Browsing by Author "Arlitt, Martin"

Now showing 1 - 6 of 6
  • Thumbnail Image
    ItemOpen Access
    Auto-Scaling Containerized Microservice Applications
    (2021-09) MirzaEbrahim Mostofi, Vahid; Krishnamurthy, Diwakar; Arlitt, Martin; Drew, Steve; Medeiros de Souza, Roberto
    The microservices architecture is being increasingly used to build complex applications. Many such applications are customer-facing. Hence, they face workload fluctuations and need to respond to end user requests quickly in spite of such fluctuations. Furthermore, application owners typically prefer to allocate resources efficiently using containerization technology so that operational costs are kept low. These two requirements are typically implemented within an auto-scaler module. The third requirement, unique to microservice applications, is the need for owners to roll out updates in an agile and frequent manner. Hence, an auto-scaler appropriate for microservices should be designed to support this requirement. Unfortunately, current auto-scaling techniques do not satisfy these three key requirements simultaneously. I develop a novel auto-scaler called TRIM that addresses this open issue. TRIM exploits properties of real-life microservice workloads. Specifically, my analysis of a large dataset consisting of $24,000$ production microservice applications reveals a novel insight that a small number of workload patterns are encountered frequently over any given time period. Furthermore, these popular patterns tend to be popular during a subsequent time period. TRIM pre-computes resource allocations for these small number of popular patterns quickly and re-uses these allocations at runtime when appropriate. I develop MOAT, a novel heuristic optimization technique that ensures that the pre-computed allocations satisfy response time targets efficiently. Using a variety of analytical, on-premise, and public cloud systems, I show that MOAT and TRIM outperform state-of-the-art baselines in terms of all three requirements described previously. For example, I consider an on-premise system subjected to a $24$ hour workload derived from a real-life microservice application. By quickly pre-computing resource allocations for just 5 popular workload patterns in this workload, TRIM achieves up to 70% lesser response time violations and up to 20% reduced costs compared to an industry-standard auto-scaling technique.
  • Thumbnail Image
    ItemOpen Access
    Characterization of Periodic Network Traffic
    (2017) Haffey, Mackenzie; Williamson, Carey; Arlitt, Martin; Williamson, Carey; Arlitt, Martin; Aycock, John; Fong, Philip
    This thesis focuses on characterizing periodic communications in network traffic, which we refer to as network heartbeats. Heartbeat traffic can be used to assess the overall health of an operational network, based on the presence/absence of heartbeats for known network services, and also to detect unexpected/undesired network services, such as malicious traffic. We use a simple and flexible SQL-based method to detect a wide range of heartbeats in network traffic, using seven weeks of connection logs from a campus edge network. Our results show that heartbeat analysis is effective for detecting P2P, gaming, cloud, scanning, and botnet traffic flows, which often have periodic signatures.
  • Thumbnail Image
    ItemOpen Access
    Facebook Meets the Virtualized Enterprise
    (2008-07-15T22:16:05Z) Simmonds, Robert; Curry, Roger; Kiddle, Cameron; Markatchev, Nayden; Tan, Tingxi; Arlitt, Martin; Walker, Bruce
    “Web 2.0” and “cloud computing” are revolutionizing the way IT infrastructure is accessed and managed. Web 2.0 technologies such as blogs, wikis and social networking platforms provide Internet users with easier mechanisms to produce Web content and to interact with each other. Cloud computing technologies are aimed at running applications as services over the Internet on a scalable infrastructure. They enable businesses that do not have the capital or technical expertise to support their own infrastructure to get access to computing on demand. They could also be used by large businesses to more efficiently manage their own infrastructure as an “internal cloud”. In this paper we explore the advantages of using Web 2.0 and cloud computing technologies in an enterprise setting to provide employees with a comprehensive and transparent environment for utilizing applications. To demonstrate the effectiveness of this approach we have developed an environment that uses Facebook (a social networking platform) to provide access to the Fire Dynamics Simulator (a legacy application). The application is supported using Virtual Appliances that are hosted in an internal cloud computing infrastructure that adapts dynamically to user demands. Initial feedback suggests this approach provides a much better user experience than the traditional standalone use of the application. It also simplifies the management and increases the effective utilization of the underlying IT resources.
  • Thumbnail Image
    ItemOpen Access
    NetFlix and Twitch Traffic Characterization
    (2015-09-30) Laterman, Michel; Williamson, Carey; Arlitt, Martin
    Streaming video content is the largest contributor to inbound network traffic at the University of Calgary. Over five months, from December 2014 - April 2015, over 2.7 petabytes of traffic on 49 billion connections was observed. This thesis presents traffic characterizations for two large video streaming services, namely NetFlix and Twitch. These two services contribute a significant portion of inbound bytes. NetFlix provides TV series and movies on demand. Twitch offers live streaming of video game play. These services share many characteristics, including asymmetric connections, content delivery mechanisms, and content popularity patterns. This thesis sheds light on the usage of modern video streaming services on an edge network. It's one of only a few studies to utilize long-term network-level data. To the best of our knowledge, it's one of the first studies that uses network-level data for Twitch traffic characterization, and content characterization for NetFlix and Twitch.
  • Thumbnail Image
    ItemOpen Access
    Pandemic Effects on Campus Network Traffic
    (2023-03-27) Karamollahi, Mehdi; Williamson, Carey; Arlitt, Martin; Krishnamurthy, Diwakar; Willett, Wesley; Haque, Israat; Claypool, Mark; Reardon, Joel
    The first wave of the COVID-19 pandemic hit North America in March 2020, disrupting personal and professional lives, and leading to work-from-home mandates in many jurisdictions. The lockdown measures started at the University of Calgary on March 13, 2020, and the university switched to fully remote learning and working. Although the lockdown measures evolved over the following months, the pandemic significantly affected how people used the campus network in both the short term and the longer term. In this dissertation, we use three years (i.e., 2019, 2020, and 2021) of empirical network traffic measurement data from the University of Calgary’s campus network to study the effects of the pandemic on our post-secondary education environment. The highlights from overall changes on our campus include: changes to inbound and outbound traffic volumes; reduced traffic asymmetry; significant growth in videoconferencing traffic; structural changes in workday traffic patterns; and more global distribution of campus network users. The research in this dissertation takes an applied approach, focusing on the performance implications of these pandemic-related traffic changes as viewed through the lens of a campus edge network. We first study videoconferencing applications and remote access services used during the lockdown. We study their traffic volumes, directionality, and diurnal patterns and characterize them as observed from our campus edge network. Next, we investigate the increase in inbound scanning activities on our campus network during the lockdown. We characterize this traffic, identify the legitimate and suspicious actors involved, and discuss three specific examples of security-related incidents observed on our campus network. The methodology used in these case studies could be used for similar studies. We also study our campus community and analyze the pandemic effects on different sub-communities. Our findings show that the network usage patterns are highly correlated with the physical presence requirements on campus. Moreover, many asymmetries in connection counts and traffic byte volumes are evident, as the pandemic led to many shifts in application usage. Lastly, we select Zoom as the most prevalent videoconferencing application adopted by our campus community and further analyze its network traffic. We investigate connection-level and packet-level Zoom traffic, identify its structure, and identify the root cause of performance problems with Zoom sessions on our campus network. To mitigate these problems, we propose multiple solutions and evaluate them quantitatively using workload modelling and simulation.
  • Thumbnail Image
    ItemOpen Access
    Traffic Analysis of Two Scientific Web Sites
    (2015-12-15) Liu, Yang; Williamson, Carey; Williamson, Carey; Arlitt, Martin; Donovan, Eric
    This thesis presents a workload characterization study of two scientific Web sites at the University of Calgary based on a four-month period of observation (from January 1, 2015 to April 30, 2015). The Aurora site is a scientific site for auroral researchers, providing auroral images collected from remote cameras deployed in northern Canada. The ISM site is a scientific site providing lecture materials to about 400 undergraduate students in the ASTR 209 course. Three main observations emerge from our workload characterization study. First, scientific Web sites can generate extremely large volumes of Internet traffic, even when the user community is seemingly small. Second, robot traffic and real-world events can have surprisingly large impacts on network traffic. Third, a large fraction of the observed network traffic is highly redundant, and can be reduced significantly with more efficient networking solutions.