Cloud Services
April, 2023
Applications
At the user endpoint, only an evergreen browser is required. Major releases will be tested on the current versions of:
When using outdated and/or other browsers, your mileage may vary. 3D features require hardware acceleration.
The Lighthouse audits, being run with slow network connectivity, result in these spectacular scores of a typical Vision 42 ® application:
100% - Performance
100% - Progressive Web App
100% - Accessibility
100% - Best Practices
100% - SEO
For details, see the attached report. Even on an emulated mid-range smartphone from 2020, the score stays nearly 100% across the board. Note we make use of HTTP/2 and IPv6. An application's performance is regularly assessed. In case of diminishing performance, measures will be taken to speed up the application. Among possible improvements are code refactoring, database tuning, and adding hardware.
Each application exists in at least three environments: development, quality assurance, and production. Our release management is based on promotion to the next environment, to limit human error. All versions of the source code, the database schemata, the documentation, and the configuration files are immutably stored in a blockchain, in case a rollback is required. All data in the quality assurance environment is copied daily from production.
Application bugs will be fixed free of charge.
Infrastructure
Applications made by Vision 42 need infrastructure (servers, storage, and network) to run on. Vision 42 selects reliable and accomplished infrastructure-as-a-service (IaaS) providers to host the applications. At the time of writing, these providers are TransIP and Google.
Although our IaaS providers guarantee and deliver 99.99% uptime, the availability of a Vision 42 application can be slightly lower. Keeping potential release issues in mind, the effective availability will be at least 99.9%. Internet and customer network issues are out of our control, so the perceived availability at certain user endpoints might be lower. Free or captive-portal based Internet access, like in hotels, are often not stable enough and can cause sub-optimal operation. Intermediate cyber attacks of the distributed denial of service (DDoS) type, have been known to cause multi-hour Internet connection saturation [1]. Our current operational status and detailed uptime statistics can be consulted on a public status page. In case of calamities, news will be published on our website or on Twitter.
The primary data center is state-of-the-art and is located in Amsterdam, The Netherlands. See the attached fact sheet for details. Our IaaS providers have a massive bandwidth available.
Vision 42 applications run on redundant physical servers. All servers have been over-dimensioned, so they have power to spare. Currently, the specifications of a single virtual server are:
CPU model: AMD EPYC (dedicated)
CPU usage: < 15% (typical)
Memory: 8 GB
Disk: Non-Volatile Memory Express (NVMe) - 3 replicas
File system: ZFS
Network: 2.5 Gbps
Although present, swap space is barely used.
Our Domain Name System (DNS) provider is Cloudflare.
[1] The Netherlands has a powerful National Scrubbing Center against DDoS attacks.

Backup and Restore
Although our primary IaaS vendor provides backups, we choose not to place all of our eggs in one basket. That is why we make use of an independent backup-as-a-service (BaaS) provider. At the time of writing, this provider is rsync.net.
Geographically, the secondary data center is located in Denver, USA [2]. In contrast with Amsterdam [1], Denver is located one mile above sea-level. The geographical difference between both locations provide protection against complete data-loss caused by most natural disasters. Additionally, offsite backups are kept at two separate locations in Belgium [3] [4]. Data for optional machine learning resides in Frankfurt [5]. In case a big asteroid destroys The Netherlands, Colorado, and Belgium at once, we fear your data may be lost. At that moment, it will be the least of your problems...
All databases are continuously being replicated offsite. All other files are copied offsite with hourly intervals. Daily, a full backup of the databases is performed, verified, and stored in both data centers and all offsite locations. We keep many snapshots, so we can restore up to 3 years ago. Our recover point objective (RPO) is minutes at most.
Using transaction logs, we can restore to any point in time. In case of accidental deleting of data, we can remove the harmful commands from the log and restore up to the current situation, without any data loss.
Recover time objective (RTO) depends on the service level agreement (SLA) of your support formula and will typically be next business day at most.
We offer a takeout service to leaving clients. A large ZIP-file will be provided, containing your data in easily readable CSV-files.
Monitoring
Our infrastructure and the applications are continuously being monitored. Alerts are sent to the engineering staff, who often solve the infrequent problems before a customer notices. Monitoring includes:
Infrastructure availability
Application availability
Database integrity:
Table or index entries that are out of sequence
Misformatted records
Missing pages
Missing or surplus index entries
UNIQUE, CHECK, and NOT NULL constraint errors
Integrity of the freelist
Sections of the database that are used more than once, or not at all
Foreign key constraints that are violated
Replication of databases
Backup of application files and transaction logs
Backup of configuration files
Snapshots
Available disk and backup capacity
Intrusion or privilege escalation attempts
Intrusion Detection System (IDS)
Unsanctioned open ports
Expiration of certificates
Unauthorized certificates, using certificate transparency logs
Undelivered email
Domain Name System (DNS) configuration
Warnings and errors of web applications
The timely execution of scheduled tasks
Scheduled tasks that are not being monitored
Network performance
External email issues
Even the monitoring itself
Security
All servers run FreeBSD or Linux operating systems. We proactively patch and upgrade all system software. If an intrusion or a privilege escalation attempt is detected, automatic countermeasures will fire immediately.
Instead of passwords, salted cryptographic hashes are stored in the database. Weak passwords are not allowed. Brute-force attacks are crippled by delaying failed authorization attempts.
All external network connections are encrypted, both the client/server as the server/server connections. Certificates are always being verified. Strong certificates are used for our servers. Additionally, we make use of:
Transport Layer Security (TLS)
Certification Authority Authorization (CAA)
Certificate Transparency (CT)
Online Certificate Status Protocol (OCSP) stapling, including Must-Staple
Domain Name System Security Extensions (DNSSEC)
Secure Shell fingerprint records (SSHFP)
Service binding and parameter specification (DNS SVCB and HTTPS RRs)
Intrusion Detection System (IDS)
We hold these security ratings (reports attached):
"A+" final SSL grade on ImmuniWeb
"A+" overall rating on Qualys SSL Labs
"A+" on securityheaders.io
"A+" on moz://a Observatory
"10/10" score on Mail Tester
Our TLS implementation is compliant with:
Payment Card Industry Data Security Standard (PCI DSS) requirements
Health Insurance Portability and Accountability Act (HIPAA) guidance
National Institute of Standards and Technology (NIST) guidelines
Our email is protected from eavesdropping and impersonation by:
Opportunistic TLS (StartTLS)
Sender Policy Framework (SPF)
DomainKeys Identified Mail (DKIM)
Domain-based Message Authentication, Reporting and Conformance (DMARC)
Every write operation can be traced to a specific user, an IP address, and a browser type. Relevant server logging is being kept for 10 weeks:
Web access, error, and TLS
E-mail
SFTP
FTP




