How do I plan and select a database server? (3) Evaluate the CPU of the database server based on the calculation of TPC-C)

The TPC-C benchmark is a performance benchmark established by the transaction processing board dedicated to demonstrating online transaction processing performance (OLTP, it is measured to enable customers to evaluate the performance of different online transaction processing systems that run in a standard database in a controllable state.

TPC-C testing includes five typical OLTP transactions that are:

New Order: a user submits a new order

Payment: updates the user's account balance to reflect a payment

Delivery: Delivery of orders (implemented through a batch of transactions)

Order Status: return the status of the user's latest order

Inventory level: monitor the inventory of the current Warehouse

The transaction processing of the TPC-C is implemented on a database of 9 tables, including: Update, insert, delete, terminate, and access to the primary and secondary keys, 90% of the response time for each transaction should be less than or equal to 5 seconds, and the inventory level response time can be less than 20 seconds.

The throughput value of the TPC-C is the direct result of the terminal activity level, for example, each warehouse has 10 terminals, the above five transactions are available on each terminal, A remote terminal simulator is used to perform necessary transaction hybrid operations during performance testing. This hybrid represents a complete order business processing process: entry, payment, inspection, delivery. More professionally, this necessary mix is defined as generating an equal number of new order and payment transactions, and generating a delivery transaction in every 10 new order transactions, one order status check transaction and one inventory Level Check Transaction

The remote terminal simulator is also used to measure the response time of each transaction and simulate the input time and thinking time. The input time refers to the time spent inputting data on the terminal, thinking time refers to the time taken by the operator before the terminal reads the transaction result and initiates the next transaction request. Every thing has a minimum input time and a minimum thinking time. In addition, the response time must be under a given limit value.

TPC-C benchmark test results-TPC-C throughput (tpmc), representing the maximum continuous performance of the system, which is defined as how many new order transactions the system can process per minute, at the same time, the system is still processing four other transaction types (payment, order status, delivery, and inventory level ). All five TPC-C transactions have a limited user response time requirement in which the response time for the new order transaction is less than 5 seconds. Therefore, if a system's TPC-C value is 100 tpmc/min, it means that the system can generate 100 new order transactions while processing other mixed TPC-C transactions per minute.

How to evaluate servers with TPC-C

From the above, we can see that the TPC-C testing benchmark is mainly used to test the number of online transactions that can be processed by the host server per minute, the Unit result of the test is TPM value (transaction per minute, that is, the number of transactions processed per minute ).

Although the TPC-C objectively reflects the system processing performance of various computer manufacturers, and the testing benchmark is constantly improved to be closer to the real application of the transaction environment, however, it still cannot be fully consistent with the various practical applications, and the host systems that participate in the TPC test have been properly optimized. Therefore, when selecting a host server as a carrier in the actual business application system, you must consider multiple factors to meet the production requirements of the application system to the greatest extent possible.

The following formula is an empirical methodology summarized by IBM in the practical application of the financial integrated business system. It basically reflects the host processing capability requirements of financial business characteristics:

TPM = task x 80% x s x F/(t x C)

Where:

Task: calculates the peak transaction volume for the daily business.

T: indicates the daily peak transaction time. Assume that the daily 80% transaction volume is concentrated in 4 hours per day, that is, within 240 minutes: t = 240.

S: Ratio of complexity of actual banking transaction operations to standard TPC-C benchmark environmental transactions. Because of the complexity of the actual financial business transactions and TPC? There is a big difference in transactions in the C standard test, and a reasonable corresponding value must be set. Taking ordinary savings business transactions as an example, a transaction often needs to open a large number of database tables at the same time, take out the relevant data for operation, relative to the complexity of TPC-C standard transactions, much more complex; based on scientific statistics, the complexity of each transaction operation compared with that in the TPC standard test can be set to 10 to 10 ~ 20.

C: The CPU balance of the host. Practical application experience shows that the CPU usage of a host server is higher than 80%, indicating that the CPU usage is too high, leading to a system bottleneck. When the utilization rate is 75%, it is in the best utilization status. Therefore, when calculating host performance indicators, you must consider CPU redundancy and set C = 75%.

F: for the future of the system 3 ~ 5 years of redundant reserved business volume development.

In conclusion, in order to ensure the online business processing performance requirements, we can calculate the processing capability required by the host and obtain the corresponding model and configuration accordingly.

Trackback: http://biz.chinabyte.com/462/2293962_2.shtml