Antonios Katsarakis

consistency replication performance distributed fault-tolerance protocols strong eurosys rdma linearizability scale concurrency transactions asplos dlht dandelion resizable high performance memory-aware non-blocking hashtable asynchronous ndsi sosp osdi locality fast low-latency invalidations hermes skew fully numa out scale-out transformations actions map large spark rdd reduce dltree u2pc batching hpdc high-performance requests billion lock-free parallel occupancy consensus almost-local alr faults crash reads local linearizable cap impossibility theorem l2aw coherence viva thesis phd systems database serializability strict available reliable throughput poster fault datastore propagation early broadcast paxos latency sequential consistency caching symmetric writes sequential 2018 eurosys18 hot ccnuma cc lstm 2017 machine deep neural isca tpu cost accelerator google convolutional hardware learning networks asic cnn tensor model critical frameworks paper data deán matei hadoop jeff processing review big yarn jargons big data master slave issues rdds shuffle framework goals bottleneck