publication

2025

1. SDM’25

   Staleness-Alleviated Distributed GNN Training via Online Dynamic-Embedding Prediction

   Guangji Bai, Ziyang Yu, Zheng Chai,  Yue Cheng,  and Liang Zhao

   In Proceedings of SIAM International Conference on Data Mining 2025

   PDF

2024

1. SoCC’24

   FedCaSe: Enhancing Federated Learning with Heterogeneity-aware Caching and Scheduling

   Redwan Ibne Seraj Khan, Arnab K. Paul,  Yue Cheng, Xun Jian,  and Ali R. Butt

   In Proceedings of the ACM Symposium on Cloud Computing 2024

   PDF Code
2. VLDB’24

   Everything You Always Wanted to Know About Storage Compressibility of Pre-Trained ML Models but Were Afraid to Ask

   Zhaoyuan Su, Ammar Ahmed, Zirui Wang, Ali Anwar,  and Yue Cheng

   In 50th International Conference on Very Large Data Bases 2024

   arXiv PDF Code
3. VLDB’24

   Algorithmic Complexity Attacks on Dynamic Learned Indexes

   Rui Yang, Evgenios M. Kornaropoulos,  and Yue Cheng

   In 50th International Conference on Very Large Data Bases 2024

   arXiv PDF Code
4. USENIX ATC’24

   ALPS: An Adaptive Learning, Priority OS Scheduler for Serverless Functions

   Yuqi Fu, Ruizhe Shi, Haoliang Wang, Songqing Chen,  and Yue Cheng

   In 2024 USENIX Annual Technical Conference (USENIX ATC 24) 2024

   PDF Code Talk
5. SIGMETRICS’24

   A Closer Look into IPFS: Accessibility, Content, and Performance

   Ruizhe Shi, Ruizhi Cheng, Bo Han,  Yue Cheng,  and Songqing Chen

   In ACM SIGMETRICS / IFIP Performance 2024

   PDF
6. arXiv

   Beyond Efficiency: A Systematic Survey of Resource-Efficient Large Language Models

   Guangji Bai, Zheng Chai, Chen Ling, Shiyu Wang, Jiaying Lu, Nan Zhang, Tingwei Shi, Ziyang Yu, Mengdan Zhu, Yifei Zhang, Carl Yang,  Yue Cheng,  and Liang Zhao

   In 2024

   arXiv Website

2023

1. Dissertation

   Towards Elastic and Cost-effective Stateful Serverless Systems

   Jingyuan Zhang

   PhD Dissertation 2023

   HTML
2. BigData’23

   Towards Cost-effective and Resource-aware Aggregation at Edge for Federated Learning

   Ahmad Khan, Yuze Li, Xinran Wang, Sabaat Haroon, Haider Ali,  Yue Cheng, Ali R. Butt,  and Ali Anwar

   In 2023 IEEE International Conference on Big Data 2023

   PDF
3. ASPLOS’23

   λFS: A Scalable and Elastic Distributed File System Metadata Service using Serverless Functions

   Benjamin Carver, Runzhou Han, Jingyuan Zhang, Mai Zheng,  and Yue Cheng

   In 28th ACM International Conference on Architectural Support for Programming Languages and Operating Systems 2023

   arXiv PDF Code
4. VLDB’23

   InfiniStore: Elastic Serverless Cloud Storage

   Jingyuan Zhang, Ao Wang, Xiaolong Ma, Benjamin Carver, Nicholas John Newman, Ali Anwar, Lukas Rupprecht, Dimitrios Skourtis, Vasily Tarasov, Feng Yan,  and Yue Cheng

   In 49th International Conference on Very Large Data Bases 2023

   arXiv PDF Code
5. USENIX FAST’23

   SHADE: Enable Fundamental Cacheability for Distributed Deep Learning Training

   Redwan Ibne Seraj Khan, Ahmad Hossein Yazdani, Yuqi Fu, Arnab K. Paul, Bo Ji, Xun Jian,  Yue Cheng,  and Ali R. Butt

   In 21th USENIX Conference on File and Storage Technologies (FAST 23) 2023

   PDF Code Talk

2022

1. DRBSD-8 ’22

   Understanding Impact of Lossy Compression on Derivative-related Metrics in Scientific Datasets

   Zhaoyuan Su, Sheng Di, Ali Murat Gok,  Yue Cheng,  and Franck Cappello

   In Proceedings of the 8th International Workshop on Data Analysis and Reduction for Big Scientific Data 2022
2. Preprint

   InfiniStore: Elastic Serverless Cloud Storage

   Jingyuan Zhang, Ao Wang, Xiaolong Ma, Benjamin Carver, Nicholas John Newman, Ali Anwar, Lukas Rupprecht, Dimitrios Skourtis, Vasily Tarasov, Feng Yan,  and Yue Cheng

   In Preprint 2022

   arXiv Code
3. SC’22

   SFS: Smart OS Scheduling for Serverless Functions

   Yuqi Fu, Li Liu, Haoliang Wang,  Yue Cheng,  and Songqing Chen

   In Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis 2022

   Best Student Paper Award Finalist
   arXiv PDF Code

2021

1. SoCC’21

   Mind the Gap: Broken Promises of CPU Reservations in Containerized Multi-Tenant Clouds

   Li Liu, Haoliang Wang, An Wang, Mengbai Xiao,  Yue Cheng,  and Songqing Chen

   In Proceedings of the ACM Symposium on Cloud Computing 2021

   Abstract PDF Code

   Containerization is becoming increasingly popular, but unfortunately, containers often fail to deliver the anticipated performance with the allocated resources. In this paper, we first demonstrate the performance variance and degradation are significant (by up to 5x) in a multi-tenant environment where containers are co-located. We then investigate the root cause of such performance degradation. Contrary to the common belief that such degradation is caused by resource contention and interference, we find that there is a gap between the amount of CPU a container reserves and actually gets. The root cause lies in the design choices of today’s Linux scheduling mechanism, which we call Forced Runqueue Sharing and Phantom CPU Time. In fact, there are fundamental conflicts between the need to reserve CPU resources and Completely Fair Scheduler’s work-conserving nature, and this contradiction prevents a container from fully utilizing its requested CPU resources. As a proof-of-concept, we implement a new resource configuration mechanism atop the widely used Kubernetes and Linux to demonstrate its potential benefits and shed light on future scheduler redesign. Our proof-of-concept, compared to the existing scheduler, improves the performance of both batch and interactive containerized apps by up to 5.6x and 13.7x.

2. SC’21

   FedAT: A High-Performance and Communication-Efficient Federated Learning System with Asynchronous Tiers

   Zheng Chai, Yujing Chen, Ali Anwar, Liang Zhao,  Yue Cheng,  and Huzefa Rangwala

   In Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis 2021

   Abstract arXiv PDF

   Federated learning (FL) involves training a model over massive distributed devices, while keeping the training data localized and private. This form of collaborative learning exposes new tradeoffs among model convergence speed, model accuracy, balance across clients, and communication cost, with new challenges including: (1) straggler problem—where clients lag due to data or (computing and network) resource heterogeneity, and (2) communication bottleneck—where a large number of clients communicate their local updates to a central server and bottleneck the server. Many existing FL methods focus on optimizing along only one single dimension of the tradeoff space. Existing solutions use asynchronous model updating or tiering-based, synchronous mechanisms to tackle the straggler problem. However, asynchronous methods can easily create a communication bottleneck, while tiering may introduce biases that favor faster tiers with shorter response latencies.To address these issues, we present FedAT, a novel Federated learning system with Asynchronous Tiers under Non-i.i.d. training data. FedAT synergistically combines synchronous, intra-tier training and asynchronous, cross-tier training. By bridging the synchronous and asynchronous training through tiering, FedAT minimizes the straggler effect with improved convergence speed and test accuracy. FedAT uses a straggler-aware, weighted aggregation heuristic to steer and balance the training across clients for further accuracy improvement. FedAT compresses uplink and downlink communications using an efficient, polyline-encoding-based compression algorithm, which minimizes the communication cost. Results show that FedAT improves the prediction performance by up to 21.09% and reduces the communication cost by up to 8.5\texttimes, compared to state-of-the-art FL methods.

3. USENIX ATC’21

   FaaSNet: Scalable and Fast Provisioning of Custom Serverless Container Runtimes at Alibaba Cloud Function Compute

   Ao Wang, Shuai Chang, Huangshi Tian, Hongqi Wang, Haoran Yang, Huiba Li, Rui Du,  and Yue Cheng

   In 2021 USENIX Annual Technical Conference (USENIX ATC 21) 2021

   PDF Code Talk
4. OPT’21

   Community-based Layerwise Distributed Training of Graph Convolutional Networks

   Hongyi Li, Junxiang Wang, Yongchao Wang,  Yue Cheng,  and Liang Zhao

   In The 13th International OPT Workshop on Optimization for Machine Learning (OPT’21) 2021

   arXiv
5. Thesis

   Wukong: A Fast, Cost-Effective, and Easy-to-Use Serverless DAG Engine

   Benjamin Carver

   Master’s Thesis 2021

   HTML

2020

1. ICDM ’21

   Toward Model Parallelism for Deep Neural Network Based on Gradient-Free ADMM Framework

   Junxiang Wang, Zheng Chai,  Yue Cheng,  and Liang Zhao

   In 2020 IEEE International Conference on Data Mining (ICDM) 2020

   arXiv
2. SoCC’20

   Wukong: A Scalable and Locality-Enhanced Framework for Serverless Parallel Computing

   Benjamin Carver, Jingyuan Zhang, Ao Wang, Ali Anwar, Panruo Wu,  and Yue Cheng

   In Proceedings of the 11th ACM Symposium on Cloud Computing 2020

   Abstract arXiv PDF Code Talk

   Executing complex, burst-parallel, directed acyclic graph (DAG) jobs poses a major challenge for serverless execution frameworks, which will need to rapidly scale and schedule tasks at high throughput, while minimizing data movement across tasks. We demonstrate that, for serverless parallel computations, decentralized scheduling enables scheduling to be distributed across Lambda executors that can schedule tasks in parallel, and brings multiple benefits, including enhanced data locality, reduced network I/Os, automatic resource elasticity, and improved cost effectiveness. We describe the implementation and deployment of our new serverless parallel framework, called Wukong, on AWS Lambda. We show that Wukong achieves near-ideal scalability, executes parallel computation jobs up to 68.17X faster, reduces network I/O by multiple orders of magnitude, and achieves 92.96% tenant-side cost savings compared to numpywren.

3. OPTML-ICML’20

   Tunable Subnetwork Splitting for Model-parallelism of Neural Network Training

   Junxiang Wang, Zheng Chai,  Yue Cheng,  and Liang Zhao

   In Beyond First Order Methods in ML Systems 2020

   arXiv Code
4. HPDC’20

   TiFL: A Tier-Based Federated Learning System

   Zheng Chai, Ahsan Ali, Syed Zawad, Stacey Truex, Ali Anwar, Nathalie Baracaldo, Yi Zhou, Heiko Ludwig, Feng Yan,  and Yue Cheng

   In Proceedings of the 29th International Symposium on High-Performance Parallel and Distributed Computing 2020

   Abstract arXiv PDF

   Federated Learning (FL) enables learning a shared model acrossmany clients without violating the privacy requirements. One of the key attributes in FL is the heterogeneity that exists in both resource and data due to the differences in computation and communication capacity, as well as the quantity and content of data among different clients. We conduct a case study to show that heterogeneity in resource and data has a significant impact on training time and model accuracy in conventional FL systems. To this end, we propose TiFL, a Tier-based Federated Learning System, which divides clients into tiers based on their training performance and selects clients from the same tier in each training round to mitigate the straggler problem caused by heterogeneity in resource anddata quantity. To further tame the heterogeneity caused by non-IID (Independent and Identical Distribution) data and resources, TiFL employs an adaptive tier selection approach to update the tiering on-the-fly based on the observed training performance and accuracy. We prototype TiFL in a FL testbed following Google’s FL architecture and evaluate it using the state-of-the-art FL benchmarks. Experimental evaluation shows that TiFL outperforms the conventional FL in various heterogeneous conditions. With the proposed adaptive tier selection policy, we demonstrate that TiFL achieves much faster training performance while achieving the same or better test accuracy across the board.

5. USENIX FAST’20

   InfiniCache: Exploiting Ephemeral Serverless Functions to Build a Cost-Effective Memory Cache

   Ao Wang, Jingyuan Zhang, Xiaolong Ma, Ali Anwar, Lukas Rupprecht, Dimitrios Skourtis, Vasily Tarasov, Feng Yan,  and Yue Cheng

   In 18th USENIX Conference on File and Storage Technologies (FAST 20) 2020

   PDF Talk Website Press Blog

2019

1. TPDS

   MOANA: Modeling and Analyzing I/O Variability in Parallel System Experimental Design

   Kirk W. Cameron, Ali Anwar,  Yue Cheng, Li Xu, Bo Li, Uday Ananth, Jon Bernard, Chandler Jearls, Thomas Lux, Yili Hong, Layne T. Watson,  and Ali R. Butt

   IEEE Transactions on Parallel and Distributed Systems 2019

   HTML
2. USENIX NSDI’19

   HyperFaaS: A Truly Elastic Serverless Computing Framework

   Jingyuan Zhang, Ao Wang, Min Li, Yuan Chen,  and Yue Cheng

   In USENIX Symposium on Networked Systems Design and Implementation 2019

   Poster
3. PDSW’19

   In Search of a Fast and Efficient Serverless DAG Engine

   Benjamin Carver, Jingyuan Zhang, Ao Wang,  and Yue Cheng

   In 2019 IEEE/ACM Fourth International Parallel Data Systems Workshop (PDSW) 2019

   arXiv PDF Code
4. Cloud’19

   Bolt: Towards a Scalable Docker Registry via Hyperconvergence

   Michael Littley, Ali Anwar, Hannan Fayyaz, Zeshan Fayyaz, Vasily Tarasov, Lukas Rupprecht, Dimitrios Skourtis, Mohamed Mohamed, Heiko Ludwig,  Yue Cheng,  and Ali R. Butt

   In 2019 IEEE 12th International Conference on Cloud Computing (CLOUD) 2019

   PDF
5. USENIX OpML’19

   Towards Taming the Resource and Data Heterogeneity in Federated Learning

   Zheng Chai, Hannan Fayyaz, Zeshan Fayyaz, Ali Anwar, Yi Zhou, Nathalie Baracaldo, Heiko Ludwig,  and Yue Cheng

   In 2019 USENIX Conference on Operational Machine Learning (OpML 19) 2019

   PDF Talk
6. VEE’19

   VCPU as a Container: Towards Accurate CPU Allocation for VMs

   Li Liu, Haoliang Wang, An Wang, Mengbai Xiao,  Yue Cheng,  and Songqing Chen

   In Proceedings of the 15th ACM SIGPLAN/SIGOPS International Conference on Virtual Execution Environments 2019

   Abstract PDF

   With our increasing reliance on cloud computing, accurate resource allocation of virtual machines (or domains) in the cloud have become more and more important. However, the current design of hypervisors (or virtual machine monitors) fails to accurately allocate resources to the domains in the virtualized environment. In this paper, we claim the root cause is that the protection scope is erroneously used as the resource scope for a domain in the current virtualization design. Such design flaw prevents the hypervisor from accurately accounting resource consumption of each domain. In this paper, using virtual CPUs as a container we propose to redefine the resource scope of a domain, so that the new resource scope is aligned with all the CPU consumption incurred by this domain. As a demonstration, we implement a novel system, called VASE (vCPU as a container), on top of the Xen hypervisor. Evaluations on our testbed have shown our proposed approach is effective in accounting system-wide CPU consumption incurred by domains, while introducing negligible overhead to the system.

2018

1. Book Chapter

   SDN helps Big Data to optimize Storage

   Ali R. Butt, Ali Anwar,  and Yue Cheng

   Book Chapter, Big Data and Software Defined Networks, Editor: Javid Taheri. IET, ISBN 978-1-78561-304-3. 2018
2. BigData’18

   Analyzing Alibaba’s Co-located Datacenter Workloads

   Yue Cheng, Ali Anwar,  and Xuejing Duan

   In 2018 IEEE International Conference on Big Data (Big Data) 2018

   PDF
3. SC’18

   BESPOKV: Application Tailored Scale-Out Key-Value Stores

   Ali Anwar,  Yue Cheng, Hai Huang, Jingoo Han, Hyogi Sim, Dongyoon Lee, Fred Douglis,  and Ali R. Butt

   In SC18: International Conference for High Performance Computing, Networking, Storage and Analysis 2018

   PDF Code
4. ApSys’18

   Characterizing Co-Located Datacenter Workloads: An Alibaba Case Study

   Yue Cheng, Zheng Chai,  and Ali Anwar

   In Proceedings of the 9th Asia-Pacific Workshop on Systems 2018

   arXiv
5. USENIX FAST’18

   Improving Docker Registry Design Based on Production Workload Analysis

   Ali Anwar, Mohamed Mohamed, Vasily Tarasov, Michael Littley, Lukas Rupprecht,  Yue Cheng, Nannan Zhao, Dimitrios Skourtis, Amit S. Warke, Heiko Ludwig, Dean Hildebrand,  and Ali R. Butt

   In 16th USENIX Conference on File and Storage Technologies (FAST 18) 2018

   PDF Code Talk
6. IPDPS’18

   Chameleon: An Adaptive Wear Balancer for Flash Clusters

   Nannan Zhao, Ali Anwar,  Yue Cheng, Mohammed Salman, Daping Li, Jiguang Wan, Changsheng Xie, Xubin He, Feiyi Wang,  and Ali Butt

   In 2018 IEEE International Parallel and Distributed Processing Symposium (IPDPS) 2018

   PDF

2017

1. Doctoral Dissertation

   Workload-aware Efficient Storage Systems

   Yue Cheng

   PhD Dissertation 2017

   HTML

2016

1. Internet Computing

   Provider versus Tenant Pricing Games for Hybrid Object Stores in the Cloud

   Yue Cheng, M. Safdar Iqbal, Aayush Gupta,  and Ali R. Butt

   IEEE Internet Computing 2016

   HTML
2. USENIX ATC’16

   Erasing Belady’s Limitations: In Search of Flash Cache Offline Optimality

   Yue Cheng, Fred Douglis, Philip Shilane, Grant Wallace, Peter Desnoyers,  and Kai Li

   In 2016 USENIX Annual Technical Conference (USENIX ATC 16) 2016

   PDF Talk
3. HotStorage’16

   ClusterOn: Building Highly Configurable and Reusable Clustered Data Services Using Simple Data Nodes

   Ali Anwar,  Yue Cheng, Hai Huang,  and Ali R. Butt

   In 8th USENIX Workshop on Hot Topics in Storage and File Systems (HotStorage 16) 2016

   PDF Talk
4. HPDC’16

   MOS: Workload-Aware Elasticity for Cloud Object Stores

   Ali Anwar,  Yue Cheng, Aayush Gupta,  and Ali R. Butt

   In Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing 2016

   Abstract PDF

   The use of cloud object stores has been growing rapidly in recent years as they combine key advantages such as HTTP-based RESTful APIs, high availability, elasticity with a "pay-as-you-go" pricing model that allows applications to scale as needed. The current practice is to either use a single set of configuration parameters or rely on statically configured storage policies for a cloud object store deployment, even when the store is used to support different types of applications with evolving requirements. This crucial mismatch between the different applications requirements and capabilities of the object store is problematic and should be addressed to achieve high efficiency and performance.In this paper, we propose MOS, a Micro Object Storage architecture, which supports independently configured microstores each tuned dynamically to the needs of a particular type of workload. We also design an enhancement, MOS++, that extends MOS’s capabilities through fine-grained resource management to effectively meet the tenants’ SLAs while maximizing resource efficiency. We have implemented a prototype of MOS ++ in OpenStack Swift using Docker containers. Our evaluation shows that MOS ++ can effectively support heterogeneous workloads across multiple tenants. Compared to default and statically configured object store setups, for a two-tenant setup, MOS++ improves the sustained access bandwidth by up to 79% for a large-object workload, while reducing the 95th percentile latency by up to 70.2% for a small-object workload.

5. VarSys’16

   Towards Managing Variability in the Cloud

   Ali Anwar,  Yue Cheng,  and Ali R. Butt

   In 2016 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW) 2016

   PDF

2015

1. PDSW’15

   Taming the Cloud Object Storage with MOS

   Ali Anwar,  Yue Cheng, Aayush Gupta,  and Ali R. Butt

   In Proceedings of the 10th Parallel Data Storage Workshop 2015

   Abstract PDF

   Cloud object stores today are deployed using a single set of configuration parameters for all different types of applications. This homogeneous setup results in all applications experiencing the same service level (e.g., data transfer throughput, etc.). However, the vast variety of applications expose extremely different latency and throughput requirements. To this end, we propose MOS, a Micro Object Storage architecture with independently configured microstores each tuned dynamically for a particular type of workload. We then expose these microstores to the tenant who can then choose to place their data in the appropriate microstore according the latency and throughput requirements of their workloads. Our evaluation shows that compared with default setup, MOS can improve the performance up to 200% for small objects and 28% for large objects while providing opportunity of tradeoff between two.

2. HotCloud’15

   Pricing Games for Hybrid Object Stores in the Cloud: Provider vs. Tenant

   Yue Cheng, M. Safdar Iqbal, Aayush Gupta,  and Ali R. Butt

   In 7th USENIX Workshop on Hot Topics in Cloud Computing (HotCloud 15) 2015

   PDF Talk
3. HPDC’15

   CAST: Tiering Storage for Data Analytics in the Cloud

   Yue Cheng, M. Safdar Iqbal, Aayush Gupta,  and Ali R. Butt

   In Proceedings of the 24th International Symposium on High-Performance Parallel and Distributed Computing 2015

   Abstract PDF Talk

   Enterprises are increasingly moving their big data analytics to the cloud with the goal of reducing costs without sacrificing application performance. Cloud service providers offer their tenants a myriad of storage options, which while flexible, makes the choice of storage deployment non trivial. Crafting deployment scenarios to leverage these choices in a cost-effective manner - under the unique pricing models and multi-tenancy dynamics of the cloud environment - presents unique challenges in designing cloud-based data analytics frameworks.In this paper, we propose CAST, a Cloud Analytics Storage Tiering solution that cloud tenants can use to reduce monetary cost and improve performance of analytics workloads. The approach takes the first step towards providing storage tiering support for data analytics in the cloud. CAST performs offline workload profiling to construct job performance prediction models on different cloud storage services, and combines these models with workload specifications and high-level tenant goals to generate a cost-effective data placement and storage provisioning plan. Furthermore, we build CAST++ to enhance CAST’s optimization model by incorporating data reuse patterns and across-jobs interdependencies common in realistic analytics workloads. Tests with production workload traces from Facebook and a 400-core Google Cloud based Hadoop cluster demonstrate that CAST++ achieves 1.21X performance and reduces deployment costs by 51.4% compared to local storage configuration.

4. EuroSys’15

   An In-Memory Object Caching Framework with Adaptive Load Balancing

   Yue Cheng, Aayush Gupta,  and Ali R. Butt

   In Proceedings of the Tenth European Conference on Computer Systems 2015

   Abstract PDF Talk

   The extreme latency and throughput requirements of modern web applications are driving the use of distributed in-memory object caches such as Memcached. While extant caching systems scale-out seamlessly, their use in the cloud — with its unique cost and multi-tenancy dynamics — presents unique opportunities and design challenges.In this paper, we propose MBal, a high-performance in-memory object caching framework with adaptive Multiphase load Balancing, which supports not only horizontal (scale-out) but vertical (scale-up) scalability as well. MBal is able to make efficient use of available resources in the cloud through its fine-grained, partitioned, lockless design. This design also lends itself naturally to provide adaptive load balancing both within a server and across the cache cluster through an event-driven, multi-phased load balancer. While individual load balancing approaches are being lever-aged in in-memory caches, MBal goes beyond the extant systems and offers a holistic solution wherein the load balancing model tracks hotspots and applies different strategies based on imbalance severity – key replication, server-local or cross-server coordinated data migration. Performance evaluation on an 8-core commodity server shows that compared to a state-of-the-art approach, MBal scales with number of cores and executes 2.3x and 12x more queries/second for GET and SET operations, respectively.

2013

1. SoCC’13

   High Performance In-Memory Caching through Flexible Fine-Grained Services

   Yue Cheng, Aayush Gupta, Anna Povzner,  and Ali R. Butt

   In Proceedings of the 4th Annual Symposium on Cloud Computing 2013

   Abstract PDF Poster

   In-memory object caches are extensively used in today’s web installations [1, 6]. Most existing systems adopt monolithic storage models and engineer optimizations on specific workload characteristics [3, 6] or operations [4, 5]. Such optimizations are insufficient as large-scale cloud workloads typically exhibit both temporal and spatial shifts - requirements vary within the same deployment over time and different parts of the same workload demonstrate different access patterns. To this end, we propose a caching tier that supports differentiated services in multiple dimensions. Since there is no best "one-size-fits-all" solution for all workload requirements, we argue that a fine-grained modular design will provide both high performance and flexibility in supporting multiple services.