Michael Stumm

Michael Stumm

Auto-generated profile, based on publicly available information

Michael Stumm

University of Toronto

Research Focus & Metrics

Academic Career Length

32Years (1995-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
1 March 1995 in The Journal of Supercomputing
Verified
Journal Article
Closed access
PDF file
RURonald C. Unrau
Ronald C. Unrau
OKOrran Krieger
Orran Krieger
BGBenjamin Gamsa
Benjamin Gamsa
MSMichael Stumm
Michael Stumm

30 November 2022 in ACM Transactions on Storage
Verified
Journal Article
Open access
RLRuibin Li
Ruibin Li
XRXiang Ren
Xiang Ren
XZXu Zhao
Xu Zhao
SHSiwei He
Siwei He
MSMichael Stumm
Michael Stumm
DYDing Yuan
Ding Yuan
Persistent byte-addressable memory (PM) is poised to become prevalent in future computer systems. PMs are significantly faster than disk storage, and accesses to PMs are governed by the Memory Management Unit (MMU) just as accesses with volatile RAM. These unique characteristics shift the bottleneck from I/O to operations such as block address lookup—for example, in write workloads, up to 45% of the overhead in ext4-DAX is due to building and searching extent trees to translate file offsets to addresses on persistent memory. We propose a novel contiguous file system, ctFS, that eliminates most of the overhead associated with indexing structures such as extent trees in the file system. ctFS represents each file as a contiguous region of virtual memory, hence a lookup from the file offset to the address is simply an offset operation, which can be efficiently performed by the hardware MMU at a fraction of the cost of software-maintained indexes. Evaluating ctFS on real-world workloads such as LevelDB shows it outperforms ext4-DAX and SplitFS by 3.6× and 1.8×, respectively.

1 December 1995 in The Journal of Supercomputing
Verified
Journal Article
Closed access
PDF file
KFKeith Farkas
Keith Farkas
ZVZvonko Vranesic
Zvonko Vranesic
MSMichael Stumm
Michael Stumm

22 April 2024
Verified
Conference Paper
Open access
SSSari Sultan
Sari Sultan
KSKia Shakiba
Kia Shakiba
ALAlbert Lee
Albert Lee
PCPaul Chen
Paul Chen
MSMichael Stumm
Michael Stumm
In-memory caches play a pivotal role in optimizing distributed systems by significantly reducing query response times. Correctly sizing these caches is critical, especially considering that prominent organizations use terabytes and even petabytes of DRAM for these caches. The Miss Ratio Curve (MRC) and Working Set Size (WSS) are the most widely used tools for sizing these caches. Modern cache workloads employ Time-to-Live (TTL) limits to define the lifespan of cached objects, a feature essential for ensuring data freshness and adhering to regulations like GDPR. Surprisingly, none of the existing MRC and WSS tools accommodate TTLs. Based on 28 real-world cache workloads that contain 113 billion accesses, we show that taking TTL limits into consideration allows an average of 69% (and up to 99%) lower memory footprint for in-memory caches without a degradation in the hit rate. This paper describes how TTLs can be integrated into today's most important MRC generation and WSS estimation algorithms. We also describe how the widely used HyperLogLog (HLL) cardinality estimator can be extended to accommodate TTLs, and show how it can be used to efficiently estimate the WSS. Our extended algorithms maintain comparable performance levels to the original algorithms. All our extended approximate algorithms are efficient, run in constant space, and enable more resource-efficient and cost-effective cache management.

10 July 2025
Verified
Conference Paper
Open access
KSKia Shakiba
Kia Shakiba
MSMichael Stumm
Michael Stumm