Mapping the SLA Landscape for High Performance Clouds

By Dr. Ivona Brandic

February 7, 2011

Cloud computing represents the convergence of several concepts in IT, ranging from virtualization, distributed application design, grid computing, and enterprise IT management–resulting to be a promising paradigm for on demand provision of ICT infrastructures.

During the past few years significant effort has been made in the sub-fields of cloud research, including the development of various federation mechanisms, cloud security, virtualization and service management techniques.

While a wealth of work has been accomplished to suit the technological development of clouds, there has yet been very little work done in the area of the market mechanisms that support them.

As we learned in the past, however (consider the case of grid technologies) appropriate market models for virtual goods, ease of use of those markets, low thresholds for entering the market for traders and buyers, and the appropriate processes for the definition and management of virtual goods have remained challenging issues. The way these topics are addressed will decide whether cloud computing will take root as a self-sustaining state-of-the-art technology.
 
The current cloud landscape is characterized by two market mechanisms: either users can select products from one of the big players with their sets of well-defined, but rigid offerings; or they rely on off-line relationships to cloud providers with niche products.

This division is especially marked in the area of HPC given the comprehensive special requirements needed, including specific security infrastructures, compliance to legal guidelines, massive scalability or support for parallel code execution, among others. HPC thus suffers from a low number of comparable choices, thus resulting in low liquidity of current cloud markets and provider/vendor lock in.

Sufficient market liquidity is essential for dynamic and open cloud markets. Liquid markets are characterized by a high number of matches for bids and offers. With the low market liquidity traders have the high risk of not being able to trade resources, while users have the risks of not being able to find suitable products.

A crucial factor in achieving high market liquidity is the existence of standardized goods. Virtual goods, as this is the case in clouds, however, exhibit high variability in product description. That means, that very similar or almost identical goods can be described in various ways with different attributes and parameters.

As shown in Table 1 below, computing resources traded in a PaaS fashion can be described through different non-standardized attributes, e.g., CPU cores, incoming bandwidth, processor types, required storage. Thus, high variability in the description of goods results again in low market liquidity. Another important characteristic of virtual goods is that they change and evolve over time following various technological trends. For example the attribute number of cores appeared just with the introduction of multi core architectures.

Table 1: Example SLA parameters

Incoming Bandwidth >10 MBit/s
Outgoing Bandwidth >12 MBit/s
Storage >1024 GB
Availability >99%
CPU Cores >16

Based on aforementioned observations, two challenging questions have been identified:

  • How can users’ demand and traders’ offers be channeled towards standardized products, which can evolve and adapt over time and reflect users’ needs and traders’ capabilities?
  • Which mechanism do we need to achieve sufficient market liquidity, where traders have high probability to sell their products and where users have sufficient probability to buy products they require.

To counteract this problem we make use of Service Level Agreements (SLAs), which are traditionally used to establish contracts between cloud traders and buyers.

Table 1 shows a typical SLA with the parameters and according values expressing non-functional requirements for the service usage. SLA templates represent popular SLA formats containing all attributes and parameters but without any values and are usually used to channel demand and offer of a market. Private templates are utilized at the buyers and traders infrastructures and reflect the needs of the particular stakeholder in terms of SLA parameters they use to establish a contract. Typical SLA parameters used at the PaaS level are depicted in Table 1 and include availability, inbound bandwidth, outgoing bandwidth, etc. Considering the high variability of virtual goods in cloud markets, the probability is high that public templates used in marketplaces to attract buyers and sellers and private templates of cloud stakeholders do not match.

One could think that traditional approaches like semantic technologies, e.g., ontologies, can be used to channel variety of SLA templates. Also public templates, which can be downloaded and utilized within the local business / scientific applications could counter act the problem of the variety of SLA templates. However, usage of ontologies is a highly static approach where the dynamics of the changing demand / supply of the market and evolving products cannot be captured. Moreover, utilization of public SLA templates in private business processes or scientific applications is in many cases not possible since it requires changes of the local applications.

In the context of the Austrian national FoSII project (DSG group, Vienna University of Technology), we are investigating self-governing cloud Computing infrastructures necessary for the attainment of established Service Level Agreements (SLAs). Timely prevention of SLA violations requires advanced resource monitoring and knowledge management. In particular, we are developing novel techniques for mapping low-level resource metrics to high-level SLAs and bridging the gap between metrics monitored by the arbitrary monitoring tools and SLA metrics guaranteed to the user, which are usually application based.

We apply various knowledge management techniques, as for example Case Based Reasoning for the prevention of SLA violations before they occur while reducing energy consumption. In collaboration with the Seoul National University we are exploring novel models for SLA mapping to counteract the problem of heterogeneous public and private templates in cloud markets. SLA mapping approach facilitates market participants to define translations from their private templates to public SLA templates while keeping their private temples unchanged. The effects of the SLA mapping approach are twofold:

  • It increases market liquidity since slightly different private templates are channeled towards few publicly available public templates. Consequently, public templates can be frequently adapted based on the supplied, aggregated, and analyzed SLA mappings. Thus, publicly available SLA temples reflect the demand and supply of the markets and can be easily adapted.
  • By clustering supplied SLA mappings different groups of cloud buyers with similar demand can be identified. Thus, based on the information obtained from the clustering information, products for a specific group of users can be tailored. This includes also generation of product niches, which are usually neglected in traditional markets.

SLA mapping is used to bridge the gap between inconsistent parts of two SLA templates – usually between the publicly available template and the private template. For the implementation of the SLA mappings we use XSLT, a declarative XML-based language for the transformation of XML documents. Thereby the original document is not changed, rather the new document is created based on the content of the original document. Thus, if the original document is the private template of the cloud user, which differs from the public template, transformations based on the XSLT can be defined transforming the private into the public template.

Thereby we distinguish two different types of mappings:

1. Ad-hoc SLA mapping. Such mappings define translations between a parameter existing in both, private and public SLA template. We differ simple ad-hoc mapping i.e., mapping of different values for an SLA attribute or an SLA element, e.g., mapping between the names CPU Cores and Number Of Cores of an SLA parameter, and complex ad-hoc mapping, i.e., mapping between different functions for calculating a value of an SLA parameter. An example for the complex mapping would be a unit for expressing a value of an SLA parameter Price from EUR to USD, where translation have to be defined from one function for calculating price to another one. Although, simple and complex mappings appear to be rather trivial, contracts cannot be established between non-matching templates without human intervention of without the overhead of the semantic layer – which anyway has to be managed manually.

2. Future SLA mapping defines a wish for adding a new SLA parameter supported by the application to a public SLA template, or a wish for deleting an existing SLA parameter from a public template. Unlike ad-hoc mapping, future mapping cannot be applied immediately, but possibly in the future. For example a buyer could express the need for a specific SLA parameter, which does not exist yet, but can be integrated into the public templates after the observation of the supplied SLA mappings.

So far we have implemented the first prototype of the VieSLAF (Vienna Service Level Agreement) middleware for the management of SLA mappings allowing users and traders to define, manage, and apply their mappings. In our recent work we developed simulation models for the definition of market settings suitable for the evaluation of the SLA mapping approach in a real world scenario. Based on the applied SLA mappings we defined utility and cost models for users and providers. Thereafter, we applied three different methods for the evaluation of the supplied SLA mappings during a specific time span. We simulated market conditions with a number of market participants entering and leaving the market with different distributions of SLA parameters, thus, requiring different SLA mapping scenarios.

Our first observations show promising results where we achieve good high net utilities considering utilities and costs of doing SLA mappings vs. doing nothing (i.e., not achieving a match in the market). Moreover, in our simulations we applied clustering algorithms where we isolated clusters of SLA templates, which can be used as a starting point for the definition of various cloud products. Utilities achieved when applying clustering algorithms outperforms the costs for doing SLA mappings and doing nothing.

However, those are only preliminary results and the whole potential of SLA mappings is still not fully exploited. Integration into IDEs like Eclipse, where cloud stakeholders can define SLA mapping using suitable Domain Specific Languages, e.g., visual modeling languages, is an open research issue and could facilitate definition of SLA mapping by domains specialists.

The process of defining SLA mapping fully is still in the early stages; for now, these mappings are defined manually by the end users. However, with the development of the appropriate infrastructures and middleware mapping could be done in an automatic way. For example, if the attribute Price has to be translated to Euro a third party service delivering the current USD/Euro exchange rate could be included in an autonomic way facilitating not only mapping between different attributes, but also the proper generation of the according attribute values.

Aggregated and analyzed SLA maps can deliver important information about the demand and structure of the market, thus, facilitating development of open and dynamic cloud markets. Thereby, market rules and structures can be adapted on demand based on the current developments of the products and market participants.

About the Author

Dr. Ivona Brandic is Assistant Professor at the Distributed Systems Group, Information Systems Institute, Vienna University of Technology (TU Wien).

Prior to that, she was Assistant Professor at the Department of Scientific Computing, Vienna University. She received her PhD degree from Vienna University of Technology in 2007. From 2003 to 2007 she participated in the special research project AURORA (Advanced Models, Applications and Software Systems for High Performance Computing) and the European Union’s GEMSS (Grid-Enabled Medical Simulation Services) project.

She is involved in the European Union’s SCube project and she is leading the Austrian national FoSII (Foundations of Self-governing ICT Infrastructures) project funded by the Vienna Science and Technology Fund (WWTF). She is Management Committee member of the European Commission’s COST Action on Energy Efficient Large Scale Distributed Systems. From June-August 2008 she was visiting researcher at the University of Melbourne. Her interests comprise SLA and QoS management, Service-oriented architectures, autonomic computing, workflow management, and large scale distributed systems (cloud, grid, cluster, etc.).
 

Subscribe to HPCwire's Weekly Update!

Be the most informed person in the room! Stay ahead of the tech trends with industry updates delivered to you every week!

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion XL — were added to the benchmark suite as MLPerf continues Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing power it brings to artificial intelligence.  Nvidia's DGX Read more…

Call for Participation in Workshop on Potential NSF CISE Quantum Initiative

March 26, 2024

Editor’s Note: Next month there will be a workshop to discuss what a quantum initiative led by NSF’s Computer, Information Science and Engineering (CISE) directorate could entail. The details are posted below in a Ca Read more…

Waseda U. Researchers Reports New Quantum Algorithm for Speeding Optimization

March 25, 2024

Optimization problems cover a wide range of applications and are often cited as good candidates for quantum computing. However, the execution time for constrained combinatorial optimization applications on quantum device Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at the network layer threatens to make bigger and brawnier pro Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HBM3E memory as well as the the ability to train 1 trillion pa Read more…

MLPerf Inference 4.0 Results Showcase GenAI; Nvidia Still Dominates

March 28, 2024

There were no startling surprises in the latest MLPerf Inference benchmark (4.0) results released yesterday. Two new workloads — Llama 2 and Stable Diffusion Read more…

Q&A with Nvidia’s Chief of DGX Systems on the DGX-GB200 Rack-scale System

March 27, 2024

Pictures of Nvidia's new flagship mega-server, the DGX GB200, on the GTC show floor got favorable reactions on social media for the sheer amount of computing po Read more…

NVLink: Faster Interconnects and Switches to Help Relieve Data Bottlenecks

March 25, 2024

Nvidia’s new Blackwell architecture may have stolen the show this week at the GPU Technology Conference in San Jose, California. But an emerging bottleneck at Read more…

Who is David Blackwell?

March 22, 2024

During GTC24, co-founder and president of NVIDIA Jensen Huang unveiled the Blackwell GPU. This GPU itself is heavily optimized for AI work, boasting 192GB of HB Read more…

Nvidia Looks to Accelerate GenAI Adoption with NIM

March 19, 2024

Today at the GPU Technology Conference, Nvidia launched a new offering aimed at helping customers quickly deploy their generative AI applications in a secure, s Read more…

The Generative AI Future Is Now, Nvidia’s Huang Says

March 19, 2024

We are in the early days of a transformative shift in how business gets done thanks to the advent of generative AI, according to Nvidia CEO and cofounder Jensen Read more…

Nvidia’s New Blackwell GPU Can Train AI Models with Trillions of Parameters

March 18, 2024

Nvidia's latest and fastest GPU, codenamed Blackwell, is here and will underpin the company's AI plans this year. The chip offers performance improvements from Read more…

Nvidia Showcases Quantum Cloud, Expanding Quantum Portfolio at GTC24

March 18, 2024

Nvidia’s barrage of quantum news at GTC24 this week includes new products, signature collaborations, and a new Nvidia Quantum Cloud for quantum developers. Wh Read more…

Alibaba Shuts Down its Quantum Computing Effort

November 30, 2023

In case you missed it, China’s e-commerce giant Alibaba has shut down its quantum computing research effort. It’s not entirely clear what drove the change. Read more…

Nvidia H100: Are 550,000 GPUs Enough for This Year?

August 17, 2023

The GPU Squeeze continues to place a premium on Nvidia H100 GPUs. In a recent Financial Times article, Nvidia reports that it expects to ship 550,000 of its lat Read more…

Shutterstock 1285747942

AMD’s Horsepower-packed MI300X GPU Beats Nvidia’s Upcoming H200

December 7, 2023

AMD and Nvidia are locked in an AI performance battle – much like the gaming GPU performance clash the companies have waged for decades. AMD has claimed it Read more…

DoD Takes a Long View of Quantum Computing

December 19, 2023

Given the large sums tied to expensive weapon systems – think $100-million-plus per F-35 fighter – it’s easy to forget the U.S. Department of Defense is a Read more…

Synopsys Eats Ansys: Does HPC Get Indigestion?

February 8, 2024

Recently, it was announced that Synopsys is buying HPC tool developer Ansys. Started in Pittsburgh, Pa., in 1970 as Swanson Analysis Systems, Inc. (SASI) by John Swanson (and eventually renamed), Ansys serves the CAE (Computer Aided Engineering)/multiphysics engineering simulation market. Read more…

Choosing the Right GPU for LLM Inference and Training

December 11, 2023

Accelerating the training and inference processes of deep learning models is crucial for unleashing their true potential and NVIDIA GPUs have emerged as a game- Read more…

Intel’s Server and PC Chip Development Will Blur After 2025

January 15, 2024

Intel's dealing with much more than chip rivals breathing down its neck; it is simultaneously integrating a bevy of new technologies such as chiplets, artificia Read more…

Baidu Exits Quantum, Closely Following Alibaba’s Earlier Move

January 5, 2024

Reuters reported this week that Baidu, China’s giant e-commerce and services provider, is exiting the quantum computing development arena. Reuters reported � Read more…

Leading Solution Providers

Contributors

Comparing NVIDIA A100 and NVIDIA L40S: Which GPU is Ideal for AI and Graphics-Intensive Workloads?

October 30, 2023

With long lead times for the NVIDIA H100 and A100 GPUs, many organizations are looking at the new NVIDIA L40S GPU, which it’s a new GPU optimized for AI and g Read more…

Shutterstock 1179408610

Google Addresses the Mysteries of Its Hypercomputer 

December 28, 2023

When Google launched its Hypercomputer earlier this month (December 2023), the first reaction was, "Say what?" It turns out that the Hypercomputer is Google's t Read more…

AMD MI3000A

How AMD May Get Across the CUDA Moat

October 5, 2023

When discussing GenAI, the term "GPU" almost always enters the conversation and the topic often moves toward performance and access. Interestingly, the word "GPU" is assumed to mean "Nvidia" products. (As an aside, the popular Nvidia hardware used in GenAI are not technically... Read more…

Shutterstock 1606064203

Meta’s Zuckerberg Puts Its AI Future in the Hands of 600,000 GPUs

January 25, 2024

In under two minutes, Meta's CEO, Mark Zuckerberg, laid out the company's AI plans, which included a plan to build an artificial intelligence system with the eq Read more…

Google Introduces ‘Hypercomputer’ to Its AI Infrastructure

December 11, 2023

Google ran out of monikers to describe its new AI system released on December 7. Supercomputer perhaps wasn't an apt description, so it settled on Hypercomputer Read more…

China Is All In on a RISC-V Future

January 8, 2024

The state of RISC-V in China was discussed in a recent report released by the Jamestown Foundation, a Washington, D.C.-based think tank. The report, entitled "E Read more…

Intel Won’t Have a Xeon Max Chip with New Emerald Rapids CPU

December 14, 2023

As expected, Intel officially announced its 5th generation Xeon server chips codenamed Emerald Rapids at an event in New York City, where the focus was really o Read more…

IBM Quantum Summit: Two New QPUs, Upgraded Qiskit, 10-year Roadmap and More

December 4, 2023

IBM kicks off its annual Quantum Summit today and will announce a broad range of advances including its much-anticipated 1121-qubit Condor QPU, a smaller 133-qu Read more…

  • arrow
  • Click Here for More Headlines
  • arrow
HPCwire