Academia.eduAcademia.edu

Figure 1 – uploaded by Ferhat Khendek

See full PDFdownloadDownload figure
The required measurements for our evaluation have been obtained from a real deployment with six compute nodes. The upgrade scenario for the six nodes was executed ten times, and

Table 1 The required measurements for our evaluation have been obtained from a real deployment with six compute nodes. The upgrade scenario for the six nodes was executed ten times, and

Related Figures (13)

Fig. 2. Partial resource graph for the illustrative example
Fig. 2. Partial resource graph for the illustrative example
Fig. 3. A framework for IaaS cloud system upgrade
Fig. 3. A framework for IaaS cloud system upgrade
Fig. 4. The iterative process of the IaaS upgrade approach
Fig. 4. The iterative process of the IaaS upgrade approach
Fig. 5. Upgrade control graph for the illustrative example
Fig. 5. Upgrade control graph for the illustrative example
Fig. 6. Two different cases for our evaluation scenario TABLE I. SCALING PARAMETERS FOR THE TENANTS IN OUR EVALUATION SCENARIO
Fig. 6. Two different cases for our evaluation scenario TABLE I. SCALING PARAMETERS FOR THE TENANTS IN OUR EVALUATION SCENARIO
Fig. 7. Comparison of the total duration of the upgrade for evaluation scenario (a)
Fig. 7. Comparison of the total duration of the upgrade for evaluation scenario (a)
Fig. 8. Comparison of the total duration of the upgrade for evaluation scenario (b)
Fig. 8. Comparison of the total duration of the upgrade for evaluation scenario (b)
Fig. 10. Comparison of the average outage at the application level for evaluation scenario (b) Fig. 9. Comparison of the average outage at the application level for evaluation scenario (a)
Fig. 10. Comparison of the average outage at the application level for evaluation scenario (b) Fig. 9. Comparison of the average outage at the application level for evaluation scenario (a)
Fig. 11. Comparison of the outage for each VM during the upgrade of VMs. Table II and Table III demonstrate the comparison o SLA violations during upgrade scenarios (a) and (b) respectively. In our evaluation we calculated the number o times SLA violations occur for each tenant and we considered the average as the number of SLA violations per tenant. The number of impacted VMs per tenant in each SLA violation indicates how many VMs are impacted. For the average tota duration of SLA violations per tenant, we calculated the tota
Fig. 11. Comparison of the outage for each VM during the upgrade of VMs. Table II and Table III demonstrate the comparison o SLA violations during upgrade scenarios (a) and (b) respectively. In our evaluation we calculated the number o times SLA violations occur for each tenant and we considered the average as the number of SLA violations per tenant. The number of impacted VMs per tenant in each SLA violation indicates how many VMs are impacted. For the average tota duration of SLA violations per tenant, we calculated the tota
TABLE II. SLA VIOLATION RELATED MEASUREMENT RESULTS FOR ALL POSSIBLE BATCH SELECTIONS FOR EVALUATION SCENARIO (A) Internal validity threats: The first threat to validity is that we conducted our evaluation by calculating the measurements for a cluster of ten compute nodes considering the measurements obtained from a real deployment of six compute nodes. Experimental results from a real deployment of ten compute nodes might differ from the results calculated in our evaluation.
TABLE II. SLA VIOLATION RELATED MEASUREMENT RESULTS FOR ALL POSSIBLE BATCH SELECTIONS FOR EVALUATION SCENARIO (A) Internal validity threats: The first threat to validity is that we conducted our evaluation by calculating the measurements for a cluster of ten compute nodes considering the measurements obtained from a real deployment of six compute nodes. Experimental results from a real deployment of ten compute nodes might differ from the results calculated in our evaluation.
TABLE A. DEFINITION FOR THE PARAMETERS USED IN THE PROPOSED APPROACH
TABLE A. DEFINITION FOR THE PARAMETERS USED IN THE PROPOSED APPROACH
Related topics:
Computer ScienceService Level Agreement
580 California St., Suite 400
San Francisco, CA, 94104
© 2025 Academia. All rights reserved