Squid Web Cache wiki

Squid Web Cache documentation

πŸ”— Feature: Squid Appliance

πŸ”— Details

There area number of features and design ideas that I think, if implemented, would make this a very useful project. This would be a way to get a system up and running with Squid extremely quickly with good performance, but without a lot of technical knowledge. Tweaking manually would offer better performance. People seeking the best possible performance or more advanced configurations (like reverse proxies) should be looking at building their setup manually.

Offering a quick install method like this means that organizations can drop a Squid cache into place without needing to devote much in the way of man hours. This significantly reduces the risk to opportunity cost for organizations. If the attempt doesn’t work, then the organization can remove the cache and only be out a few man hours instead of days.

πŸ”— Disk Layout

Partitioning and formatting should happen automatically on install.

Normal boot/tmp/etc rules apply. However, I want to mention having separate partitions for the OS and storage that would stay persistent across an OS reinstall. Placing the cache, and a backup of configuration files on a second partition would allow a person to format/reinstall the OS + Squid without losing their cache data or configuration data (more on this below).

For the OS partition EXT3 is probably a good choice as it is stable. It may be possible to specify this partition as a small fixed size. Then redirect the logs to wherever the backup of the configuration files is.

A backup of the configuration files is relatively static and should not make a difference on files system choice for the cache ( orpersistent storage) partition. If the cache is setup as it’s own direct access partition, then a single smaller partition should be created to hold the configuration files.

According to this page the best file system to set up the cache is probably ReiserFS or EXT3.

πŸ”— Cache File Options

COSS is apparently the recommended cache system now for small objects. There does not seem to be any information as to if it is better installed on a file system or with direct partition access. Presumably direct partition access would be best.

I don’t see any information on which of the other cache_dir types (aufs, diskd or ufs) are best for storing large files.

The size of the two caches (one for small files and the other for large) would be determined at install time. The installer is presented with a few usage scenarios, and the cache sizes are set automatically from there. This way speed and bandwidth savings are balanced properly for the user.

πŸ”— Upgrading

Squid could be upgraded by itself as new versions come out. However it is probably that users would also want to upgrade the OS at the same time to take advantage of fixes and improvements. To facilitate this, the install CD should have an β€œUpgrade” option. This option would format the OS partition and install on there, then read the stored configuration files off of the persistent storage partition. These would be used to automatically configure Squid using whatever options had been chosen previously.

KnoppMyth makes use of this, but must update the schema of the stored database it uses. I don’t believe there is enough data and options to warrant the need of a full database running. This should make upgrading and restoring configurations very strait forward.

πŸ”— GUI

I can’t see any practical reason to have X installed on the system. Is there anything on a proxy like this that would benefit from X?

Some simple after the fact configuration and system information could be provided by a web interface.

πŸ”— Statistics

Statistics should be provided through some sort of web interface. A quick and/or detailed summary of hit % broken down by day and time. What packages are available that do this?

Possibly some sort of -tail display of the logs, to see the last items requested. Bonus points if they are formatted nicely.

πŸ”— Authentication

Adding authentication can make the complexity of the project an order of magnitude more complex. Possibly an IP list would be easy. If people want this, they are best off with a larger all inclusive project like SmoothWall, IPCop, or simply rolling their own.

πŸ”— Filtering

Adding filtering can make the complexity of the project an order of magnitude more complex. Possibly a domain list would be easy to implement. Users could try installing SquidGuard, DansGuardian, etc, or trying one of the packages mentioned above in Authentication.

πŸ”— Operating System

The BestOsForSquid says almost any Unix-like OS would work. It should be a very common distro with a large user base so that regular updates are likely, and questions regarding obscure OS errors might be answered.

πŸ”— Squid Version

If possible, the latest versions of the 2.x and 3.x branches should be made available as choices during install. Both are experiencing active development and currently have different feature sets that users might need.

What types of things change depending on the version? Are the log files different? Are common settings in the configuration files completely different?

πŸ”— Transparent

Option to set if the proxy is transparent or not. Some people will want an inline bridge, others may not. What would need to change on this option?

πŸ”— Install Options Tree

A prototype of install options should be made, along with what specifically changes for each install option.

πŸ”— Install type

πŸ”— Inline Proxy

πŸ”— Transparent Proxy

πŸ”— Proxy Purpose

πŸ”— IP Information

πŸ”— Squid Version

πŸ”— Other Considerations

Categories: WantedFeature

Navigation: Site Search, Site Pages, Categories, πŸ”Ό go up