Jan 25th, 2019Demonstrate trace generators and discuss the issues and improvement opportunities of MITHRIL.

Jan 18th, 2019Study the MITHRIL paper and plan to build a trace generator to evaluate the algorithms and others.

Jun 11th, 2018Safal: Keeping a pointer to last NR block on stack S isn’t very efficient...currently trying to do the second idea of making Q behave like ARC.

Jun 6th, 2018 Discussed the metrics used, it was suggested that perhaps the Q2 hit ratio alone is not sufficient to decide when to change Q size, it was suggested to compare the Q2 hit ratio with the LIRS hit ratio, perhaps factoring in the cache allocation percentage to S and Q.

May 29th, 2018 Discussion about the ARC, LRU, LIRS algorithms.

May 25th, 2018 In the LIRS algorithm, there is a parameter, which is size of the stack Q. By default it is set as 1% of the cache size. However, for the highest hit ratio the parameter may need to be set differently with different workloads or even dynamically for one workload. A major effort during the summer is to investigate impact of the parameter and strategies to adjust it, or equivalently to come up with an improved LIRS algorithm.