Umeå University's logo

umu.sePublications
Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Impact of etcd deployment on Kubernetes, Istio, and application performance
Umeå University, Faculty of Science and Technology, Department of Computing Science.ORCID iD: 0000-0001-5860-6695
Umeå University, Faculty of Science and Technology, Department of Computing Science.ORCID iD: 0000-0003-0106-3049
Umeå University, Faculty of Science and Technology, Department of Computing Science.ORCID iD: 0000-0002-2633-6798
Show others and affiliations
2020 (English)In: Software, practice & experience, ISSN 0038-0644, E-ISSN 1097-024X, Vol. 50, no 10, p. 1986-2007Article in journal (Refereed) Published
Abstract [en]

This experience article describes lessons learned as we conducted experiments in a Kubernetes-based environment, the most notable of which was that the performance of both the Kubernetes control plane and the deployed application depends strongly and in unexpected ways on the performance of the etcd database. The article contains (a) detailed descriptions of how networking with and without Istio works in Kubernetes, based on the Flannel Container Networking Interface (CNI) provider in VXLAN mode with IP Virtual Server (IPVS)-backed Kubernetes Services, (b) a comprehensive discussion about how to conduct load and performance testing using a closed-loop workload generator, and (c) an open source experiment framework useful for executing experiments in a shared cloud environment and exploring the resulting data. It also shows that statistical analysis may reveal the data resulting from such experiments to be misleading even when careful preparations are made, and that nondeterministic behavior stemming from etcd can affect both the platform as a whole and the deployed application. Finally, it is demonstrated that using high-performance backing storage for etcd can reduce the occurrence of such nondeterministic behaviors by a statistically significant (P < .05) margin. The implication of this experience article is that systems researchers studying the performance of applications deployed on Kubernetes cannot simply consider their specific application to be under test. Instead, the particularities of the underlying Kubernetes and cloud platform must be taken into account, in particular because their performance can impact that of etcd.

Place, publisher, year, edition, pages
John Wiley & Sons, 2020. Vol. 50, no 10, p. 1986-2007
Keywords [en]
cloud computing, distributed systems, etcd, Kubernetes, performance
National Category
Computer Sciences
Identifiers
URN: urn:nbn:se:umu:diva-174336DOI: 10.1002/spe.2885ISI: 000556599500001Scopus ID: 2-s2.0-85089109897OAI: oai:DiVA.org:umu-174336DiVA, id: diva2:1459766
Available from: 2020-08-20 Created: 2020-08-20 Last updated: 2023-03-23Bibliographically approved
In thesis
1. Managing cloud resource scarcity
Open this publication in new window or tab >>Managing cloud resource scarcity
2020 (English)Doctoral thesis, comprehensive summary (Other academic)
Abstract [en]

According to the Infrastructure-as-a-Service conceptualization of cloud computing, Infrastructure Providers offer utility-like pay-as-you-go access to computing resources (e.g., data processing, networks, and storage) to Service Providers, who use those resources to host applications for the benefit of end users. The quantity of resources available to Infrastructure Providers at any given moment is limited, as is the quantity of resources allocated to the applications of each Service Provider.

This thesis examines the management of cloud resource scarcity from the perspectives of both Infrastructure and Service Providers, with the aim of finding ways to ensure that the end user experience is minimally affected.

We consider three main strategies for managing cloud resource scarcity. First, we explore ways to efficiently construct collaborative federations of autonomous and independent Infrastructure Providers that allow local resource scarcity to be masked by extension using capacity from remote sites. Second, we consider how scheduling both within a cloud site and across a federation can be made aware of restrictions imposed by Service Providers for, e.g., performance or legal reasons. Third, we suggest ways of making applications conscious of resource availability so that they can apply quality elasticity under resource constraints.

The thesis is the culmination of 11 years of work within academia and industry. Based on the unique perspective granted by this long experience, the introductory chapters present a historical view of each subtopic mentioned above. Specifically, they discuss how cloud computing has evolved in conjunction with ways of developing applications to the symbiotic benefit of both, leading to the emergence of cloud-native software that allows Infrastructure Providers to use their infrastructure more efficiently and offer it more affordably while simultaneously granting Service Providers improved availability and performance in cloud-based environments.

Abstract [sv]

I molnet (cloud computing), ur Infrastructure-as-a-Service-perspektivet, ger infrastruktursleverantörer tillgång till datorresurser (såsom databehandlings-, nätverks- och lagringskapacitet) med en löpande betalningsmodell till tjänsteleverantörer så att dessa kan erbjuda applikationer till fördel för slutanvändare. För infrastruktursleverantörer finns det en begränsning i mängden datorresurser som är tillgängliga vid varje givet tillfälle. För tjänsteleverantörer finns det en begränsning i hur många sådana resurser som är allokerade till deras applikationer.

Det övergripande målet med denna doktorsavhandling är att studera olika sätt att hantera resursbrister i molnet ur både infrastruktursleverantörers och tjänsteleverantörers perspektiv, så att slutanvändares upplevelse blir minimalt påverkad.

Våra ansatser till att hantera resursbrister i molnet kommer från tre huvudsakliga områden. Först utforskar vi federationer av autonoma och oberoende infrastruktursleverantörer, som möjliggör att lokala resursbrister kan maskeras genom att utöka och nyttja resurser från andra leverantörer. Därnäst undersöker vi hur schemaläggning både inom och mellan molnsajter i en federation kan göras medveten om begränsningar som tjänsteleverantörer kräver av exempelvis prestanda- eller legala skäl. Slutligen föreslår vi olika sätt hur applikationer kan göras medvetna om nuvarande resurstillgång och göras kvalitetselastiska.

Avhandlingen utgör kulmen av 11 års arbete inom akademins och industrins värld. Baserat på de unika möjligheter en sådan erfarenhet ger presenteras även ett historiskt perspektiv av dessa områden i de inledande kapitlen. I dessa kapitel diskuterar vi hur molnet har utvecklats tillsammans med hur applikationer levereras till slutkunder och hur ett symbiotiskt förhållande uppstått dem emellan. Resultatet är mjukvara som är cloud-native, vilket möjliggör för infrastruktursleverantörer att till högre grad effektivt utnyttja sin infrastruktur och erbjuda tillgång till den på ett mer kostnadseffektivt sätt, samt ger tjänsteleverantörer ökad tillgänglighet och prestanda i molnbaserade miljöer.

Place, publisher, year, edition, pages
Umeå: Umeå universitet, 2020. p. 109
Series
Report / UMINF, ISSN 0348-0542 ; 20.07
Keywords
cloud computing, edge computing, kubernetes, federation, scheduling, quality elasticity
National Category
Computer Sciences
Identifiers
urn:nbn:se:umu:diva-174509 (URN)978-91-7855-328-0 (ISBN)978-91-7855-329-7 (ISBN)
Public defence
2020-10-16, Aula Biologica, Biologihuset, Umeå, 10:00 (English)
Opponent
Supervisors
Available from: 2020-09-25 Created: 2020-08-26 Last updated: 2020-10-20Bibliographically approved

Open Access in DiVA

fulltext(1318 kB)382 downloads
File information
File name FULLTEXT02.pdfFile size 1318 kBChecksum SHA-512
4721360508fec3f0caa24dca901abd1691ce2846167a8f6a91738a9fb3b4c8575059ee9728a1a73ded64ae759a6d863afc7f6bbc4a72c5dc59e21384e4eb642a
Type fulltextMimetype application/pdf

Other links

Publisher's full textScopus

Authority records

Larsson, LarsKlein, CristianElmroth, Erik

Search in DiVA

By author/editor
Larsson, LarsKlein, CristianElmroth, Erik
By organisation
Department of Computing Science
In the same journal
Software, practice & experience
Computer Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 500 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

doi
urn-nbn

Altmetric score

doi
urn-nbn
Total: 725 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf