What exactly is std::atomic
C++ atomics and memory ordering
This is used to prevent the CPU from doing the instruction reordering.
An operation (or set of operations) is atomic, linearizable, indivisible or uninterruptible if it appears to the rest of the system to occur at once without being interrupted. Atomicity is a guarantee of isolation from interrupts, signals, concurrent processes and threads.
it is the requirement that one thread of execution never enter its critical section at the same time that another concurrent thread of execution enters its own critical section
不允许多线程对 critical section （可以是 share variables）的同时访问。
下面是用 linked list 的删除操作来说明。
两个线程同时进行删除操作，分别想删除 i 和 i+1，同时执行了
In mongoDB version before v3.2, $geoNear operation was sometimes slow for queries on dense datasets.
After version v3.2, this algorithm had been upgrade. This is a brief explanation about the upgrade.
The $geoNear algorithm iteratively expands its search in distance intervals (the red annulus shown below), aiming to have a few hundred documents per interval. Searching all documents in an interval is accomplished by finding an index cell covering. This covering ensures that all of the documents in the interval can be found using an index scan. The documents in the covering but not in the interval are filtered out afterwards. After all of the documents in an interval are found, they are sorted and returned.
This is an article about partitioning geo-index at SOCC17 from UCB by Anand.
The era of the Internet of things is imminent. Big data of IoT is inherently geospatial in nature and has unpredictable skews in space and time.
In order to overcome the skews and give the best performance, partition and re-balance are critical to the database and key-value store. However, frequent re-balance operation will harm insertion performance.
This is the paper presented at 2007’s SOSP by Amazon.
In order to provide high availability Dynamo sacrifices consistency under certain failure scenarios ().
Amazon’s primary usecase for Dynamo is its shopping cart, where it’s really important to be highly available, or customers should be able to view and add items to shopping cart dispite of datacenter failure.
In a cluster consists of millions of components, it always have small but significant number of server or network components failing at any giving time. Dynamo should be able to treats these as normal case without impacting availability or performance.