Leming TPC-E described Benchmark
Prospects
In recent years, IT field presents rocket development, technological idealism and sense of national identity inspired innovation domestic databases, thanks to the support of national policy, domestic database has been flourishing. Although the database product has been the actual application, but to be widely recognized by the industry, one of the world out of the country, and even issued a challenge to the traditional mainstream database, we need to select domestic and international recognized and authoritative benchmark test proved.
Database is the core of an enterprise, regardless of how the company's business, the database can only recognition task, unknown to these tasks is mainly divided into two categories: one is transaction processing, one is intelligent analysis. For transaction processing, you can select domestic and international recognition of the TPC-E benchmark testing.
Introduction benchmark
TPC-E benchmark is the stock trading business scene as a carrier test, select the stock trading business scenario in mind as it is typical of the financial industry, compared with other industries have higher performance requirements, type of operation and rich, with a wide range of representatives sex, classic OLTP system. TPC-E benchmark test using a unified data model to test, that is, the measured test data and do not care about the database itself, but using the same principles to generate stock trade data.
TPC-E benchmark reality stock-trading operations based on abstract 10 kinds of database transactions and their mixing ratio, describes the relationship between the client, securities companies, stock exchanges, as follows:
Resolve
On the surface, the whole structure as the core securities company, securities companies operating always represent the wishes of customers, so in essence, testing or buying and selling stocks around customer behavior unfold, customers for stock transaction request, the securities company after receipt of the request on behalf of clients the request is sent to the stock exchange, trading results and feedback received to the customer. In order to maintain a clean environment and updated test data before the tests, the reference also provides two kinds of matters unrelated to the business scenarios that transaction cleanup services, data maintenance services.
Customer body acts as a very important role, which determines the size of the data, the number of users based on a reasonable set of database capabilities before the test, but should meet customer number is not an integer multiple of less than 5000 and 1000. Customers accounts to buy shares, an average of five per customer account, the average number of shares traded each account is 10, so the average number of shares for each customer transaction is 50. Customer orders proposed transaction request in 60% of market orders (buy 30%, 30% sold) expressed as the current market price transactions; the other 40% of limit orders (buy 20%, 20% sell), represents the specified price before the transaction. Each successful submission of an order, will produce a corresponding result, TPC-E benchmark rate produce a result that is used to measure system performance, called the transaction throughput.
We continue to emphasize the TPC-E benchmark is a standard height simulation, the simulation is reflected in the rich variety of services, reasonable mixing ratio, strict constraints on response time, throughput, etc. and the validity of the transaction. On the other hand, TPC-E benchmark is a basic standard, testing is in this configuration, the database whether they have the basic skills, rather than the ability of the test database in the end how strong, which is why its results must meet the criteria throughput the range of 80% -102% to be considered valid.
实际应用系统数据库中的数据可能与股票行业相去甚远，TPC-E基准测试使用股票交易行业数据进行测试，却说体现的是实际应用系统中数据库的性能，该如何理解？
Database is a technology database performance testing is reflected in the ability of the database, no matter what kind of applications, for databases are CRUD, so choose the stock trading industry for testing, operating rich enough to show capability to science comprehensive assessment of the performance of the database, and the use of uniform data to test the results are comparable.
