[DB Seminar] Fall 2017: Joy Arulraj
For the first time in 25 years, a new non-volatile memory (NVM) category is being created that is expected to be 1000 times faster than current durable storage devices. The advent of NVM will fundamentally change the dichotomy between memory and durable storage in database systems (DBMSs). These new NVM devices are almost as fast as DRAM, but all writes to it are potentially persistent even after power loss. Existing DBMSs are unable to take full advantage of this technology because their internal architectures are predicated on the assumption that memory is volatile. With NVM, many of the components of legacy DBMSs are unnecessary and will degrade the performance of data intensive applications.
In this talk, I will present our research on understanding the implications of NVM for database systems. I will begin by discussing why adapting existing database systems to use NVM like a faster SSD will not be sufficient to leverage the unique characteristics of NVM. I will then describe the design of a new NVM-aware DBMS code named Peloton. Peloton uses a new logging and recovery protocol, called write-behind logging, that reduces the recovery time of the system by more than two orders of magnitude compared to the traditional write-ahead logging protocol. Besides improving application availability, we found that write-behind logging can help improve space utilization of the NVM device and extend its lifetime.
I will conclude with promising initial results from our ongoing investigation on supporting real-time analytics on NVM. We are now extending Peloton to run on a multi-tier storage hierarchy comprising of DRAM, NVM, and SSD. We are excited about the development of Peloton as a research vehicle for investigating the changes required for supporting emergent storage technologies.