Apache Mesos, a two-level resource scheduler, provides resource sharing across multiple users in a multi-tenant clustered environment. Computational resources (i.e., CPU, memory, disk, etc.) are distributed according to the Dominant Resource Fairness (DRF) policy. Mesos frameworks (users) receive resources based on their current usage and are responsible for scheduling their tasks within the allocation. We have observed that multiple frameworks can cause fairness imbalance in a multi-user environment. For example, a greedy framework consuming more than its fair share of resources can deny resource fairness to others. The user with the least Dominant Share is considered first by the DRF module to get its resource allocation. However, the default DRF implementation, in Apache Mesos' Master allocation module, does not consider the overall resource demands of the tasks in the queue for each user/framework. This lack of awareness can lead to poor performance as users without any pending task may receive more resource offers, and users with a queue of pending tasks can starve due to their high dominant shares. In a multi-tenant environment, the characteristics of frameworks and workloads must be understood by cluster managers to be able to define fairness based on not only resource share but also resource demand and queue wait time. We have developed a policy driven queue manager, Tromino, for an Apache Mesos cluster where tasks for individual frameworks can be scheduled based on each framework's overall resource demands and current resource consumption. Dominant Share and demand awareness of Tromino and scheduling based on these attributes can reduce (1) the impact of unfairness due to a framework specific configuration, and (2) unfair waiting time due to higher resource demand in a pending task queue. In the best case, Tromino can significantly reduce the average waiting time of a framework by using the proposed Demand-DRF aware policy.
more »
« less
Exploring the Fairness and Resource Distribution in an Apache Mesos Environment
Apache Mesos, a cluster-wide resource manager, is widely deployed in massive scale at several Clouds and Data Centers. Mesos aims to provide high cluster utilization via fine grained resource co-scheduling and resource fairness among multiple users through Dominant Resource Fairness (DRF) based allocation. DRF takes into account different resource types (CPU, Memory, Disk I/O) requested by each application and determines the share of each cluster resource that could be allocated to the applications. Mesos has adopted a two-level scheduling policy: (1) DRF to allocate resources to competing frameworks and (2) task level scheduling by each framework for the resources allocated during the previous step. We have conducted experiments in a local Mesos cluster when used with frameworks such as Apache Aurora, Marathon, and our own framework Scylla, to study resource fairness and cluster utilization. Experimental results show how informed decision regarding second level scheduling policy of frameworks and attributes like offer holding period, offer refusal cycle and task arrival rate can reduce unfair resource distribution. Bin-Packing scheduling policy on Scylla with Marathon can reduce unfair allocation from 38% to 3%. By reducing unused free resources in offers we bring down the unfairness from to 90% to 28%. We also show the effect of task arrival rate to reduce the unfairness from 23% to 7%.
more »
« less
- Award ID(s):
- 1740263
- NSF-PAR ID:
- 10069273
- Date Published:
- Journal Name:
- 2018 IEEE 11th International Conference on Cloud Computing (CLOUD)
- Page Range / eLocation ID:
- 434 to 441
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
We first consider the static problem of allocating resources to (i.e., scheduling) multiple distributed application frameworks, possibly with different priorities and server preferences, in a private cloud with heterogeneous servers. Several fair scheduling mechanisms have been proposed for this purpose. We extend prior results on max-min fair (MMF) and proportional fair (PF) scheduling to this constrained multiresource and multiserver case for generic fair scheduling criteria. The task efficiencies (a metric related to proportional fairness) of max- min fair allocations found by progressive filling are compared by illustrative examples. In the second part of this paper, we consider the online problem (with framework churn) by implementing variants of these schedulers in Apache Mesos using progressive filling to dynamically approximate max-min fair allocations. We evaluate the implemented schedulers in terms of overall execution time of realistic distributed Spark workloads. Our experiments show that resource efficiency is improved and execution times are reduced when the scheduler is “server specific” or when it leverages characterized required resources of the workloads (when known).more » « less
-
Obeid, Iyad ; Selesnick, Ivan ; Picone, Joseph (Ed.)The goal of this work was to design a low-cost computing facility that can support the development of an open source digital pathology corpus containing 1M images [1]. A single image from a clinical-grade digital pathology scanner can range in size from hundreds of megabytes to five gigabytes. A 1M image database requires over a petabyte (PB) of disk space. To do meaningful work in this problem space requires a significant allocation of computing resources. The improvements and expansions to our HPC (highperformance computing) cluster, known as Neuronix [2], required to support working with digital pathology fall into two broad categories: computation and storage. To handle the increased computational burden and increase job throughput, we are using Slurm [3] as our scheduler and resource manager. For storage, we have designed and implemented a multi-layer filesystem architecture to distribute a filesystem across multiple machines. These enhancements, which are entirely based on open source software, have extended the capabilities of our cluster and increased its cost-effectiveness. Slurm has numerous features that allow it to generalize to a number of different scenarios. Among the most notable is its support for GPU (graphics processing unit) scheduling. GPUs can offer a tremendous performance increase in machine learning applications [4] and Slurm’s built-in mechanisms for handling them was a key factor in making this choice. Slurm has a general resource (GRES) mechanism that can be used to configure and enable support for resources beyond the ones provided by the traditional HPC scheduler (e.g. memory, wall-clock time), and GPUs are among the GRES types that can be supported by Slurm [5]. In addition to being able to track resources, Slurm does strict enforcement of resource allocation. This becomes very important as the computational demands of the jobs increase, so that they have all the resources they need, and that they don’t take resources from other jobs. It is a common practice among GPU-enabled frameworks to query the CUDA runtime library/drivers and iterate over the list of GPUs, attempting to establish a context on all of them. Slurm is able to affect the hardware discovery process of these jobs, which enables a number of these jobs to run alongside each other, even if the GPUs are in exclusive-process mode. To store large quantities of digital pathology slides, we developed a robust, extensible distributed storage solution. We utilized a number of open source tools to create a single filesystem, which can be mounted by any machine on the network. At the lowest layer of abstraction are the hard drives, which were split into 4 60-disk chassis, using 8TB drives. To support these disks, we have two server units, each equipped with Intel Xeon CPUs and 128GB of RAM. At the filesystem level, we have implemented a multi-layer solution that: (1) connects the disks together into a single filesystem/mountpoint using the ZFS (Zettabyte File System) [6], and (2) connects filesystems on multiple machines together to form a single mountpoint using Gluster [7]. ZFS, initially developed by Sun Microsystems, provides disk-level awareness and a filesystem which takes advantage of that awareness to provide fault tolerance. At the filesystem level, ZFS protects against data corruption and the infamous RAID write-hole bug by implementing a journaling scheme (the ZFS intent log, or ZIL) and copy-on-write functionality. Each machine (1 controller + 2 disk chassis) has its own separate ZFS filesystem. Gluster, essentially a meta-filesystem, takes each of these, and provides the means to connect them together over the network and using distributed (similar to RAID 0 but without striping individual files), and mirrored (similar to RAID 1) configurations [8]. By implementing these improvements, it has been possible to expand the storage and computational power of the Neuronix cluster arbitrarily to support the most computationally-intensive endeavors by scaling horizontally. We have greatly improved the scalability of the cluster while maintaining its excellent price/performance ratio [1].more » « less
-
more » « less
Summary Fair allocation has been studied intensively in both economics and computer science. Many existing mechanisms that consider fairness of resource allocation focus on a single resource. With the advance of cloud computing that centralizes multiple types of resources under one shared platform, multi‐resource allocation has come into the spotlight. In fact, fair/efficient multi‐resource allocation has become a fundamental problem in any shared computer system. The widely used solution is to partition resources into bundles that contain fixed amounts of different resources, so that multiple resources are abstracted as a single resource. However, this abstraction cannot satisfy different demands from heterogeneous users, especially on ensuring fairness among users competing for resources with different capacity limits. A promising approach to this problem is dominant resource fairness (DRF), which tries to equalize each user's dominant share (share of a user's most highly demanded resource, that is, the largest fraction of any resource that the user has required for a task), but this method may still suffer from significant loss of efficiency (i.e., some resources are underused). This article develops a new allocation mechanism based on DRF aiming to balance fairness and efficiency. We consider fairness not only in terms of a user's dominant resource, but also in another resource dimension which is secondarily desired by this user. We call this allocation mechanism 2‐dominant resource fairness (2‐DF). Then, we design a non‐trivial on‐line algorithm to find a 2‐DF allocation and extend this concept to
k ‐dominant resource fairness (k ‐DF). -
null (Ed.)Contemporary GPUs support multiple kernels to run concurrently on the same streaming multiprocessors (SMs). Recent studies have demonstrated that such concurrent kernel execution (CKE) improves both resource utilization and computational throughput. Most of the prior works focus on partitioning the GPU resources at the cooperative thread array (CTA) level or the warp scheduler level to improve CKE. However, significant performance slowdown and unfairness are observed when latency-sensitive kernels co-run with bandwidth-intensive ones. The reason is that bandwidth over-subscription from bandwidth-intensive kernels leads to much aggravated memory access latency, which is highly detrimental to latency-sensitive kernels. Even among bandwidth-intensive kernels, more intensive kernels may unfairly consume much higher bandwidth than less-intensive ones. In this article, we first make a case that such problems cannot be sufficiently solved by managing CTA combinations alone and reveal the fundamental reasons. Then, we propose a coordinated approach for CTA combination and bandwidth partitioning. Our approach dynamically detects co-running kernels as latency sensitive or bandwidth intensive. As both the DRAM bandwidth and L2-to-L1 Network-on-Chip (NoC) bandwidth can be the critical resource, our approach partitions both bandwidth resources coordinately along with selecting proper CTA combinations. The key objective is to allocate more CTA resources for latency-sensitive kernels and more NoC/DRAM bandwidth resources to NoC-/DRAM-intensive kernels. We achieve it using a variation of dominant resource fairness (DRF). Compared with two state-of-the-art CKE optimization schemes, SMK [52] and WS [55], our approach improves the average harmonic speedup by 78% and 39%, respectively. Even compared to the best possible CTA combinations, which are obtained from an exhaustive search among all possible CTA combinations, our approach improves the harmonic speedup by up to 51% and 11% on average.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/CLOUD.2018.00061
