Memory resource balancing for virtualized computing

Virtualization has become a common abstraction layer in modern data centers. By multiplexing hardware resources into multiple virtual machines (VMs) and thus enabling several operating systems to run on the same physical platform simultaneously, it can effectively reduce power consumption and building size or improve security by isolating VMs. In a virtualized system, memory resource management plays a critical role in achieving high resource utilization and performance. Insufficient memory allocation to a VM will degrade its performance dramatically. On the contrary, over-allocation causes waste of memory resources. Meanwhile, a VM’s memory demand may vary significantly. As a result, effective memory resource management calls for a dynamic memory balancer, which, ideally, can adjust memory allocation in a timely manner for each VM based on their current memory demand and thus achieve the best memory utilization and the optimal overall performance. In order to estimate the memory demand of each VM and to arbitrate possible memory resource contention, a widely proposed approach is to construct an LRU-based miss ratio curve (MRC), which provides not only the current working set size (WSS) but also the correlation between performance and the target memory allocation size. Unfortunately, the cost of constructing an MRC is nontrivial. In this dissertation, we first present a low overhead LRU-based memory demand tracking scheme, which includes three orthogonal optimizations: AVL-based LRU organization, dynamic hot set sizing and intermittent memory tracking. Our evaluation results show that, for the whole SPEC CPU 2006 benchmark suite, after applying the three optimizing techniques, the mean overhead of MRC construction is lowered from 173% to only 2%. Based on current WSS, we then predict its trend in the near future and take different strategies for different prediction results. When there is a sufficient amount of physical memory on the host, it locally balances its memory resource for the VMs. Once the local memory resource is insufficient and the memory pressure is predicted to sustain for a sufficiently long time, a relatively expensive solution, VM live migration, is used to move one or more VMs from the hot host to other host(s). Finally, for transient memory pressure, a remote cache is used to alleviate the temporary performance penalty. Our experimental results show that this design achieves 49% center-wide speedup.

Nitrogen Oxides in the North Atlantic Troposphere: Impacts of Boreal Wildfire and Anthropogenic Emissions

Nitrogen oxides play a crucial role in the budget of tropospheric ozone (O sub(3)) and the formation of the hydroxyl radical. Anthropogenic activities and boreal wildfires are large sources of emissions in the atmosphere. However, the influence of the transport of these emissions on nitrogen oxides and O sub(3) levels at hemispheric scales is not well understood, in particular due to a lack of nitrogen oxides measurements in remote regions. In order to address these deficiencies, measurements of NO, NO sub(2) and NO sub(y) (total reactive nitrogen oxides) were made in the lower free troposphere (FT) over the central North Atlantic region (Pico Mountain station, 38 degree N 28 degree W, 2.3 km asl) from July 2002 to August 2005. These measurements reveal a well-defined seasonal cycle of nitrogen oxides (NO sub(x) = NO+NO sub(2) and NO sub(y)) in the background central North Atlantic lower FT, with higher mixing ratios during the summertime. Observed NO sub(x) and NO sub(y) levels are consistent with long-range transport of emissions, but with significant removal en-route to the measurement site. Reactive nitrogen largely exists in the form of PAN and HNO sub(3) ( similar to 80-90% of NO sub(y)) all year round. A shift in the composition of NO sub(y) from dominance of PAN to dominance of HNO sub(3) occurs from winter-spring to summer-fall, as a result of changes in temperature and photochemistry over the region. Analysis of the long-range transport of boreal wildfire emissions on nitrogen oxides provides evidence of the very large-scale impacts of boreal wildfires on the tropospheric NO sub(x) and O sub(3) budgets. Boreal wildfire emissions are responsible for significant shifts in the nitrogen oxides distributions toward higher levels during the summer, with medians of NO sub(y) (117-175 pptv) and NO sub(x) (9-30 pptv) greater in the presence of boreal wildfire emissions. Extreme levels of NO sub(x) (up to 150 pptv) and NO sub(y) (up to 1100 pptv) observed in boreal wildfire plumes suggest that decomposition of PAN to NO sub(x) is a significant source of NO sub(x), and imply that O sub(3) formation occurs during transport. Ozone levels are also significantly enhanced in boreal wildfire plumes. However, a complex behavior of O sub(3) is observed in the plumes, which varies from significant to lower O sub(3) production to O sub(3) destruction. Long-range transport of anthropogenic emissions from North America also has a significant influence on the regional NO sub(x) and O sub(3) budgets. Transport of pollution from North America causes significant enhancements on nitrogen oxides year-round. Enhancements of CO, NO sub(y) and NO sub(x) indicate that, consistent with previous studies, more than 95% of the NO sub(x) emitted over the U.S. is removed before and during export out of the U.S. boundary layer. However, about 30% of the NO sub(x) emissions exported out of the U.S. boundary layer remain in the airmasses. Since the lifetime of NO sub(x) is shorter than the transport timescale, PAN decomposition and potentially photolysis of HNO sub(3) provide a supply of NO sub(x) over the central North Atlantic lower FT. Observed Delta O sub(3)/ Delta NO sub(y) and large NO sub(y) levels remaining in the North American plumes suggest potential O sub(3) formation well downwind from North America. Finally, a comparison of the nitrogen oxides measurements with results from the global chemical transport (GCT) model GEOS-Chem identifies differences between the observations and the model. GEOS-Chem reproduces the seasonal variation of nitrogen oxides over the central North Atlantic lower FT, but does not capture the magnitude of the cycles. Improvements in our understanding of nitrogen oxides chemistry in the remote FT and emission sources are necessary for the current GCT models to adequately estimate the impacts of emissions on tropospheric NO sub(x) and the resulting impacts on the O sub(3) budget.