Coded Caching Schemes With Reduced Subpacketization From Linear Block Codes

Tang, Li; Ramamoorthy, Aditya

doi:10.1109/tit.2018.2800059

Cited by 109 publications

(91 citation statements)

References 27 publications

(78 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In that sense, we expect the Map cost of both coded schemes to be approximately r times higher than that of the uncoded implementation of the algorithm. Indeed, if we look at our scheme for r = 4 we see that 25.91 5.71 ≈ 4.54 is a good approximation to r. • The encoding time of the coded schemes (which is the time it takes so that all servers form the encoded packets that they will be transmitting afterwards) is not directly comparable to the packing of the uncoded approach which stores each intermediate value serially in a continuous memory array to ensure that a single TCP connection is initiated for each intermediate value. Nevertheless, we have a significant benefit over the prior scheme.…”

Section: Terasort Experimental Results and Discussionmentioning

confidence: 99%

See 1 more Smart Citation

Leveraging Coding Techniques for Speeding up Distributed Computing

Konstantinidis

Ramamoorthy

2018

2018 IEEE Global Communications Conference (GLOBECOM)

Self Cite

View full text Add to dashboard Cite

Distributed computing frameworks such as MapReduce are often used to process large computational jobs. They operate by partitioning each job into smaller tasks executed on different servers. The servers also need to exchange intermediate values to complete the computation. Experimental evidence suggests that this so-called Shuffle phase can be a significant part of the overall execution time for several classes of jobs. Prior work has demonstrated a natural tradeoff between computation and communication whereby running redundant copies of jobs can reduce the Shuffle traffic load, thereby leading to reduced overall execution times. For a single job, the main drawback of this approach is that it requires the original job to be split into a number of files that grows exponentially in the system parameters. When extended to multiple jobs (with specific function types), these techniques suffer from a limitation of a similar flavor, i.e., they require an exponentially large number of jobs to be executed. In practical scenarios, these requirements can significantly reduce the promised gains of the method. In this work, we show that a class of combinatorial structures called resolvable designs can be used to develop efficient coded distributed computing schemes for both the single and multiple job scenarios considered in prior work. We present both theoretical analysis and exhaustive experimental results (on Amazon EC2 clusters) that demonstrate the performance advantages of our method. For the single and multiple job cases, we obtain speed-ups of 4.69x (and 2.6x over prior work) and 4.31x over the baseline approach, respectively.

show abstract

Section: Terasort Experimental Results and Discussionmentioning

confidence: 99%

“…The set of blocks A SP C is given by the collection of all B i,l for 1 ≤ i ≤ k and 0 ≤ l ≤ q − 1 so that |A SP C | = kq. The following lemma (proved in [25]) shows that this construction always yields a resolvable design. Lemma 1.…”

Section: Definitionmentioning

confidence: 94%

Leveraging Coding Techniques for Speeding up Distributed Computing

Konstantinidis

Ramamoorthy

2018

2018 IEEE Global Communications Conference (GLOBECOM)

Self Cite

View full text Add to dashboard Cite

show abstract

“…A prior work, [4], had already initiated the study of connections between coded caching and combinatorial designs by looking at caching schemes constructed via resolvable designs, for which the authors provide a coding theoretic construction. We see the current work as being a further attempt to explore the applications of combinatorial designs for the purpose of designing practically useful coded caching schemes.…”

Section: Discussionmentioning

confidence: 99%

“…This idea has been further extended in [3] by using strong edge coloring of an associated bipartite graph. In [4], resolvable designs derived from linear block codes have been used to reduce subpacketization. All of these schemes offered reductions in subpacketization as compared to [1], at the cost of some increase in the rate, for constant memory fraction M N .…”

Section: Introductionmentioning

confidence: 99%

Coded Caching based on Combinatorial Designs

Agrawal

Sree

Krishnan

2019

2019 IEEE International Symposium on Information Theory (ISIT)

View full text Add to dashboard Cite

We consider the standard broadcast setup with a single server broadcasting information to a number of clients, each of which contains local storage (called cache) of some size, which can store some parts of the available files at the server. The centralized coded caching framework, introduced in [1], consists of a caching phase and a delivery phase, both of which are carefully designed in order to use the cache and the channel together optimally. Starting from [1], various combinatorial structures have been used to construct coded caching schemes. In this work, we propose a binary matrix model to construct the coded caching scheme. The ones in such a caching matrix indicate uncached subfiles at the users. Identity submatrices of the caching matrix represent transmissions in the delivery phase. Using this model, we then propose several novel constructions for coded caching based on the various types of combinatorial designs. While most of the schemes constructed in this work (based on existing designs) have a high cache requirement (uncached fraction being Θ( 1 √ K ) or Θ( 1 K ), K being the number of users), they provide a rate that is either constant or decreasing (O( 1 K )) with increasing K, and moreover require competitively small levels of subpacketization (being O(K i ), 1 ≤ i ≤ 3), which is an extremely important parameter in practical applications of coded caching. We mark this work as another attempt to exploit the well-developed theory of combinatorial designs for the problem of constructing caching schemes, utilizing the binary caching model we develop.

show abstract

“…The set of blocks A SP C is given by the collection of all B i,l for 1 ≤ i ≤ k and 0 ≤ l ≤ q − 1 so that |A SP C | = kq. The following lemma (see [14] for a proof in a different context) shows that this construction yields a resolvable design.…”

Section: Definitionmentioning

confidence: 94%

CAMR: Coded Aggregated MapReduce

Konstantinidis

Ramamoorthy

2019

2019 IEEE International Symposium on Information Theory (ISIT)

Self Cite

View full text Add to dashboard Cite

Many big data algorithms executed on MapReducelike systems have a shuffle phase that often dominates the overall job execution time. Recent work has demonstrated schemes where the communication load in the shuffle phase can be traded off for the computation load in the map phase. In this work, we focus on a class of distributed algorithms, broadly used in deep learning, where intermediate computations of the same task can be combined. Even though prior techniques reduce the communication load significantly, they require a number of jobs that grows exponentially in the system parameters. This limitation is crucial and may diminish the load gains as the algorithm scales. We propose a new scheme which achieves the same load as the state-of-the-art while ensuring that the number of jobs as well as the number of subfiles that the data set needs to be split into remain small.

show abstract

Coded Caching Schemes With Reduced Subpacketization From Linear Block Codes

Cited by 109 publications

References 27 publications

Leveraging Coding Techniques for Speeding up Distributed Computing

Leveraging Coding Techniques for Speeding up Distributed Computing

Coded Caching based on Combinatorial Designs

CAMR: Coded Aggregated MapReduce

Contact Info

Product

Resources

About