bgbenchmark.org
Publications - BG Benchmark
http://www.bgbenchmark.org/BG/publications.html
Sumita Barahmand and Shahram Ghandeharizadeh. BG: A Benchmark to Evaluate Interactive Social Networking Actions. Proceedings of the biennial Conference on Innovative Data Systems Research (CIDR), 2013. USC DBLAB Tech Report 2012-06. Sumita Barahmand and Shahram Ghandeharizadeh. D-Zipfian: A Decentralized Implementation of Zipfian. Proceedings of the Sixth International Workshop on Testing Database Systems (ACM SIGMOD DBTest), 2013. USC DBLAB Tech Report 2012-04. USC DBLAB Tech Report 2012-08. Extensions ...
bgbenchmark.org
Rating Data Stores - BG Benchmark
http://www.bgbenchmark.org/BG/coordinator.html
Figure 1. MongoDB's SoAR and Socialites Rating for a given workload. Highest number of completed actions per second that satisfy the specified SLA. Highest number of simultaneous threads that issue requests against the data store and satisfy the specified SLA. It quantifies the multithreading capability of the data store and whether it suffers from limitations such as the convoy phenomena that diminish its throughput with a large number of simultaneous requests. Figure 2. BG's rating architecture. These ...
bgbenchmark.org
Downloads - BG Benchmark
http://www.bgbenchmark.org/BG/downloads.html
17 Oct 2015 . 2014-2015 USC Database Lab.
bgbenchmark.org
Developers Manual - BG Benchmark
http://www.bgbenchmark.org/BG/manual.html
To evaluate a data store using BG, one must follow these eight steps:. Create a schema for the target data store,. Design an implementation of alternative BG actions to fine tune schema of Phase 1,. Initialize client component of the target data store,. Implement schema using the target data store,. Load schema with data,. Implement alternative BG actions,. Quantify Response time, throughput and amount of unpredictable data measurements of a data store for a specific setting,. BG consists of 3 components:.
bgbenchmark.org
Image Size Matters - BG Benchmark
http://www.bgbenchmark.org/BG/observations.html
The presence of a profile image and its size may impact system performance dramatically. Using BG, an evaluator may benchmark a data store with different image sizes for member profile. This includes the possibility of no image at all. See BG's data model. VP) actions to observe a service time equal to or faster than 100 milliseconds. SQL-X: A commercial strength relational database management system. Due to licensing agreement, we cannot disclose the identity of this system. 2014-2015 USC Database Lab.
bgbenchmark.org
Unpredictable Data - BG Benchmark
http://www.bgbenchmark.org/BG/unpredictabledata.html
Unpredictable data is either stale, inconsistent, or simply invalid data produced by a data store. For example, the design of a Cache Augmented SQL (CASQL) system. May incur dirty reads. Or suffer from race conditions that leave the cache and the database in an inconsistent state. A data store may employ an eventual consistency. Technique that produces either stale or inconsistent data for some time. The requirements of an application dictate whether these techniques are appropriate or not. J Dean and S&...
bgbenchmark.org
BG Visualization Deck - BG Benchmark
http://www.bgbenchmark.org/BG/bgdeck.html
BG visualization deck enables a user to specify parameter settings for BGCoord, initiate rating of a data store, and monitor the rating process. Once the rating is initiated, the BGCoord aggregates the rating parameters from the BGClients and reports metrics such as throughput, response time and BGClient health statistics such as average disk queue length and CPU utilization to BG's visualization deck for display. 2014-2015 USC Database Lab.
bgbenchmark.org
Social Actions - BG Benchmark
http://www.bgbenchmark.org/BG/socialactions.html
BG emulates members of a social networking site performing simple operations named actions. These are as follows:. View Profile, VP: Member A retrieves the profile information of Member B. A member may retrieve self profile. List Friends, LF: Member A lists friends of Member B. View Friend Requests, VFR: Member A views pending friend requests. Invite Friend, IF: Member A invites Member B to be a friend. Accept Friend Request, AFR: Member A accepts Member B's friendship invitation. Reply to a Tweet.
bgbenchmark.org
Data Model - BG Benchmark
http://www.bgbenchmark.org/BG/datamodel.html
Figure 1. Conceptual data model of BG's database. Relational data model of BG's database. Username, pw, firstname, lastname, gender, job, jdate, ldate, address, email, tel). Type, body, doc). Timestamp, type, content). JSON-like data model of BG's database. Dob: "10.24.1986". Jdate: "01.01.2011". Ldate: "09.20.2012". Address: "941, Milkyway, San Jose". Tel: "323 000 0000". Doc: "19.08.2012". Timestamp: "19.08.2012 10:00:00". Timestamp: "19.08.2012 14:20:00". 2014-2015 USC Database Lab.
bgbenchmark.org
D-Zipfian - BG Benchmark
http://www.bgbenchmark.org/BG/dzipfian.html
BG uses the Zipfian distribution to generate its workload of requests. This distribution enables BG to emulate scenarios such as 20% of members generating 80% of requests. The exponent of the distribution controls the degree of skewness. BGCoord partitions members of the social network across the N BGClients logically, requiring each BGClient to generate requests for its fragment of members. This enables BG to scale to a large number of BGClients. Decentralized Zipfian. 2014-2015 USC Database Lab.
