ICM Server Cache 
The ICM server cache or Internet server cache saves HTTP(S) objects before they are sent to the client. The next time an object is requested, the application gets the content directly from the cache before sending it to the client.
The HTTP request handler uses the ICM server cache when, for example, response pages need to be reused, such as the entry page of an online shop application. The ICM server cache saves the pages before they are sent to the client. When the page is next called, the application gets the page directly from the ICM and sends it to the client, provided that the expiry period has not run out. In this case the work process does not have to be opened.
Note
Objects in the MIME repository are saved to the cache by default. Method calls are used to activate the ICM server cache. This is described later.
Implementation Considerations
Using the ICM server cache can increase performance considerably.
Benchmarking tests for cache hits in the main memory have resulted in latent response times of less than one millisecond per request, and a total run of under 3,000 requests per second on a 4 CPU hardware.
These results are based on a strong parallel and multi-threaded architecture that supports simultaneous read and write accesses with versioning. Furthermore, a patented indexing algorithm is used to access the cache directory quickly, for example for long Web URLs as cache keys.
Integration
The ICM server cache is part of the Internet Communication Manager (ICM).
Features
The following graphic displays the architecture of the ICM server cache.
ICM Server Cache: Structure
The ICM server cache contains the following:
Two-level cache hierarchy: Memory cache and disk cache
Dynamic Caching Technology
Active Caching
UFO caching (unfound objects)
Browser-dependent caching
Two-Level Cache Hierarchy
The ICM server cache consists of a two-level hierarchy: A very fast main memory repository (memory cache) built on a disk-drive-based disk cache. This has the advantage that both the speed of the main memory access and the large storage capacity of the hard disk repository can be used.
Dynamic Caching Technology
Conventional Web-caching scenarios are based on HTTP proxies and usually only support caching of static contents, such as GIF pictures. The Internet server cache, on the other hand, can store dynamic Web content as well as static content, such as JSP and BSPs. This is especially important as Web applications are becoming increasingly dynamic.
Active Caching
An additional difference between the existing standard Web caching solutions and the Internet server cache consists of the cache control technology, which is integrated in the Internet Communication Framework.
Active caching means that the application has full control over how current the objects are that are in the cache. This happens by using asynchronous content invalidation, which is triggered by application-dependent events. Even this is contrary to the standard HTTP caching, with which the application only has limited control over how current the objects in the cache actually are. With the standard HTTP caching technology, expiry time heuristics are usually used.
UFO Caching (Unfound Objects)
The Internet server cache supports caching invalid requests. These are requests that lead to errors on the application server or on the database back end, such as “not found” errors. As a result, the back end is protected from overload situations in which the system is flooded with invalid Web requests.
Caution
Note that this function is only used as general protection and cannot protect the system from more polished attacks.
GZIP Compression
If the client supports GZIP compression (accept-encoding in the header field) and the server also uses GZIP to compress the response (content-encoding in the header field), then the response is saved to the cache in compressed form. GZ=1 is set in the cache key (see below).
Caution
Note that GZIP compression is only supported for HTTP1.1. If HTTP1.0 is used as a protocol, then the response is not compressed.
Browser-dependent caching
An HTML page may appear differently in different browsers, since the HTML code is not always interpreted in the same way.
This is why the ICM server cache can store the pages according to the browser type. If the same page is requested by a different browser, the request cannot be retrieved from the cache. To inform the runtime environment of whether the content is browser-dependent, you can set the relevant flag in the page's properties. By default the flag is not set.
More Information
How Does the ICM Server Cache Work?
The following documentation describes how the cache identifies objects, stores and invalidates them.
Profile Parameters for Configuring the ICM Server Cache
The following parameters are used to parameterize the ICM Server Cache:
icm/HTTP/server_cache_<xx>/size_MB
icm/HTTP/server_cache_<xx>/memory_size_MB
icm/HTTP/server_cache_<x>/max_entries
icm/HTTP/server_cache_<xx>/clear
icm/HTTP/server_cache_<xx>/expiration
icm/HTTP/server_cache_<xx>/max_name_len
Monitoring
You can monitor the ICM server cache using the Web administration interface or the ICM monitor (AS ABAP).