EuroSys'10
Q-Clouds: Managing Performance Interference Effects for QoS-Aware Clouds
* Motivation: In existing systems, customers are charged based upon the amount of resources used or reserved, but no guarantees are made regarding the application level performance or quality-of-service (QoS) that the given resources will provide.
As cloud providers continue to utilize virtualization technologies in their systems, this can become problematic. In particular, the consolidation of multiple customer applications onto multicore servers introduces performance interference between collocated workloads, significantly impacting
application QoS.
* Abstract:
1.To address this challenge, we advocate that the cloud should transparently provision additional resources as necessary to achieve the performance that customers would have realized if they were running in isolation("Head Room").
2. For example, an application using one core of a multicore processor may experience significantly reduced performance when another application simultaneously runs on an adjacent core, due to an increased miss rate in the last level cache (LLC) [8, 13, 39].(互相干扰的一个例子:LLC增加)。
3. One approach to deal with this indeterminism is to improve virtualization
technologies, through better resource partitioning in both hardware and software, and remove performance interference altogether. A benefit of this approach is that the cloud can continue to be agnostic of application QoS and maintain
simple resource capacity based provisioning and billing. However, perfect resource partitioning can be difficult and costly to implement, and even if accomplished, may result in inefficient use of resources [36].(完全的隔离是牺牲性能的、不合理)。
4. First, Q-Clouds employs application feedback to build multiinput multi-output (MIMO) models that capture interference relationships between applications. An advantage of the approach is that the system does not need to determine the
underlying sources of interference. The MIMO model is used in a closed loop controller to tune resource allocations and achieve specified performance levels for each VM.(MIMO Model, 优点是不需要知道底层interference细节)。
Q-Clouds: Managing Performance Interference Effects for QoS-Aware Clouds
* Motivation: In existing systems, customers are charged based upon the amount of resources used or reserved, but no guarantees are made regarding the application level performance or quality-of-service (QoS) that the given resources will provide.
As cloud providers continue to utilize virtualization technologies in their systems, this can become problematic. In particular, the consolidation of multiple customer applications onto multicore servers introduces performance interference between collocated workloads, significantly impacting
application QoS.
* Abstract:
1.To address this challenge, we advocate that the cloud should transparently provision additional resources as necessary to achieve the performance that customers would have realized if they were running in isolation("Head Room").
2. For example, an application using one core of a multicore processor may experience significantly reduced performance when another application simultaneously runs on an adjacent core, due to an increased miss rate in the last level cache (LLC) [8, 13, 39].(互相干扰的一个例子:LLC增加)。
3. One approach to deal with this indeterminism is to improve virtualization
technologies, through better resource partitioning in both hardware and software, and remove performance interference altogether. A benefit of this approach is that the cloud can continue to be agnostic of application QoS and maintain
simple resource capacity based provisioning and billing. However, perfect resource partitioning can be difficult and costly to implement, and even if accomplished, may result in inefficient use of resources [36].(完全的隔离是牺牲性能的、不合理)。
4. First, Q-Clouds employs application feedback to build multiinput multi-output (MIMO) models that capture interference relationships between applications. An advantage of the approach is that the system does not need to determine the
underlying sources of interference. The MIMO model is used in a closed loop controller to tune resource allocations and achieve specified performance levels for each VM.(MIMO Model, 优点是不需要知道底层interference细节)。