Scalable pCT Image Reconstruction Delivered as a Cloud Service

Chard, Ryan; Madduri, Ravi; Karonis, Nicholas T.; Chard, Kyle; Duffin, Kirk L.; nez, Caesar E. Ordo; Uram, Thomas D.; Fleischauer, Justin; Foster, Ian; Papka, Michael E.; Winans, John R.

doi:10.1109/tcc.2015.2457423

“…(a) Design and development of a cloud-based service for reconstructing proton computed tomography (pCT) images on-demand [17]. I implemented the service which utilises more than 100 GPU-enabled cloud instances to reconstruct medical images on-demand.…”

Section: Contributionsmentioning

confidence: 99%

“…The first service (cf. Section 3.1) is a scalable Proton Computed Tomography (pCT) image reconstruction service [17]. The pCT service provides on-demand medical image reconstructions by employing more than 100 GPU-enabled cluster instances to reconstruct 100GB datasets.…”

Section: Science As a Service -Use Casesmentioning

confidence: 99%

“…The provisioning algorithm has two primary phases. The first phase involves the management of existing infrastructure (lines [9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24][25][26]. This phase first checks the state of instance requests to identify requests that have been fulfilled and either cancel or repurpose unnecessary requests (e.g., a previously idle job may have been deployed to a newly idle instance).…”

Section: Provisioning Algorithmmentioning

confidence: 99%

See 1 more Smart Citation

Improving the Performance of Cloud-based Scientific Services

Chard¹

Self Cite

0

View full text Add to dashboard Cite

<p>Cloud computing provides access to a large scale set of readily available computing resources at the click of a button. The cloud paradigm has commoditised computing capacity and is often touted as a low-cost model for executing and scaling applications. However, there are significant technical challenges associated with selecting, acquiring, configuring, and managing cloud resources which can restrict the efficient utilisation of cloud capabilities. Scientific computing is increasingly hosted on cloud infrastructure—in which scientific capabilities are delivered to the broad scientific community via Internet-accessible services. This migration from on-premise to on-demand cloud infrastructure is motivated by the sporadic usage patterns of scientific workloads and the associated potential cost savings without the need to purchase, operate, and manage compute infrastructure—a task that few scientific users are trained to perform. However, cloud platforms are not an automatic solution. Their flexibility is derived from an enormous number of services and configuration options, which in turn result in significant complexity for the user. In fact, naïve cloud usage can result in poor performance and excessive costs, which are then directly passed on to researchers. This thesis presents methods for developing efficient cloud-based scientific services. Three real-world scientific services are analysed and a set of common requirements are derived. To address these requirements, this thesis explores automated and scalable methods for inferring network performance, considers various trade-offs (e.g., cost and performance) when provisioning instances, and profiles application performance, all in heterogeneous and dynamic cloud environments. Specifically, network tomography provides the mechanisms to infer network performance in dynamic and opaque cloud networks; cost-aware automated provisioning approaches enable services to consider, in real-time, various trade-offs such as cost, performance, and reliability; and automated application profiling allows a huge search space of applications, instance types, and configurations to be analysed to determine resource requirements and application performance. Finally, these contributions are integrated into an extensible and modular cloud provisioning and resource management service called SCRIMP. Cloud-based scientific applications and services can subscribe to SCRIMP to outsource their provisioning, usage, and management of cloud infrastructures. Collectively, the approaches presented in this thesis are shown to provide order of magnitude cost savings and significant performance improvement when employed by production scientific services.</p>

show abstract

“…The projected cost for reconstructing various size datasets under different cluster configurations with both On-demand (solid line) and low Spot price (dashed line) is shown in Figure 3.3. Whereas, a dedicated cluster, such as Gaea, has an initial cost of approximately $800,000, and an annual cost of over $150,000 in staffing and operational expenses [17].…”

Section: Pct Reconstruction Servicementioning

confidence: 99%

“…As the capabilities of cloud providers have increased, cloud platforms have become a viable alternative to HPC platforms for hosting large-scale analyses [11,12,13,14]. The cloud is accessible and affordable [15], with low barriers of entry [16], and reduced operational costs [17] when compared to dedicated HPC platforms. It presents an opportunity to accelerate scientific discovery, providing researchers access to enormous on-demand and elastically scalable compute resources at the click of a button [18].…”

Section: Introductionmentioning

confidence: 99%

Improving the Performance of Cloud-based Scientific Services

Chard¹

Self Cite

0

View full text Add to dashboard Cite

<p>Cloud computing provides access to a large scale set of readily available computing resources at the click of a button. The cloud paradigm has commoditised computing capacity and is often touted as a low-cost model for executing and scaling applications. However, there are significant technical challenges associated with selecting, acquiring, configuring, and managing cloud resources which can restrict the efficient utilisation of cloud capabilities. Scientific computing is increasingly hosted on cloud infrastructure—in which scientific capabilities are delivered to the broad scientific community via Internet-accessible services. This migration from on-premise to on-demand cloud infrastructure is motivated by the sporadic usage patterns of scientific workloads and the associated potential cost savings without the need to purchase, operate, and manage compute infrastructure—a task that few scientific users are trained to perform. However, cloud platforms are not an automatic solution. Their flexibility is derived from an enormous number of services and configuration options, which in turn result in significant complexity for the user. In fact, naïve cloud usage can result in poor performance and excessive costs, which are then directly passed on to researchers. This thesis presents methods for developing efficient cloud-based scientific services. Three real-world scientific services are analysed and a set of common requirements are derived. To address these requirements, this thesis explores automated and scalable methods for inferring network performance, considers various trade-offs (e.g., cost and performance) when provisioning instances, and profiles application performance, all in heterogeneous and dynamic cloud environments. Specifically, network tomography provides the mechanisms to infer network performance in dynamic and opaque cloud networks; cost-aware automated provisioning approaches enable services to consider, in real-time, various trade-offs such as cost, performance, and reliability; and automated application profiling allows a huge search space of applications, instance types, and configurations to be analysed to determine resource requirements and application performance. Finally, these contributions are integrated into an extensible and modular cloud provisioning and resource management service called SCRIMP. Cloud-based scientific applications and services can subscribe to SCRIMP to outsource their provisioning, usage, and management of cloud infrastructures. Collectively, the approaches presented in this thesis are shown to provide order of magnitude cost savings and significant performance improvement when employed by production scientific services.</p>

show abstract