Srikanth Kandula

About

I am a senior principal researcher at Microsoft Research. My interests are broadly in building and analyzing networked systems. Of late, I have worked on big-data platforms and datacenter networks. I completed my PhD in Computer Science from MIT in 2008.

News:

• 03/22: Handling workload and data drift in learned cardinality estimation models using, in part, query GANs; at SIGMOD; kudos to Beibin and Yao.
• 10/21: Pre-training summarization models for structured datasets at VLDB; kudos to Yao and team.
• 10/21: A simple problem-agnostic parallelization method for granular optimization problems at SOSP; kudos to Deepak, Firas, Fiodar and team.
• 10/21: Using nyquist theorem, we show that datacenter measurements can be under-sampled by up to 1000x without loss; kudos to Behnaz and team.
• 08/21: “distinguished reviewer” at VLDB.
• 09/20: Extended version of data-induced predicates.
• 08/20: A new parallel data-generator for zipf-skewed TPC-H.
• 08/20: Approximate partition selection using summary statistics; @VLDB
• 07/20: Faster TE by decomposing multi-commodity flow; @NSDI.
• 10/19: Data-induced predicates; @VLDB (‘best paper runner-up‘).
• 08/19: Lucky to be part of VL2 which received the test-of-time award from SIGCOMM.
• 05/19: Experiences from shipping samplers and QO rules; @VLDB; talk.
• 07/18: Network + job-aware scheduling of cache tiers; @SoCC; `best paper’.
• 06/18: Probabilistic predicates; @SIGMOD; demo (`best demo’); blog post.

Current project: Lazy approximations, Cluster scheduling, SWAN++

Past projects: Seawall, Flyways, CloudCmp, Netmedic, Broom, EXpose, VL2, FatVAP, Sherlock, TeXCP, Flare, Kill-Bots