GB/T 47177.2-2026Quantum computing service platform - Part 2: Performance evaluation (English PDF)

6.1 Performance Testing of Quantum Computer Systems

6.1.1 Performance Indicators The performance requirements of quantum computing service platforms for quantum computer systems mainly include the following indicators.

a) Number of qubits.

b) Connectivity.

c) Bit coherence.

---Longitudinal relaxation time;

---Lateral relaxation time.

d) Operational fidelity.

---Read fidelity;

---Single-bit gate fidelity;

---Two-bit gate fidelity.

e) Operation time.

---Single-bit gate operation duration;

---Two-bit gate operation duration.

6.1.2 Test Method According to GB/T 47468-2026.

6.2 Performance Testing of the Classical Simulator for Quantum Resources

6.2.1 Number of Classic Emulators The number of classical simulators refers to the number of classical simulators that a quantum computing service platform can provide for parallel execution. This is defined here. This refers to the number of full-amplitude simulators supported. The testing steps for the number of classic simulators are as follows:

a) Under a limited simulation computation method, evaluate the highest quantum level that each individual simulator in the evaluation system can support when performing quantum simulations. The number of bits (basically determined by the upper limit of memory capacity), denoted as Nm;

b) Due to limitations in system scheduling and storage capacity, the standard for the number of running qubits is set to the number of qubits that can be executed in parallel. Eye, denoted as Nm-2;

c) Open multiple emulators and run them simultaneously on random lines;

d) Gradually increase the number of parallel processes, ensuring the system can run stably simultaneously without memory errors or resource queuing. The maximum number of successful parallel runs is the number of classic simulators. Example. A classic simulator can simulate a maximum of 35 qubits at full amplitude by default. During testing, 33 qubits will be activated in parallel. The full-amplitude simulation calculation is performed using a randomly sampled circuit library. The maximum number of parallel executions the system can support at this point is equal to the number of classic simulators. This is limited by... Due to system scheduling and storage limitations, the actual number of classic emulators that can run may be lower than the theoretical maximum. Theoretically, it can support four classic emulators simultaneously, but in practice... Typically, a maximum of two classical simulators are supported. Therefore, if several classical simulators are deployed within a quantum computing service platform, they can all perform parallel testing simultaneously. The number of concurrently running simulators on the final system is the maximum number of classic simulators.

6.2.2 Bit Scale The bit size refers to the highest number of qubits that a classical simulator provided by a quantum computing service platform can accept for a task. The approximate solution method that does not support problem partitioning or subspace partitioning is not supported. For a given classical simulator, the maximum number of qubits supported is... The larger the value, the stronger the computing power. The test steps for bit-scale are as follows:

a) Select a single type of classical simulator provided by the quantum computing service platform as the test object;

b) Construct a randomized benchmark quantum circuit with circuit parameters of number of qubits N and circuit depth D, where N starts from an initial value N0. Initially, increment by step size DeltaN, while keeping D at a fixed value of D0;

c) Start and initialize the classic simulator under test, select the execution backend (CPU or GPU), and disable optimizations such as approximation, truncation, and chunking. Options;

d) Perform quantum circuit simulation for each value of N and output the final state sampling result, recording the running state;

e) Gradually increase N until the first "out of memory" or "allocation failure" error occurs. Record the maximum number of bits without an error as. Nmax is the bit size.

6.2.3 Maximum Line Depth Maximum line depth refers to the maximum depth of the simulated lines supported by the classical simulator provided by the quantum computing service platform. The test procedure for maximum line depth is as follows:

a) Select a single type of classical simulator provided by the quantum computing service platform as the test object;

b) Construct a randomized benchmark quantum circuit with the parameters of N qubits and D layers, where N remains a fixed value. N0,D starts from the initial value D0 and increases by a step size DeltaD;

6.3 Service Performance Testing

6.3.1 Task Response Time Task response time refers to the time taken by a quantum computing service platform to resolve, schedule, and execute computing tasks. The shorter the time, the better. The stronger the performance of the service platform, the better. The steps for testing task response time are as follows:

a) Submit a computational task to the quantum computing service platform;

b) The platform responds to the task request and records the time point t1 at this point;

c) The platform processes and executes the task until it is completed, and records the time point t2 at this point;

d) Extract the task parsing time (Tparse), task scheduling time (Tschedule), resource switching time (Thandoff), and actual time from the tool logs. The raw data of Toperation for task execution time;

e) Calculate the task response time according to formula (1).

6.3.2 Task Concurrency Task concurrency refers to the ability of a quantum computing service platform to process multiple quantum computing tasks simultaneously, reflecting the system's ability to handle multiple tasks at the same time. The maximum number of tasks that can be handled within a given time period. The steps for testing task concurrency are as follows:

a) Use log analysis methods to collect the service platform's operational logs, including task request logs, system operation logs, etc., and statistically analyze them. The number of task requests processed within a given time period, and the task execution time, etc.

b) Analyze the service platform's operation logs, outputting the peak number of task requests processed per unit of time (QPSpeak), and the actual number of requests executed per unit of time. Find the average time T.

c) Calculate the task concurrency according to formula (2).

6.3.3 Load Balancing Recovery Time Load balancing recovery time refers to the time a quantum computing service platform can recover from node failures in tasks involving high concurrency and resource allocation. Or, under conditions of uneven load, the time required for the system to recover to a balanced state. The test steps for load balancing recovery time are as follows:

a) Use the load balancing monitoring tool built into the quantum computing service platform to pre-configure monitoring parameters (including but not limited to load balancing). (Data such as node CPU utilization, memory usage, task request response latency, and backend server liveness status) ensures the tool can function effectively. Collect and record the above data in a timely manner;

b) Simulate node failure manually and record the time point Tfp of the failure occurrence using monitoring tools;

c) Continuously monitor the system status until the load balancing state is restored to normal system operation, and record the time point Trp at this point;

6.3.5 Throughput

6.3.5.1 HTTP Throughput HTTP throughput is a measure of a quantum computing service platform's ability to securely process HTTP requests per unit of time, reflecting its... Performance of security capabilities such as DDoS attack resistance in high-concurrency scenarios. The steps for testing HTTP throughput are as follows:

a) Enable application firewall and other security attack protection functions on the quantum computing service platform;

b) Test the HTTP throughput of the quantum computing service platform using testing instruments or tools;

c) With page sizes set to 64KB, 32KB, 16KB, and 1KB, the measurement platform experienced no application-layer service transmission failures. It can also properly identify and block HTTP throughput under attack conditions.

6.3.5.2 HTTPS Throughput HTTPS throughput refers to the efficiency with which a quantum computing service platform can handle encrypted requests when security encryption and protection mechanisms are enabled. And maintain the comprehensive ability to ensure the stability of services such as security verification. The steps for testing HTTP throughput are as follows:

a) Enable security attack protection functions related to the quantum computing service platform;

b) Test the HTTPS throughput of the quantum computing service platform using test instruments or tools;

c) When selecting test subjects, high-strength cipher suites that meet national cryptographic management requirements and possess forward security should be chosen;

d) With page sizes set to 64KB, 32KB, 16KB, and 1KB, the measurement platform was tested under conditions of no application-layer service transmission loss. HTTPS throughput under conditions of failure and normal identification and blocking of attacks.

6.3.6 New Connection Speed

6.3.6.1 HTTP New Connection Rate HTTP new connection rate is used to measure the ability of a quantum computing service platform to resist high-frequency connection attacks (such as CC attacks, HTTP attacks, etc.). In the event of a flood, it provides the ability to quickly establish and securely handle legitimate requests. The steps for testing the HTTP new connection speed are as follows:

a) Use testing instruments or tools to test the HTTP new connection rate of the quantum computing service platform;

b) Set the page size to 64KB and measure the HTTP connection establishment rate of the platform in the case of application layer business transmission failure.

6.3.6.2 HTTPS New Connection Speed HTTPS new connection rate refers to the speed at which a quantum computing service platform can quickly establish a new connection under high-strength encrypted handshake and security protection pressure. The comprehensive capability to establish trusted communication channels and defend against resource exhaustion attacks. The steps for testing the speed of newly established HTTPS connections are as follows:

a) Use testing instruments or tools to test the rate of newly established HTTPS connections on the quantum computing service platform;

b) When selecting test subjects, high-strength cipher suites that meet national cryptography management requirements and possess forward security should be chosen;

c) Set the page size to 64KB and measure the rate of establishing new HTTPS connections on the platform when no application layer service transmission fails.