Web Cache Concepts

Web Cache Concepts

FortiGate web caching is a form of object caching that accelerates web applications and web servers by reducing bandwidth usage, server load, and perceived latency. Web caching supports caching of HTTP 1.0 and HTTP 1.1 web sites. See RFC 2616 for information about web caching for HTTP 1.1.

Web caching supports caching of Flash content over HTTP but does not cache audio and video streams including Flash videos and streaming content that use native streaming protocols such as RTMP.

The first time a file is received by web caching it is cached in the format it is received in, whether it be compressed or uncompressed. When the same file is requested by a client but in a different compression format, the cached file is converted to the new compressed format before being sent to the client.

There are three significant advantages to using web caching to improve HTTP and WAN performance:

  • reduced bandwidth consumption because fewer requests and responses go over the WAN or Internet. l reduced web server load because there are fewer requests for web servers to handle.
  • reduced latency because responses for cached requests are available from a local FortiGate unit instead of from across the WAN or Internet.

You can use web caching to cache any web traffic that passes through the FortiGate unit, including web pages from web servers on a LAN, WAN or on the Internet. You apply web caching by enabling the web caching option in any security policy. When enabled in a security policy, web caching is applied to all HTTP sessions accepted by the security policy. If the security policy is an explicit web proxy security policy, the FortiGate unit caches explicit web proxy sessions.

Turning on web caching for HTTP and HTTPS traffic

Web caching can be applied to any HTTP or HTTPS traffic by enabling web caching in a security policy that accepts the traffic. This includes IPv4, IPv6, WAN optimization and explicit web proxy traffic. Web caching caches all HTTP traffic accepted by a policy on TCP port 80.

You can add web caching to a policy to:

  • Cache Internet HTTP traffic for users on an internal network to reduce Internet bandwidth use. Do this by selecting the web cache option for security policies that allow users on the internal network to browse web sites on the

Internet.

  • Reduce the load on a public facing web server by caching objects on the FortiGate unit. This is a reverse proxy with web caching configuration. Do this by selecting the web cache option for a security policy that allows users on the Internet to connect to the web server.
  • Cache outgoing explicit web proxy traffic when the explicit proxy is used to proxy users in an internal network who are connecting to the web servers on the Internet. Do this by selecting the web cache option for explicit web proxy security policies that allow users on the internal network to browse web sites on the Internet.
  • Combine web caching with WAN optimization. You can enable web caching in any WAN optimization security policy. This includes manual, active, and passive WAN optimization policies and WAN optimization tunnel policies.

Turning on web caching for HTTPS traffic

You can enable web caching on both the client-side and the server-side FortiGate units or on just one or the other. For optimum performance you can enable web caching on both the client-side and server-side FortiGate units. In this way only uncached content is transmitted through the WAN optimization tunnel. All cached content is access locally by clients from the client side FortiGate unit.

One important use for web caching is to cache software updates (for example, Windows Updates or iOS updates. When updates occur a large number of users may all be trying to download these updates at the same time. Caching these updates will be a major performance improvement and also have a potentially large impact on reducing Internet bandwidth use. You may want to adjust the maximum cache object size to make sure these updates are cached. See Turning on web caching for HTTP and HTTPS traffic on page 270.

Turning on web caching for HTTPS traffic

Web caching can also cache the content of HTTPS traffic on TCP port 443. With HTTPS web caching, the FortiGate unit receives the HTTPS traffic on behalf of the client, opens up the encrypted traffic and extracts content to be cached. Then FortiGate unit re-encrypts the traffic and sends it on to its intended recipient. It is very similar to a man-in-the-middle attack.

You enable HTTPS web caching from the CLI in a security policy or an explicit proxy policy that accepts the traffic to be cached using webcache-https. For a firewall policy:

config firewall policy edit 0 .

. . set webcache enable set webcache-https any .

.

.

end

For an explicit web proxy policy:

config firewall proxy-policy edit 0 set proxy explicit-web .

. . set webcache enable set webcache-https any .

.

. end

Turning on web caching for HTTPS traffic

The any setting causes the FortiGate unit to re-encrypt the traffic with the FortiGate unit’s certificate rather than the original certificate. This configuration can cause errors for HTTPS clients because the name on the certificate does not match the name on the web site.

You can stop these errors from happening by configuring HTTPS web caching to use the web server’s certificate by setting webcache-https to ssl-server. This option is available for both firewall policies and explicit web proxy policies.

config firewall policy edit 0 .

. . set webcache enable set webcache-https ssl-server .

.

.

end

The ssl-server option causes the FortiGate unit to re-encrypt the traffic with a certificate that you imported into the FortiGate unit. You can add certificates using the following command:

config firewall ssl-server edit corporate-server set ip <Web-Server-IP> set port 443 set ssl-mode { full | half} set ssl-cert <Web-Server-Cert>

end Where:

Web-Server-IP is the web server’s IP address.

Web-Server-Cert is a web server certificate imported into the FortiGate unit.

The SSL server configuration also determines whether the SSL server is operating in half or full mode and the port used for the HTTPS traffic.

You can add multiple SSL server certificates in this way. When web caching processing an SSL stream if it can find a certificate that matches the web server IP address and port of one of the added SSL servers; that certificate is used to encrypt the SSL traffic before sending it to the client. As a result the client does not generate SSL certificate errors.

Web caching uses the FortiGate unit’s FortiASIC to accelerate SSL decryption/encryption performance.

Full mode SSL server configuration

The ssl-mode option determines whether the SSL server operates in half or full mode. In full mode the FortiGate unit performs both decryption and encryption of the HTTPS traffic. The full mode sequence is shown below.

Turning on web caching for HTTPS traffic

Full mode SSL server configuration

In full mode the FortiGate unit is acting as a man in the middle, decrypting and encrypting the traffic. So both the client and the web server see encrypted packets.

Usually the port of the encrypted HTTPS traffic is always 443. However, in the SSL server configuration you can set the port used for HTTPS traffic. This port is not altered by the SSL Server. So for example, if the SSL Server receives HTTPS traffic on port 443, the re-encrypted traffic forwarded to the FortiGate unit to the server or client will still use port 443.

Half mode SSL server configuration

In half mode, the FortiGate unit only performs one encryption or decryption action. If HTTP packets are received, the half mode SSL server encrypts them and converts them to HTTPS packets. If HTTPS packets are received, the SSL server decrypts them and converts them to HTTP packets.

Half mode SSL server configuration

In half mode, the FortiGate unit is acting like an SSL accelerator, offloading HTTPS decryption from the web server to the FortiGate unit. Since FortiGate units can accelerate SSL processing, the end result could be improved web site performance.

Usually the port of the encrypted traffic is always 443. However, in the SSL server configuration you can set the port used for HTTPS traffic. No matter what port is used for the HTTPS traffic, the decrypted HTTP traffic uses port 80.

Changing the ports on which to look for HTTP and HTTPS traffic to cache

Changing the ports on which to look for HTTP and HTTPS traffic to cache

By default FortiOS assumes HTTP traffic uses TCP port 80 and HTTPS traffic uses port 443. So web caching caches all HTTP traffic accepted by a policy on TCP port 80 and all HTTPS traffic on TCP port 443. If you want to cache HTTP or HTTPS traffic on other ports, you can enable security profiles for the security policy and configure a proxy options profile to that looks for HTTP and HTTPS traffic on other TCP ports. To configure a proxy options profile go to Network > Explicit Proxy.

Setting the HTTP port to Any in a proxy options profile is not compatible with web caching. If you set the HTTP port to any, web caching only caches HTTP traffic on port 80.

Web caching and HA

You can configure web caching on a FortiGate HA cluster. The recommended best practice HA configuration for web caching is active-passive mode. When the cluster is operating, all web caching sessions are processed by the primary unit only. Even if the cluster is operating in active-active mode, HA does not load-balance web caching sessions.

In a cluster, only the primary unit stores the web cache database. The databases is not synchronized to the subordinate units. So, after a failover, the new primary unit must build its web cache.

Web caching and memory usage

To accelerate and optimize disk access and to provide better throughput and less latency, web caching uses provisioned memory to reduce disk I/O and increase disk I/O efficiency. In addition, web caching requires a small amount of additional memory per session for comprehensive flow control logic and efficient traffic forwarding.

When web caching is enabled you will see a reduction in available memory. The reduction increases when more web caching sessions are being processed. If you are thinking of enabling web caching on an operating FortiGate unit, make sure its memory usage is not maxed out during high traffic periods.

In addition to using the system dashboard to see the current memory usage you can use the get test wad 2 command to see how much memory is currently being used by web caching. See get test {wad | wccpd} <test_ level> on page 1 for more information.

Changing web cache settings

In most cases, the default settings for the WAN optimization web cache are acceptable. However, you may want to change them to improve performance or optimize the cache for your configuration. To change these settings, go to WAN Opt. & Cache > Settings.

From the FortiGate CLI, you can use the config wanopt webcache command to change these WAN optimization web cache settings.

Changing web cache settings

Always revalidate

Select to always revalidate requested cached objects with content on the server before serving them to the client.

Max cache object size

Set the maximum size of objects (files) that are cached. The default size is 512000 KB and the range is 1 to 4294967 KB. This setting determines the maximum object size to store in the web cache. Objects that are larger than this size are still delivered to the client but are not stored in the FortiGate web cache.

For most web traffic the default maximum cache object size is recommended. However, since web caching can also cache larger objects such as Windows updates, Mac OS updates, iOS updates or other updates delivered using HTTP you might want to increase the object size to make sure these updates are cached. Caching these updates can save a lot of Internet bandwidth and improve performance when major updates are released by these vendors.

Negative response duration

Set how long in minutes that the FortiGate unit caches error responses from web servers. If error responses are cached, then subsequent requests to the web cache from users will receive the error responses regardless of the actual object status.

The default is 0, meaning error responses are not cached. The content server might send a client error code (4xx HTTP response) or a server error code (5xx HTTP response) as a response to some requests. If the web cache is configured to cache these negative responses, it returns that response in subsequent requests for that page or image for the specified number of minutes.

Fresh factor

Set the fresh factor as a percentage. The default is 100, and the range is 1 to 100%. For cached objects that do not have an expiry time, the web cache periodically checks the server to see if the objects have expired. The higher the Fresh Factor the less often the checks occur.

For example, if you set the Max TTL value and Default TTL to 7200 minutes (5 days) and set the Fresh Factor to 20, the web cache check the cached objects 5 times before they expire, but if you set the Fresh Factor to 100, the web cache will check once.

Max TTL

The maximum amount of time (Time to Live) an object can stay in the web cache without the cache checking to see if it has expired on the server. The default is 7200 minutes (120 hours or 5 days) and the range is 1 to 5256000 minutes (5256000 minutes in a year).

Changing web cache settings

Min TTL

The minimum amount of time an object can stay in the web cache before the web cache checks to see if it has expired on the server. The default is 5 minutes and the range is 1 to 5256000 minutes (5256000 minutes in a year).

Default TTL

The default expiry time for objects that do not have an expiry time set by the web server. The default expiry time is 1440 minutes (24 hours) and the range is 1 to 5256000 minutes (5256000 minutes in a year).

Proxy FQDN

The fully qualified domain name (FQDN) for the proxy server. This is the domain name to enter into browsers to access the proxy server. This field is for information only can be changed from the explicit web proxy configuration.

Max HTTP request length

The maximum length of an HTTP request that can be cached. Larger requests will be rejected. This field is for information only can be changed from the explicit web proxy configuration.

Max HTTP message length

The maximum length of an HTTP message that can be cached. Larger messages will be rejected. This field is for information only can be changed from the explicit web proxy configuration.

Ignore

Select the following options to ignore some web caching features.

If-modified-since By default, if the time specified by the if-modified-since (IMS) header in the client’s conditional request is greater than the last modified time of the object in the cache, it is a strong indication that the copy in the cache is stale. If so, HTTP does a conditional GET to the Overlay Caching Scheme (OCS), based on the last modified time of the cached object. Enable ignoring if-modified-since to override this behavior.
HTTP 1.1

conditionals

HTTP 1.1 provides additional controls to the client over the behavior of caches toward stale objects. Depending on various cache-control headers, the FortiGate unit can be forced to consult the OCS before serving the object from the cache. For more information about the behavior of cache-control header values, see RFC 2616.Enable ignoring HTTP 1.1 Conditionals to override this behavior.

Changing web cache settings

Pragma-no-cache Typically, if a client sends an HTTP GET request with a pragma no-cache (PNC) or cache-control no-cache header, a cache must consult the OCS before serving the content. This means that the FortiGate unit always re-fetches the entire object from the OCS, even if the cached copy of the object is fresh. Because of this behavior, PNC requests can degrade performance and increase server-side bandwidth utilization. However, if you enable ignoring Pragma-no-cache, then the PNC header from the client request is ignored. The FortiGate unit treats the request as if the PNC header is not present.
IE Reload Some versions of Internet Explorer issue Accept / header instead of Pragma no-cache header when you select Refresh. When an Accept header has only the / value, the FortiGate unit treats it as a PNC header if it is a type-N object. Enable ignoring IE reload to cause the FortiGate unit to ignore the PNC interpretation of the Accept / header.

Cache Expired Objects

Applies only to type-1 objects. When this option is selected, expired type-1 objects are cached (if all other conditions make the object cacheable).

Revalidated Pragma-no-cache

The pragma-no-cache (PNC) header in a client’s request can affect how efficiently the FortiGate unit uses bandwidth. If you do not want to completely ignore PNC in client requests (which you can do by selecting to ignore Pragma-no-cache, above), you can nonetheless lower the impact on bandwidth usage by selecting Revalidate Pragma-no-cache.

When you select Revalidate Pragma-no-cache, a client’s non-conditional PNC-GET request results in a conditional GET request sent to the OCS if the object is already in the cache. This gives the OCS a chance to return the 304 Not Modified response, which consumes less server-side bandwidth, because the OCS has not been forced to otherwise return full content.

By default, Revalidate Pragma-no-cache is disabled and is not affected by changes in the top-level profile.

Most download managers make byte-range requests with a PNC header. To serve such requests from the cache, you should also configure byte-range support when you configure the Revalidate pragma-no-cache option.

 

Forwarding URLs to forwarding servers and exempting web sites from web caching                Web Cache


Having trouble configuring your Fortinet hardware or have some questions you need answered? Check Out The Fortinet Guru Youtube Channel! Want someone else to deal with it for you? Get some consulting from Fortinet GURU!

This entry was posted in FortiGate, FortiOS 5.6 on by .

About Mike

Michael Pruett, CISSP has a wide range of cyber-security and network engineering expertise. The plethora of vendors that resell hardware but have zero engineering knowledge resulting in the wrong hardware or configuration being deployed is a major pet peeve of Michael's. This site was started in an effort to spread information while providing the option of quality consulting services at a much lower price than Fortinet Professional Services. Owns PacketLlama.Com (Fortinet Hardware Sales) and Office Of The CISO, LLC (Cybersecurity consulting firm).

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.