Admin: HTTP Proxy Cache

From Resin 3.0

(Difference between revisions)
Jump to: navigation, search
 
(One intermediate revision by one user not shown)
Line 27: Line 27:
  
 
[[Image:Cached.png]]
 
[[Image:Cached.png]]
 +
 +
In this artifical example, the hello_file.html page is a static page that's almost always served out of the cache. Only twice out of 22,000 hits did the cache miss and require an actual file access.
  
 
== Uncached Pages ==
 
== Uncached Pages ==
Line 33: Line 35:
  
 
[[Image:Uncached.png]]
 
[[Image:Uncached.png]]
 +
 +
This example shows a top uncached page that might be a good candidate for improved caching, hello_java.jsp. With 2,500 requests, it's by far the most used uncached page. So concentrating on that one page can improve the performance easily.

Latest revision as of 22:51, 14 October 2010


When you want to increase performance, one of the easiest and most effective techniques is to use Resin's proxy cache and improve the ratio of pages using that cache. A page request that avoids any database access or framework processing is a pure win for performance and latency. The proxy cache can be particularly effective in combination with load balancing because Resin's proxy cache can save load on the web-tier server itself.

Contents

Summary

You can use the summary page to get an overview of the cache effectiveness by looking at the miss ratio of the proxy cache and block cache. The lower the miss ratio, the fewer requests that take up database and framework load. Each "miss" is a request that goes through the full processing of a page.

proxy cache miss ratio
The percentage of requests that miss the caching, going on to full processing.
block cache miss ratio
The underlying disk reads and writes ("block reads/writes") paging from memory caching to disk caching.
invocation cache miss ratio
Counts the invocation misses, which is when Resin must calculate rewrite, servlet and filter dispatching for splitting apart the URL

Summary Example

In the example below, the miss ratio is 11.3%, meaning 11% of all cacheable requests to the server are missing the cache. 22,000 requests use the cache and 2,800 requests miss the cache.

Cache-overview.png

Although the miss ratio is important, it's even more important to know which pages are missing the cache. A slow page like a database access that misses the cache is more important than a fast page like a static page that misses the cache. The following two tabs give more details of which pages are getting cached.

Cached pages

For further tuning, the cached pages tab shows the top cached pages as well as their miss ratio. You can use the cached page to check how effective the caching is for individual pages.

Cached.png

In this artifical example, the hello_file.html page is a static page that's almost always served out of the cache. Only twice out of 22,000 hits did the cache miss and require an actual file access.

Uncached Pages

The uncached pages tab gives an overview of the top uncached pages, which are page you might look at carefully to see if their caching can be enabled. For example, a home page that varies only every 5 minutes can have dramatically improved performance even with a simple caching added.

Uncached.png

This example shows a top uncached page that might be a good candidate for improved caching, hello_java.jsp. With 2,500 requests, it's by far the most used uncached page. So concentrating on that one page can improve the performance easily.

Personal tools