Shared DB - Linux

Disclaimer

These results should be viewed as guidelines and not performance guarantees, since there are many variables that affect performance (file set, network configurations, hardware characteristics, etc.). If throughput is important to your implementation, OPSWAT recommends site-specific benchmarking before implementing a production solution.

Environment

Using AWS environment with the specification below:

MetaDefender Core

	OS	AWS instance type	vCPU	Memory (GB)	Network bandwidth (Gbps)	Disk type	Benchmark
MetaDefender Core #1	Ubuntu 24.04	c5.4xlarge	16	32	Up to 10	SSD	Amazon EC2 c5.4xlarge - Geekbench
MetaDefender Core #2	Ubuntu 24.04	c5.4xlarge	16	32	Up to 10	SSD	Amazon EC2 c5.4xlarge - Geekbench

RDS

OS	AWS instance type	vCPU	Memory (GB)	Network bandwidth (Gbps)	Disk type
Windows Server 2022	db.m7i.4xlarge	16	64	Up to 10	SSD

Deployment Model

Using a AWS Load Balancer to distribute files sent from the client tool to two (2) different MetaDefender Core servers applying Round Robin algorithm. With this algorithm, each MetaDefender Core server is supposed to receive same number of requests.

Client tool

A simple tool written in Python to collect files in a designated folder and submit requests to Load Balancer mentioned above.

files_to_scan = list of files to scan scan_futures =[] for file_path in files_to_scan: scan_futures.append(asyncio.create_task(self.scan_file(file_path, self.load_balancer_url)) async def scan_file(self, file_path, core_url): api_url = f'{core_url}/file' with open(file_path, 'rb') as file: file_content = file.read() headers = {'Content-Type': 'application/octet-stream', 'filename': file_path} max_retries = 20 # Maximum number of retry attempts retry_delay = 120 # Delay in seconds before retrying for attempt in range(max_retries + 1): try: starttime = time.time() response = requests.post(api_url, data=file_content, headers=headers) endtime = time.time() status_code = response.status_code if status_code == 200: self.post_wait = self.post_wait + endtime - starttime self.num_post_req += 1 response_json = response.json() data_id = response_json.get('data_id') return status_code, response_json, data_id if status_code == 503: print(f"Received status code {status_code}. Retrying in {retry_delay} seconds...") await asyncio.sleep(retry_delay) else: return None, None, None except requests.RequestException as e: return None, None, None return None, None, None

OS	AWS instance type	vCPU	Memory (GB)	Network bandwidth (Gbps)	Disk type
CentOS 7	c5.4xlarge	16	32	Up to 10	SSD

Dataset

Detailed information of dataset below will be used for testing:

File category	File type	Number of files	Total size (MB)	Average file size (MB)
Adobe	PDF	370	385 MB	1.0 MB
Executable	EXE	45	309.5 MB	6.9 MB
	MSI	15	45.75 MB	3.1 MB
Image	BMP	80	515 MB	6.4 MB
	JPG	420	237.5 MB	0.6 MB
	PNG	345	169 MB	0.5 MB
Media	MP3	135	865 MB	6.4 MB
	MP4	50	500 MB	10.0 MB
Office	DOCX	235	190 MB	0.8 MB
	DOC	225	486 MB	2.2 MB
	PPTX	365	860 MB	2.4 MB
	PPT	355	1950 MB	5.5 MB
	XLSX	340	283.5 MB	0.8 MB
	XLS	335	284.5 MB	0.8 MB
Text	CSV	100	236 MB	2.4 MB
	HTML	1075	76 MB	0.1 MB
	TXT	500	210 MB	0.4 MB
Archive	ZIP	Compressed files: 10 Extracted files: 270	Compressed size: 125.5 MB Extracted size: 156.5 MB	Avg compressed size: 12.6 MB Avg extracted size: 0.6 MB
Summary (compressed)		5000	7728.5 MB	1.55 MB average file size
Summary (extracted)		5260	7759.5 MB	1.48 MB average file size

Product Information

Product versions:

MetaDefender Core 5.16.0
Engines:
- Metascan 5: Ahnlab, Bitdefender, ClamAV, ESET, K7
- Metascan 10: Metascan 5 + Avira, Varist, IKARUS, Quick Heal, TACHYON
- Metascan MAX: Metascan 10 + Lionic, CMC, CrowdStrike Falcon ML, Aurora, Trellix, NANOV, RocketCyber, Sophos, Webroot SMD, Xvirus Anti-Malware
- Deep CDR: 7.6.0
- Proactive DLP: 3.0.0
- Archive: 7.6.0
- File type analysis: 7.6.0
- File-based vulnerability assessment: 4.57-236

MetaDefender Core settings

General settings

Turn off data retention
Turn off engine update

Archive extraction settings

Max recursion level: 99999999
Max number of extracted files: 99999999
Max total size of extracted files: 99999999
Timeout: 10 minutes
Handle archive extraction task as Failed: true
- Extracted partially: true

Metascan AV settings

Max file size: 99999999
Scan timeout: 10 minutes
Per engine scan timeout: 1 minutes

Performance test results

MetaDefender Core with single engine (technology)

Summary metrics:

System resource utilization:

MetaDefender Core with common engine packages

Summary metrics:

System resource utilization:

Recommendations

Controlling total processing time of each MD Core server:

In this deployment model, we should organize and send files in the way that it best utilizes the load of each MD Core server. It is not a good practice if one Core server is free while the other one is busy. By optimizing the distribution of files, we can ensure that each Core server is utilized efficiently, thereby improving overall system performance. Furthermore, this approach can help prevent bottlenecks and minimize the chances of system overload.

Adding proper number of MD Core servers to the cluster:

Adding more Core servers to this model will increase more load on the shared database. When adding a new MD Core server, users should monitor performance of database server such as memory/CPU consumption, disk usage, network bandwidth, request response time and so on… to see if it still can handle the load. This is important in order to maintain optimal performance and ensure that the database server can continue to efficiently serve the needs of the system.

Optimizing database server for better performance:

Continuing to add more Core servers to this model may result in increased strain on the shared database. As such, it is crucial to ensure that the database is optimized to handle the additional load effectively. Users can consider adjusting default database settings of PostgresSQL to optimize for more data load if needed. Here is where we can adjust PostgresSQL database settings: <PostgreSQL install location\version>\data\postgresql.conf.

Besides that, MD Core also supports a parameter (db_connection) for users to specify max connections that MD Core can handle, take a look at this guideline: MetaDefender Configuration.

Last updated on

Was this page helpful?