and generality. Our idea is to enable direct indexing into a given level of the page table, thus eliding the need to first fetch pointers from the preceding levels. A key contribution of our work is in showing that this can be done by simply ordering the pages containing the page table in physical memory to match the order of the virtual memory pages they map to. Doing so enables direct indexing into the page table using a base-plus-offset arithmetic. We introduce Address Translation with Prefetching (ASAP), a new approach for reducing the latency of address translation to a single access to the memory hierarchy. Upon a TLB miss, ASAP launches prefetches to the deeper levels of the page table, bypassing the preceding levels. These prefetches happen concurrently with a conventional page walk, which observes a latency reduction due to prefetching while guaranteeing that only correctly-predicted entries are consumed. ASAP requires minimal extensions to the OS and trivial microarchitectural support. Moreover, ASAP is fully legacy-preserving, requiring no modifications to the existing radix tree-based page table, TLBs and other software and hardware mechanisms for address translation. Our evaluation on a range of memory-intensive workloads shows that under SMT colocation, ASAP is able to reduce page walk latency by an average of 25% (42% max) in native execution, and 45% (55% max) under virtualization.

ASSOC COMPUTING MACHINERY2019