David H. K. Kim

POET: a portable approach to minimizing energy under soft real-time constraints

Embedded real-time systems must meet timing constraints while minimizing energy consumption. To this end, many energy optimizations are introduced for specific platforms or specific applications. These solutions are not portable, however, and when the application or the platform change, these solutions must be redesigned. Portable techniques are hard to develop due to the varying tradeoffs experienced with different application/platform configurations. This paper addresses the problem of finding and exploiting general tradeoffs, using control theory and mathematical optimization to achieve energy minimization under soft real-time application constraints. The paper presents POET, an open-source C library and runtime system that takes a specification of the platform resources and optimizes the application execution. We test POETs ability to portably deliver predictable timing and energy reduction on two embedded systems with different tradeoff spaces - the first with a mobile Intel Haswell processor, and the second with an ARM big.LITTLE System on Chip. POET achieves the desired latency goals with small error while consuming, on average, only 1.3% more energy than the dynamic optimal oracle on the Haswell and 2.9% more on the ARM. We believe this open-source, library-based approach to resource management will simplify the process of writing portable, energy-efficient code for embedded systems.

Racing and Pacing to Idle: Theoretical and Empirical Analysis of Energy Optimization Heuristics

The problem of minimizing energy for a performance constraint (e.g., Real-time deadline or quality-of-service requirement) has been widely studied, both in theory and in practice. Theoretical models have indicated large potential energy savings, but practical concerns have made these savings hard to realize. Instead, practitioners often rely on heuristic solutions, which achieve good results in practice but tend to be system-specific in efficacy. An example is the race-to-idle heuristic, which makes all resources available until a task completes and then idles. Theory predicts poor energy savings, but practitioners have reported good empirical results. To help bridge the gap between theory and practice, this paper presents a geometrical framework for analyzing the energy optimality of resource allocation under performance constraints. The geometry of the problem allows us to derive an optimal strategy and three commonly used heuristics: 1) race-to-idle, 2) pace-to-idle a near-optimal idling strategy, and 3) no-idle which never idles. We then implement all strategies and test them empirically for seven benchmarks on four different multicore systems, including both x86 and ARM. We find that race-to-idle is near optimal on older systems, but can consume as much as 3× more energy than the optimal strategy. In contrast, pace-to-idle is never more than 12% worse than optimal.

Improved approximation for node-disjoint paths in planar graphs

We study the classical Node-Disjoint Paths (NDP) problem: given an n-vertex graph G and a collection =(s1,t1),…,(sk,tk) of pairs of vertices of G called demand pairs, find a maximum-cardinality set of node-disjoint paths connecting the demand pairs. NDP is one of the most basic routing problems, that has been studied extensively. Despite this, there are still wide gaps in our understanding of its approximability: the best currently known upper bound of O(√n) on its approximation ratio is achieved via a simple greedy algorithm, while the best current negative result shows that the problem does not have a better than Ω(log1/2−δn)-approximation for any constant δ, under standard complexity assumptions. Even for planar graphs no better approximation algorithms are known, and to the best of our knowledge, the best negative bound is APX-hardness. Perhaps the biggest obstacle to obtaining better approximation algorithms for NDP is that most currently known approximation algorithms for this type of problems rely on the standard multicommodity flow relaxation, whose integrality gap is Ω(√n) for NDP, even in planar graphs. In this paper, we break the barrier of O(√n) on the approximability of NDP in planar graphs and obtain an Õ(n9/19)-approximation. We introduce a new linear programming relaxation of the problem, and a number of new techniques, that we hope will be helpful in designing more powerful algorithms for this and related problems.

On Approximating Node-Disjoint Paths in Grids

In the Node-Disjoint Paths (NDP) problem, the input is an undirected n-vertex graph G, and a collection {(s_1,t_1),...,(s_k,t_k)} of pairs of vertices called demand pairs. The goal is to route the largest possible number of the demand pairs (s_i,t_i), by selecting a path connecting each such pair, so that the resulting paths are node-disjoint. NDP is one of the most basic and extensively studied routing problems. Unfortunately, its approximability is far from being well-understood: the best current upper bound of O(sqrt(n)) is achieved via a simple greedy algorithm, while the best current lower bound on its approximability is Omega(log^{1/2-\delta}(n)) for any constant delta. Even for seemingly simpler special cases, such as planar graphs, and even grid graphs, no better approximation algorithms are currently known. A major reason for this impasse is that the standard technique for designing approximation algorithms for routing problems is LP-rounding of the standard multicommodity flow relaxation of the problem, whose integrality gap for NDP is Omega(sqrt(n)) even on grid graphs. Our main result is an O(n^{1/4} * log(n))-approximation algorithm for NDP on grids. We distinguish between demand pairs with both vertices close to the grid boundary, and pairs where at least one of the two vertices is far from the grid boundary. Our algorithm shows that when all demand pairs are of the latter type, the integrality gap of the multicommodity flow LP-relaxation is at most O(n^{1/4} * log(n)), and we deal with demand pairs of the former type by other methods. We complement our upper bounds by proving that NDP is APX-hard on grid graphs.

New hardness results for routing on disjoint paths

In the classical Node-Disjoint Paths (NDP) problem, the input consists of an undirected n-vertex graph G, and a collection M={(s1,t1),…,(sk,tk)} of pairs of its vertices, called source-destination, or demand, pairs. The goal is to route the largest possible number of the demand pairs via node-disjoint paths. The best current approximation for the problem is achieved by a simple greedy algorithm, whose approximation factor is O(√n), while the best current negative result is an Ω(log1/2-δn)-hardness of approximation for any constant δ, under standard complexity assumptions. Even seemingly simple special cases of the problem are still poorly understood: when the input graph is a grid, the best current algorithm achieves an Õ(n1/4)-approximation, and when it is a general planar graph, the best current approximation ratio of an efficient algorithm is Õ(n9/19). The best currently known lower bound for both these versions of the problem is APX-hardness. In this paper we prove that NDP is 2Ω(√logn)-hard to approximate, unless all problems in NP have algorithms with running time nO(logn). Our result holds even when the underlying graph is a planar graph with maximum vertex degree 4, and all source vertices lie on the boundary of a single face (but the destination vertices may lie anywhere in the graph). We extend this result to the closely related Edge-Disjoint Paths problem, showing the same hardness of approximation ratio even for sub-cubic planar graphs with all sources lying on the boundary of a single face.

Portable Multicore Resource Management for Applications with Performance Constraints

Many modern software applications have performance requirements, like mobile and embedded systems that must keep up with sensor data, or web services that must return results to users within an acceptable latency bound. For such applications, the goal is not to run as fast as possible, but to meet their performance requirements with minimal resource usage, the key resource in most systems being energy. Heuristic solutions have been proposed to minimize energy under a performance constraint, but recent studies show that these approaches are not portable - heuristics that are near-optimal on one system can waste integer factors of energy on others. The POET library and runtime system provides a portable method for resource management that achieves near-optimal energy consumption while meeting soft real-time constraints across a range of devices. Although POET was originally designed and tested on embedded and mobile platforms, in this paper we evaluate it on a manycore server-class system. The larger scale of manycore systems adds some overhead to adjusting resource allocations, but POET still meets timing constraints and achieves near-optimal energy consumption. We demonstrate that POET achieves portable energy efficiency on platforms ranging from low-power ARM big.LITTLE architectures to powerful x86 server-class systems.

Approximation Algorithms for Scheduling with Resource and Precedence Constraints

We study non-preemptive scheduling problems on identical parallel machines and uniformly related machines under both resource constraints and general precedence constraints between jobs. Our first result is an O(logn)-approximation algorithm for the objective of minimizing the makespan on parallel identical machines under resource and general precedence constraints. We then use this result as a subroutine to obtain an O(logn)-approximation algorithm for the more general objective of minimizing the total weighted completion time on parallel identical machines under both constraints. Finally, we present an O(logm logn)-approximation algorithm for scheduling under these constraints on uniformly related machines. We show that these results can all be generalized to include the case where each job has a release time. This is the first upper bound on the approximability of this class of scheduling problems where both resource and general precedence constraints must be satisfied simultaneously.

Deletion Without Rebalancing in Binary Search Trees

We address the vexing issue of deletions in balanced trees. Rebalancing after a deletion is generally more complicated than rebalancing after an insertion. Textbooks neglect deletion rebalancing, and many B-tree--based database systems do not do it. We describe a relaxation of AVL trees in which rebalancing is done after insertions but not after deletions, yet worst-case access time remains logarithmic in the number of insertions. For any application of balanced trees in which the number of updates is polynomial in the tree size, our structure offers performance competitive with that of classical balanced trees. With the addition of periodic rebuilding, the performance of our structure is theoretically superior to that of many, if not all, classic balanced tree structures. Our structure needs lg lg m+ 1 bits of balance information per node, where m is the number of insertions and lg is the base-two logarithm, or lg lg n+ O(1) with periodic rebuilding, where n is the number of nodes. An insertion takes up to two rotations and O(1) amortized time, not counting the time to find the insertion position. This is the same as in standard AVL trees. Using an analysis that relies on an exponential potential function, we show that rebalancing steps occur with a frequency that is exponentially small in the height of the affected node. Our techniques apply to other types of balanced trees, notably B-trees, as we show in a companion article, and particularly red-black trees, which can be viewed as a special case of B-trees.

Almost polynomial hardness of node-disjoint paths in grids

In the classical Node-Disjoint Paths (NDP) problem, we are given an n-vertex graph G=(V,E), and a collection M={(s1,t1),…,(sk,tk)} of pairs of its vertices, called source-destination, or demand pairs. The goal is to route as many of the demand pairs as possible, where to route a pair we need to select a path connecting it, so that all selected paths are disjoint in their vertices. The best current algorithm for NDP achieves an O(√n)-approximation, while, until recently, the best negative result was a factor Ω(log1/2−єn)-hardness of approximation, for any constant є, unless NP ⊆ ZPTIME(npoly logn). In a recent work, the authors have shown an improved 2Ω(√logn)-hardness of approximation for NDP, unless NP⊆ DTIME(nO(logn)), even if the underlying graph is a subgraph of a grid graph, and all source vertices lie on the boundary of the grid. Unfortunately, this result does not extend to grid graphs. The approximability of the NDP problem on grid graphs has remained a tantalizing open question, with the best current upper bound of Õ(n1/4), and the best current lower bound of APX-hardness. In a recent work, the authors showed a 2Õ(√logn)-approximation algorithm for NDP in grid graphs, if all source vertices lie on the boundary of the grid – a result that can be seen as suggesting that a sub-polynomial approximation may be achievable for NDP in grids. In this paper we show that this is unlikely to be the case, and come close to resolving the approximability of NDP in general, and of NDP in grids in particular. Our main result is that NDP is 2Ω(log1−є n)-hard to approximate for any constant є, assuming that NP⊈RTIME(npoly logn), and that it is nΩ (1/(loglogn)2)-hard to approximate, assuming that for some constant δ>0, NP ⊈RTIME(2nδ). These results hold even for grid graphs and wall graphs, and extend to the closely related Edge-Disjoint Paths problem, even in wall graphs. Our hardness proof performs a reduction from the 3COL(5) problem to NDP, using a new graph partitioning problem as a proxy. Unlike the more standard approach of employing Karp reductions to prove hardness of approximation, our proof is a Cook-type reduction, where, given an input instance of 3COL(5), we produce a large number of instances of NDP, and apply an approximation algorithm for NDP to each of them. The construction of each new instance of NDP crucially depends on the solutions to the previous instances that were found by the approximation algorithm.

POET: A Portable Approach to Minimizing Energy Under Soft Real-time Constraints

Embedded real-time systems must meet timing constraints while minimizing energy consumption. To this end, many energy optimizations are introduced for specific platforms or specific applications. These solutions are not portable, however, and when the application or the platform change, these solutions must be redesigned. Portable techniques are hard to develop due to the varying tradeoffs experienced with different application/platform configurations. This paper addresses the problem of finding and exploiting general tradeoffs, using control theory and mathematical optimization to achieve energy minimization under soft real-time application constraints. The paper presents POET, an open-source C library and runtime system that takes a specification of the platform resources and optimizes the application execution. We test POETs ability to portably deliver predictable timing and energy reduction on two embedded systems with different tradeoff spaces - the first with a mobile Intel Haswell processor, and the second with an ARM big.LITTLE System on Chip. POET achieves the desired latency goals with small error while consuming, on average, only 1.3% more energy than the dynamic optimal oracle on the Haswell and 2.9% more on the ARM. We believe this open-source, library-based approach to resource management will simplify the process of writing portable, energy-efficient code for embedded systems.