Download Time Analysis for Distributed Storage Codes With Locality and Availability

Aktas, Mehmet Fatih; Kadhe, Swanand; Soljanin, Emina; Sprintson, Alex

doi:10.1109/tcomm.2021.3067054

Cited by 6 publications

(27 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Heterogeneous data access has been previously considered in the context of hot data download (see e.g., [22]- [25]) and load balancing (see e.g., [26]). In the hot data download context, heterogeneity arises when one of the data objects is highly popular.…”

Section: B Related Previous Workmentioning

confidence: 99%

“…In the hot data download context, heterogeneity arises when one of the data objects is highly popular. Works such as [22]- [25] analyze the expected latency experienced by requests that are replicated across the hot object's recovery sets. Previous works on load balancing heterogeneous requests to coded systems that arises when a batch of simultaneous requests consists of different numbers of requests for each of the k objects.…”

Section: B Related Previous Workmentioning

confidence: 99%

“…Simple replication of objects, that is, storing identical copies, is allowed and, as we will see later, often to a certain extent desirable. It is worth to note here that the assumptions we make in our storage model are common in the prior work, see, e.g., [13], [14], [24], [25]. We denote the coded objects as c 1 , c 2 , .…”

Section: A Data Storage Modelmentioning

confidence: 99%

“…Instead of measuring the service cost in terms of the amount of data accessed to serve one request as captured by C(λ), an alternate metric is to consider the total computing time spent serving each request, as considered in [12], [25], [67]- [69]. These papers identify regimes where coded data access or computing incurs a lower total computing time than replicated systems.…”

Section: A Performance Analysis and Networking Problemsmentioning

confidence: 99%

See 3 more Smart Citations

Service Rate Region: A New Aspect of Coded Distributed System Design

Aktas

Joshi

Kadhe

et al. 2021

IEEE Trans. Inform. Theory

Self Cite

View full text Add to dashboard Cite

Erasure coding has been recognized as a powerful method to mitigate delays due to slow or straggling nodes in distributed systems. This work shows that erasure coding of data objects can flexibly handle skews in the request rates. Coding can help boost the service rate region, that is, increase the overall volume of data access requests that the system can handle. This paper aims to postulate the service rate region as an important consideration in the design of erasure-coded distributed systems.We highlight several open problems that can be grouped into two broad threads: 1) characterizing the service rate region of a given code and finding the optimal request allocation, and 2) designing the underlying erasure code for a given service rate region. As contributions along the first thread, we find the rate regions of maximum-distance-separable, locally repairable, and simplex codes. We show the effectiveness of hybrid codes that combine replication and erasure coding in terms of code design. We also discover fundamental connections between multi-set batch codes and the problem of maximizing the service rate region.

show abstract

Section: B Related Previous Workmentioning

confidence: 99%

Section: B Related Previous Workmentioning

confidence: 99%

Section: A Data Storage Modelmentioning

confidence: 99%

Section: A Performance Analysis and Networking Problemsmentioning

confidence: 99%

See 2 more Smart Citations

Service Rate Region: A New Aspect of Coded Distributed System Design

Aktas

Joshi

Kadhe

et al. 2021

IEEE Trans. Inform. Theory

Self Cite

View full text Add to dashboard Cite

show abstract

“…Although replication has been typically preferred due its simplicity, it can be expensive in terms of storage. Erasure codes have been shown to be effective in achieving various goals such as providing reliability against node failures (e.g., [1]), ensuring availability of stored content during high demand (e.g., [2]), enabling the recovery of a data object from multiple disjoint groups of nodes (e.g., [3]), and providing fast content download (e.g., [4]- [6]).…”

Section: Introduction Motivationmentioning

confidence: 99%

Efficient Storage Schemes for Desired Service Rate Regions

Kazemi

Kurz

Soljanin

et al. 2021

2020 IEEE Information Theory Workshop (ITW)

Self Cite

View full text Add to dashboard Cite

A major concern in cloud/edge storage systems is serving a large number of users simultaneously. The service rate region is introduced recently as an important performance metric for coded distributed systems, which is defined as the set of all data access requests that can be simultaneously handled by the system. This paper studies the problem of designing a coded distributed storage system storing k files where a desired service rate region R of the system is given and the goal is 1) to determine the minimum number of storage nodes n(R) for serving all demand vectors inside the set R and 2) to design the most storage-efficient redundancy scheme with the service rate region covering the set R. Towards this goal, we propose three general lower bounds for n(R). Also, for k = 2, we characterize n(R), i.e., we show that the proposed lower bounds are tight, via designing a novel storage-efficient redundancy scheme with n(R) storage nodes and the service rate region covering R.

show abstract