[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86271-en":3,"doc-seo-86271-105":29,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},86271,1099514068365,"Aurelia","https://ap-avatar.wpscdn.com/avatar/10000253d8d9f28188e?_k=1776742907772140068",6,"Technology","MemExchange: Cloud-Scale Memory Trading","MemExchange addresses inefficient cloud memory allocation by dynamically right-sizing multi-tenant in-memory caching. Cloud providers over-provision memory for peak demand, creating cluster-wide underutilization, while memory-constrained tenants suffer higher cache miss rates despite idle capacity elsewhere. MemExchange reallocates idle memory between tenants across physical nodes using RDMA, guided by marginal-utility-based allocation derived from online Miss Ratio Curve estimation. It includes the MemExchange Tracker Communication protocol to coordinate reallocation and enable one-sided RDMA without remote CPU involvement, implemented in Memcached and evaluated at scale.","MemExchange: Cloud-Scale Memory Trading  \nAmirHossein Seyri  \nUniversity of Illinois Chicago Chicago, IL, USA [amirhossein.seyri@gmail.com](amirhossein.seyri@gmail.com)  \nAbhisek Pan  \nMicrosoft Redmond, WA, USA [abpan@microsoft.com](abpan@microsoft.com)  \nBalajee Vamanan  \nUniversity of Illinois Chicago Chicago, IL, USA [bvamanan@uic.edu](bvamanan@uic.edu)  \narXiv :2607 . 11579v1 [ cs .DC] 13 Jul 2026  \nAbstract  \nTo handle unpredictable workloads, cloud providers typically over-provision memory to meet peak demand, resulting insubstantial underutilization across datacenter clusters. At the same time, memory-constrained tenants may suffer elevated cache miss rates, even when idle capacity remains stranded elsewhere in the infrastructure. MemExchange is a clusterwide, multi-tenant memory management system that dynamically right-sizes in-memory caching tenants according to workload demand. Leveraging marginal-utility–based allocation derived from online Miss Ratio Curve (MRC) estimation, MemExchange redistributes idle memory between tenants across physical nodes using RDMA. This approach transforms the dedicated caching memory scattered across servers into a logically aggregated pool, enabling cross-node memory exchange without centralized coordination or forced tenant co-location. To support efficient remote access, we design the MemExchange Tracker Communication (MTC) protocol, an application-layer mechanism that coordinates memory reallocation and enables one-sided RDMA operations without involving remote CPUs. We implement MemExchange in Memcached and evaluate it through microbenchmarks, medium and rack-scale deployments of up to 100 CloudLab servers. Our results show up to 2. 3× lower remote-access overhead compared to TCP-based designs, a 13% increase in cluster-wide memory utilization at rack scale, and up to 63% reduction in miss rate for memory-constrained tenants under skewed workloads.  \n1 Introduction  \nMemory caches and key-value stores are essential components of modern web services. By storing frequently accessed objects in DRAM, they reduce backend database load and improve request latency by orders of magnitude. A cache hit is served directly from memory, whereas a cache miss requires retrieving the item from a backend database, often multiple network hops away and backed by storage significantly slower than DRAM. Even small improvements in cache hit rate can have tremendous impact; for example, a 1% increase in hit rate can reduce request latency by 35%[11] . Maintaining high hit rates is therefore critical for user-facing services.  \nAchieving high hit rates requires the cache’s working set to fit within its allocated memory. However, working set sizes are often dynamic and may exceed the statically provisioned memory assigned to a tenant [10, 11, 41, 42, 44] .  \nStatic allocation and manual right-sizing are inefficient for such workloads and frequently lead to over-provisioning. As a result, some tenants suffer from memory pressure and elevated miss rates, while others hold idle memory.  \nThis imbalance is amplified in multi-tenant cloud environments. Commercial cloud providers such as Microsoft Azure and AWS offer caching as a service (e.g., Redis-based offerings [33, 40]) . Multiple tenants are co-located on shared infrastructure [22, 44], which improves consolidation but leads to rigid per-tenant memory quotas. Despite extensive work on multi-tenant cache management [11, 12, 38, 41, 42], most prior systems operate within a single physical server and cannot reclaim idle memory across the cluster.  \nMeanwhile, cluster-wide memory utilization in large-scale datacenters often remains between 40% and 60%[29], leaving substantial idle capacity distributed across thousands of servers. This underutilization stems primarily from conservative over-provisioning: tenants are allocated memory based on peak demand to satisfy Service Level Objectives (SLOs), even though their steady-state working sets are frequently muc","cbCainxSYKz3kJkX","https://ap.wps.com/l/cbCainxSYKz3kJkX","pdf",706159,1,16,"English","en",105,"# Introduction\n## Background: cache misses and working sets\n## Problem: static quotas and stranded idle memory\n## Solution approach: cluster-wide memory redistribution with RDMA\n## Paper contributions and evaluation","[{\"question\":\"What problem does MemExchange target in cloud-scale caching?\",\"answer\":\"It targets the imbalance where providers over-provision memory for peak demand, leaving idle capacity stranded across the cluster while some tenants face memory pressure and elevated cache miss rates.\"},{\"question\":\"How does MemExchange decide how much memory each tenant should receive?\",\"answer\":\"It uses marginal-utility-based allocation derived from online Miss Ratio Curve (MRC) estimation to right-size in-memory caching tenants according to current workload demand.\"},{\"question\":\"How does MemExchange move memory across nodes efficiently?\",\"answer\":\"MemExchange uses RDMA to redistribute idle memory between tenants across physical nodes, treating remote memory as an overflow tier while keeping fast local access for hot data.\"}]",1784209945,40,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"memexchange-cloud-scale-memory-trading","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/technology/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/memexchange-cloud-scale-memory-trading/86271/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does MemExchange target in cloud-scale caching?","Question",{"text":75,"@type":76},"It targets the imbalance where providers over-provision memory for peak demand, leaving idle capacity stranded across the cluster while some tenants face memory pressure and elevated cache miss rates.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does MemExchange decide how much memory each tenant should receive?",{"text":80,"@type":76},"It uses marginal-utility-based allocation derived from online Miss Ratio Curve (MRC) estimation to right-size in-memory caching tenants according to current workload demand.",{"name":82,"@type":73,"acceptedAnswer":83},"How does MemExchange move memory across nodes efficiently?",{"text":84,"@type":76},"MemExchange uses RDMA to redistribute idle memory between tenants across physical nodes, treating remote memory as an overflow tier while keeping fast local access for hot data.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,113,117,122,127,130,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":111,"slug":112},50,"technology",{"id":114,"doc_module":4,"doc_module_name":45,"category_name":115,"show_sort_weight":28,"slug":116},7,"Healthcare","healthcare",{"id":118,"doc_module":4,"doc_module_name":45,"category_name":119,"show_sort_weight":120,"slug":121},8,"Research & Report",30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]