1 / 6

Cataclysm: Handling Extreme Overloads in Internet Services

Cataclysm: Handling Extreme Overloads in Internet Services. Bhuvan Urgaonkar and Prashant Shenoy University of Massachusetts Amherst. Overloads in Internet Applications. Bottleneck!. Internet apps experience overloads E.g. 20 fold increase in CNN traffic on 9/11

turner
Télécharger la présentation

Cataclysm: Handling Extreme Overloads in Internet Services

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Cataclysm: Handling Extreme Overloads in Internet Services Bhuvan Urgaonkar and Prashant Shenoy University of Massachusetts Amherst

  2. Overloads in Internet Applications Bottleneck! • Internet apps experience overloads • E.g. 20 fold increase in CNN traffic on 9/11 • Goal: Handle overloads without app downtime Ecommerce app • Key idea: Scalable policing & provisioning

  3. Cataclysm Components • Provisioner: Queuing theoretic app models • Add servers to overloaded apps • Policer: Differentiated service and scalability • Admit important requests during overload • Switch to approx. admission control at high loads • Even higher loads => multiple sentries, DNS RR d2 > d1 d1 Admitted d2 Admission Control Dropped Classifier

  4. Fraction admitted Arrival rate 1.2 250 1 200 0.8 GLD GLD 150 Fraction admitted Arrival rate 0.6 SIL SIL 100 BRZ BRZ 0.4 50 0.2 0 0 0 100 200 300 400 500 600 0 100 200 300 400 500 600 Time (sec) Time (sec) Cataclysm Policer • Admits important requests during overloads • Admitted requests see good response time • Single policer: arrival rates upto 19,000 req/s • Can add more policers at higher loads

  5. Concluding Remarks • The Cataclysm hosting platform • Scalable policer • Dynamic provisioning of servers • Implementation on Linux cluster of size 20 • Experimentation with variety of workloads

  6. Thank you! • More information: • http://lass.cs.umass.edu/papers.html

More Related