\section{Article 1: IBM study finds more than half of midmarket companies lack an integrated digtal strategy}
This article discusses about the lack of strong digital strategy in small or midsize companies. The digital strategy focuses on fusing of physical and digital resources in a company. This is mainly the job of CMO or CIO (information). In the age of Cloud computing, social networking and mobile computing, digital strategy becomes more and more important to compnay's strategic advantage over rivals.
The article gives two examples, in a social music industry and in a digital medical industry. Social music company, music mastermind, aims at enabling average person's ability to produce their own music. This marks the recent shift in music producing from traditional content provider, manufacture to average social users. To a tech startup, this promises new opportunities, as proven by the popularity of Web 2.0. Yet it also poses challenges. To a small company, to serve large scale userbase, it is not an easy task. And this is where IBM's Cloud solution comes in.
The same thing applies to digital medical industry, which now undergoes a revolution that digitalizes every physical medical records and personal DNA information. This promises better personal treatment, but also raises controversy like privacy issues when dealing doctor-patient exchange. The goal of the company, Coriel Life Science, is to make use of personal genomic data to better tailor the medical care. Security and privacy involved in this process, when the company wants to outsource the patient's personal and private genimic data (which is also huge) to a third party Cloud computing company, like IBM.
\section{Article 2: Salesforce denies any post-PRISM cloud trust problems}
Basically this article is a clarification from Salesforce to confirm the security of their Cloud platform. Security in third party Cloud causes a lot controversy especially after the revellation of PRISM incident, a US massive surveillance programme. Saleforces is a Cloud computing company which helps midsize company to utilize their sales data; It stores different kinds of company data, most being sensitive, including customer complaints, sales forecase and product information.
\section{Article 3: Presto: Interacting with petabytes of data analytics in Facebook}
This articles introduces presto: a SQL query engine that emphasizes the interactive query processing on big data.
Nowadays, typical workload that is run in Cloud includes the batch processing and real-time data serving. Their application ranges from simply data storing/retrieving to complex analytics with machine learning and data mining. In this domain, Presto is a computation engine that is optimized towards real-time interactive queries.
Architecture-wise, Presto is not MapReduce, but instead a stream processing engine to me. It looks like a distributed in-memory stream processing system, in which each stage models are with operators designed to support SQL semantics.
PL/System-wise, Presto is written in Java. The good side is Java is fast for development. However flip side is that Java add a JVM layer on top of the native operating system on where it is running. In order for high performance java system programming, the engineers in Facebook comes up with tricks to avoid low level inefficiency related to memory allocation and garbage collection.
Extensibility: Presto is computation engine which can work together with other Cloud systems. Most notably with HBase. Presto can be imposed on top of HBase to collectively support a fault-tolerant/persistent database with interactive performance for complex SQL queries. One task in their roadmap is on a high-performance HBase connector.
Presto is open sourced.
\section{Article 4: HBase Performance Testing}
This article addresses the performance evaluation for HBase. The writing is not very formal, yet the takeaway I found useful in this article is that
1.it confirms that HBase is write optimized system.
2.the methodology to do system evaluation includes multiple designs; typical ones include using benchmark to simulate normal or peak production conditions, stress testing which tests with a harder condition and may saturate the system, and sustained test which usually run test in long enough time in order to see behaves that does not surface in short time (e.g. does java GC come into play, does compaction occur?)
3.in HBase, most workload is not CPU bounded, but bounded by factor like network, diskIO, memoryLocking...
\section{Compare and Contrast}
These articles talk about emerging shift towards Cloud computing, including fusing digital and physical resource/data in small size company, security issues in Cloud in the post-PRISM age, a computation engine released by Facebook for realtime analysis in Cloud, and performance evaluation of Cloudy storage software HBase.
