2017-02-27Home
Last week, we mentioned announcement of Google Cloud Spanner. There is an "open source version", CockroachDB, built by three ex-Googlers who also founded a company around it. We would read about how Google Spanner Inspires CockroachDB To Outrun It this week.
Without Google's private network and atomic clock, CockroachDB provides serializer isolation, weaker than Spanner's linearizable isolation. CockRoachDB is based on RocksDB using etcd(Raft) for consensus. It has compatible interface with Postgres, supporting most SQL 92, some SQL 2011 and Spanner functions. They've employed jepsen for consistency analysis and found bugs leading to serializability violations. The software is still in beta at the time of writing. Let's see how it will turn out.
Now, what else ?
Apache Samza 0.12.0 has been released with the support of both unbounded and bounded sources, or "Convergence of Batch and Real-time processing" as said.
Apache Kafka has made a feature release 0.10.2.0 which includes the completion of 15 KIPs, over 200 bug fixes and improvements. Among them are supports for session windows and global table in Kafka Streams.
Sharing our experiences with Virtual Reality(VR) can't be difficult, and that's where Mixed Reality(MR) comes in. With MR, we can see the virtual environment of a VR user except the facial expressions blocked by the headset. Now Google researchers have been working on a solution to "remove" the headset.
Preprocessing is required to apply machine learning to real world datasets, which can be done separately through a large scale data processing framework such as Apache Spark and Apache Flink. Now Google announced tf.Transform to make it part of a TensorFlow graph. That is achieved through Apache Beam whose applications can be run on Spark, Flink and Google's own Cloud Dataflow.
A piece of "old" news from Google, Announcing TensorFlow Fold: Deep Learning With Dynamic Computation Graphs. TensorFlow Fold is to address challenge that data of varying size and structure don't batch together. You may read their paper for more details.
Google has found the first SHA1 collision.
The attacker could then use this collision to deceive systems that rely on hashes into accepting a malicious file in place of its benign counterpart. For example, two insurance contracts with drastically different terms.
Here is [an interesting site showing lifetimes of cryptographic hash functions](cryptographic hash functions).
The other breaking news is CloudBleed
Essentially, web requests to Cloudflare-backed sites received answers which included random information from other Cloudflare-backed sites!
Here is a list of possibly affected domains.
Dropbox have open sourced Securitybot, to "automatically confirm and aggregate suspicious behavior with employees on a distributed scale".
That's all for the week. Super happy to keep up the momentum.