TPC BENCHMARK (TM) H (779138), страница 21
Текст из файла (страница 21)
Any other operations on these data, such as data formatting or database activity,must be included in the execution and the timing of the refresh functions.5.4MetricsTPC-H defines the following primary metrics:The TPC-H Composite Query-per-Hour Metric (QphH@Size) is the performance metric, defined in Clause5.4.3;The price-performance metric is the TPC-H Price/Performance ($/QphH/@Size) and is defined in Clause5.4.4;The Availability Date of the system, defined in Clause 0 of the TPC Pricing Specification .When TPC_Energy option is chosen for reporting, the TPC-H energy metric reports the power per performance andis expressed as Watts/KQphH@Size. (see TPC-Energy specification for additional requirements)No other TPC-H primary metric exists.
However, secondary metrics and numerical quantities such as TPC-H Powerand TPC-H Throughput (defined in Clause 5.4.1 and Clause 5.4.2 respectively) and S, the number of query streamsin the throughput test, must be disclosed in the numerical quantities summary (see Clause 8.4.4).5.4.1TPC-H Power5.4.1.1 The results of the power test are used to compute the TPC-H query processing power at the chosen database size. Itis defined as the inverse of the geometric mean of the timing intervals, and must be computed as:3600 * SF24TPC-H Power@Size =i 22j 2i 1j 1 QI (i,0) * RI ( j,0)Where:TPC BenchmarkTM H Standard Specification Revision 2.17.1Page 98QI(i,0) is the timing interval, in seconds, of query Q i within the single query stream of the power test (seeClause 5.3.7)RI(j,0) is the timing interval, in seconds, of refresh function RFj within the single query stream of thepower test (see Clause 5.3.7)Size is the database size chosen for the measurement and SF the corresponding scale factor, as defined inClause 4.1.3.Comment: the power numerical quantity is based on a query per hour rate (i.e., factored by 3600).5.4.1.2 The units of TPC-H Power@Size are Queries per hour * Scale-Factor, reported to one digit after the decimal point,rounded to the nearest 0.1.5.4.1.3 The TPC-H Power can also be computed as: 13600 * exp 24TPC-H Power@Size =j 2i 22lnQIi,0lnRIj,0 * SFj 1 i 1Where:ln(x) is the natural logarithm of x5.4.1.4 If the ratio between the longest query timing interval and the shortest query timing interval in the power test isgreater than 1000 (i.e., max[QI(i,0)]/min[QI(i,0)] > 1000), then all query timing intervals which are smaller thanmax[QI(i,0)]/1000 must be increased to max[QI(i,0)]/1000.
The quantity max[QI(i,0)]/1000 must be treated as atiming interval as specified in Clause 5.3.7.5 for the purposes of computing the TPC-H Power@Size.Comment: The adjusted query timings affect only TPC-H Power@Size and no other component of the FDR.5.4.2TPC-H Throughput Numerical Quantity5.4.2.1 The results of the throughput test are used to compute TPC-H Throughput at the chosen database size. It is definedas the ratio of the total number of queries executed over the length of the measurement interval, and must becomputed as:TPC-H Throughput@Size = (S*22*3600)/Ts *SFWhere:Ts is the measurement interval defined in Clause 5.3.6S is the number of query streams used in the throughput test.Size is the database size chosen for the measurement and SF the corresponding scale factor, as defined inClause 4.1.3.5.4.2.2 The units of TPC-H Throughput@Size are Queries per hour * Scale-Factor, reported to one digit after the decimalpoint, rounded to the nearest 0.1.5.4.3The TPC-H Composite Query-Per-Hour Performance Metric5.4.3.1 The numerical quantities TPC-H Power and TPC-H Throughput are combined to form the TPC-H composite queryper-hour performance metric which must be computed as:QphH@Size =Power @ Size * Throughput @ SizeTPC BenchmarkTM H Standard Specification Revision 2.17.1Page 995.4.3.2 The units of QphH@Size are Queries per hour * Scale-Factor, reported to one digit after the decimal point, roundedto the nearest 0.1.5.4.4The TPC-H Price/Performance Metric5.4.4.1 The TPC-H Price/Performance metric at the chosen database size, TPC-H Price-per-QphH@Size , must be computed using the performance metric QphH@Size as follows:TPC-H Price-per-QphH@Size = $/QphH@SizeWhere:$ is the total system price in the reported currency.
The list of components to be priced is described inClause 7.1 of this specification. How to price the components and how to express the total systemprice are defined in Clause 7 of the TPC Pricing Specification.QphH@Size is the composite query-per-hour performance metric defined in Clause 5.4.3.Size is the database size chosen for the measurement, as defined in Clause 4.1.3.5.4.4.2 The units of Price-per-QphH@Size are expressed as in Clause 7 of TPC Pricing Specification.
In the United Statesthe price performance is expressed as USD per QphH@Size rounded to the highest cent (e.g., $12.123 must beshown as $12.13USD for price/performance).5.4.5Fair Metric Comparison5.4.5.1 Comparisons of TPC-H benchmark results measured against databases of different sizes are believed to be misleading because database performance and capabilities may not scale up proportionally with an increase in database sizeand, similarly, the system price/performance ratio may not scale down with a decrease in database size.If results measured against different database sizes (i.e., with different scale factors) appear in a printed or electroniccommunication, then each reference to a result or metric must clearly indicate the database size against which it wasobtained.
In particular, all textual references to TPC-H metrics (performance or price/performance) appearing mustbe expressed in the form that includes the size of the test database as an integral part of the metric’s name; i.e.including the “@size” suffix. This applies to metrics quoted in text or tables as well as those used to annotate chartsor graphs. If metrics are presented in graphical form, then the test database size on which metric is based must beimmediately discernible either by appropriate axis labeling or data point labeling.In addition, the results must be accompanied by a disclaimer stating:“The TPC believes that comparisons of TPC-H results measured against different database sizes are misleading anddiscourages such comparisons”.5.4.5.2 Any TPC-H result is comparable to other TPC-H results regardless of the number of query streams used during thetest (as long as the scale factors chosen for their respective test databases were the same).5.4.6Required Reporting ComponentsTo be compliant with the TPC-H standard and the TPC's fair use policies, all public references to TPC-H results fora given configuration must include the following components:The size of the test database, expressed separately or as part of the metric's names (e.g., QphH@10GB);The TPC-H Performance Metric, QphH@Size;The TPC-H Price/Performance metric, $/QphH@Size;The availability date of the priced configuration (see Clause 7 of the TPC Pricing Specification).Following are two examples of compliant reporting of TPC-H results:Example 1: At 10GB the RALF/3000 Server has a TPC-H Composite Query-per-Hour metric of 3010 when runagainst a 10GB database yielding a TPC-H Price/Performance of $1,202 per query-per-hour and will be available 1Apr-99.Example 2: The RALF/3000 Server, which will start shipping on 1-Apr-99, is rated 3,010 QphH@10GB and 1202$/QphH@10GB.TPC BenchmarkTM H Standard Specification Revision 2.17.1Page 1006: SUT AND DRIVER IMPLEMENTATION6.1Models of Tested Configurations6.1.1The tested and reported configuration(s) is composed of a driver that submits queries to a system under test (SUT).The SUT executes these queries and replies to the driver.
The driver resides on the SUT hardware and software.6.1.2Figure 3: Two driver/SUT configurations, a “host-based” and a “client/server” configuration illustrates examples ofdriver/SUT configurations. The driver is the shaded area. The diagram also depicts the driver/SUT boundary (seeClause 5.2 and Clause 5.3) where timing intervals are measured.Figure 3: Two driver/SUT configurations, a “host-based” and a “client/server” configurationHost Systems*DRIVER*QueryExecution&DatabaseAccessServer(s)Client(s)*DRIVERNetwork***QueryExecution*NetworkDatabaseAccessNetwork*Items marked by an * are optional6.2System Under Test (SUT) Definition6.2.1The SUT consists of:The host system(s) or server(s) including hardware and software supporting access to the databaseemployed in the performance test and whose cost and performance are described by the benchmark metrics;One or more client processing units (e.g., front-end processors/cores/threads, workstations, etc.) that willexecute the queries (if used);The hardware and software components needed to communicate with user interface devices;The hardware and software components of all networks required to connect and support the SUTcomponents;Data storage media sufficient to satisfy both the scaling rules in Clause 4: and the ACID properties ofClause 3: .
![](https://s.studizba.com/z.php?f=/templates/si/i/noavatar.png&w=100&h=100&t=1"){kind=avatar}
