SRE-Http and SSI-Caching
To improve performance, SRE-Http will "cache" HTML documents that
have server side includes (SSI). This caching is of the document
after the SSIs have been performed.
In simple cases SRE-Http can send this cached file, without having to repeat the actual
process of SSI lookups, etc. Needless to say, this can greatly
improve server performance.
Note: the bottom of this document contains a few SSI-Cache control options.
Basic Description
First, let us define some terms:
- Static SSIs
- Static SSIs are those that do not change over time and client.
over time. For example, the <!-- REPLACE WEBMASTER -->
should rarely change (it writes the contents of the WEBMASTER
SRE-Http parameter). Another example is <!-- INCLUDE filename
-->,
given that the contents of filename does not change.
- Dynamic SSIs
- Dynamic SSIs are those that will change over time and client.
The simplest example is <!-- REPLACE TIME -->.
Somewhat less
obvious are all the "processing" options, such as:
- <!-- #IF xx -->
Although these three options may yield unchanging values,
it's likely that
(at least occasionally), the values will depend on the client,
time of day, or some other varying factor.
- Simple cases
- Simple cases are HTML documents that only
contain static SSIs.
It's not hard to see that the caching of simple cases
yields immediate benefits. This is especially true as the number and size
of static SSIs increases. For example, if you use HEADERS,
FOOTERS, and <!-- INCLUDE filename --> SSIs to maintain
a standard set of backgrounds and menu bars, the advantages of
caching can be substantial.
In non-simple cases, caching is more complex. SRE-Http's
caching procedure solves this problem by using the following strategy:
- The raw HTML document (the one containing the SSI keyphrases) is
read in.
- Static server side includes are performed.
- Dynamic server side includes are noted, with placeholders
inserted.
- This partially processed file is saved in the cache.
- Dynamic SSI's are resolved, and the resulting document is
sent to the client.
Since SSIs are recursive, this might entail another
round of processing (say, of REPLACE keyphrases in an INCLUDEd file). This
later round is not cached.
When a new request for the raw HTML document is recieved, the
TRIGGERS and DURATION are checked. If either of these are out of date,
the cache entry is not used. Otherwise, the cached file (from step 4)
is read, the dynamic SSIs (from step 3) are resolved, and
the resulting document is sent to the client.
While not as quick as simple cases, the pre-processing of
static SSIs does improve performance.
Configuring the SRE-Http SSI-Cache
Configuration of the SRE-Http SSI-Cache is accomplished by changing
the SSI_CACHE_ON,
SSI_CACHE_SIZE, SSI_CACHE_DURATION,
and SSI_CACHE_STAMP parameters.
In addition, <!-- CACHE argument -->
server side include
keyphrases can be used to set file-specific caching options.
The SSI_CACHE parameters
The SSI_CACHE_ON parameter is used to enable/disable the
SSI-Cache. It is set in the INITFILT.80 file.
The SSI_CACHE_SIZE, SSI_CACHE_DURATION, and
SSI_CACHE_STAMP
parameters, which are set in the INIT_STA.80 file, can be customized
by ambitious administrators (others may wish to leave them be).
Modifications to these parameters
will not take effect until you restart GoServe.
- SSI_CACHE_ON
- SSI_CACHE_ON is used to enable the SSI-Cache.
You can use the SRE-Http configurator to change its value.
Turning off caching is sometimes useful (or necessary) in highly
dynamic environments; say, where INCLUDE files are constantly changing.
- SSI_CACHE_SIZE
- SSI_CACHE_SIZE is used to set the size of
the server side include cache.
You should enter the size in Kilobytes. For example, the default value of
SSI_CACHE_SIZE=5000 corresponds to a cache size of 5M.
- SSI_CACHE_DURATION
-
SSI_CACHE_DURATION is used to set (in days) the "lifespan"
of entries in the server side include cache.
Example: SSI_CACHE_DURATION=0.5 corresponds to a 12 hr. lifespan.
Note that you can use the <!-- CACHE DURATION time -->
SSI keyphrase to override this
default value on a file-by-file basis.
- SSI_CACHE_STAMP
-
The SSI_CACHE_STAMP dictates which fields to use when determining
whether a SSI-cache trigger file has changed.
SSI_CACHE_STAMP can take one of the following values:
- ALL : Use time,date, and size (if any change, the trigger is considered
to be out-of-sync)
- DATE : Just use the date
- TIME : Just use the time (hr/minute)
- DATETIME : Use the time and date
- SIZE : Just use the size
Example: SSI_CACHE_STAMP='DATETIME'
Note that you can use the <!-- CACHE TRIGGER_type stamp -->
SSI keyphrase, where stamp is one of the above values, to override this
default value on a file-by-file basis.
For further discussion of these parameters, please see INITFILT.DOC.
The CACHE keyword
To set ssi-caching options on a document specfici basis, you can use
the <!-- CACHE option --> SSI keyphrase. The following lists
the currently valid options:
- <!-- cache no --> or <!-- cache off -->
- Do not cache this document
- <!-- cache asis -->
- Use a cached version "as is" -- the first-round of dynamic SSIs will be
resolved prior to caching (they will not be updated
for future requests). However, SELECT and #IF keyphrases,
and "later rounds" (of keyphrases generated by a dynamic include) will
not be converted.
Restating-- when using
<!-- cache asis --> :
* INCLUDE files, the
REPSTRGS_FILE, and INITFILT.80 will not be included in the
trigger list.
* Conditional includes (SELECT and #IF) will
not work
<!-- cache asis --> should be used with care!
- <!-- cache notrigger -->
- Do not check the trigger files; but do check the owm
file. In particular, the various INCLUDEd files, and the SRE-Http
parameter files, will not be checked (nor will any explicitly mentioned
trigger files be checked). One can think of
notrigger
as
a subset of asis.
- <!-- cache duration nnn --> or <!-- cache days nnn -->
- Set the lifespan for this file's cache entry.
After nnn (possibly fractional) days,
the cached entry will be removed. Note that
if this is not specified, the SSI_CACHE_DURATION
parameter is used.
- <!-- cache hours hhh -->
- Same as <!-- cache duration-->, except hhh is
interpreted as hours (not as days)
- <!-- cache trigger file1 file2 .. -->
- Check the file-stamps for these fully
qualified files.
If any of them have changed (since the time the cached file was
created), then do not use the cached file (the
entry will be recreated).
Notes:
- The own file name is always implicitly included in the
trigger list (regardless of asis or duration settings).
- Depending on which static SSIs were used, and assuming
that a CACHE ASIS keyphrase is not used:
- <!-- cache trigger_type stamp -->
- Selects which field(s) of the trigger file-stamps
to check.
stamp can take one of the following values:
DATE TIME DATETIME SIZE ALL. The above description
of the SSI_CACHE_STAMP describes these values.
Note that you can only use one of these five stamps (combinations are
ignored). Furthermore, by default the SSI_CACHE_STAMP
parameter is used.
Hints, Cautions, and other miscellaneous Notes
- Since the SSI-cache will cache all files that might
have server side includes, we recommend use of the SSI_SHTML_ONLY
option. In particular, with SSI_SHTML_ONLY in place
(and a proper naming convention used), one can take advantage
of GoServe's own cache for non-SSI HTML documents.
- If a document only has static SSIs
(that is, no TIME, HITS, INTERPRET, or any other SSIs that
might depend on time, client identity,
or other varying factors); then it is treated as an asis document.
- Non-asis documents are partially cached -- the cache file is
first processed for dynamic SSIs, and then sent to the client.
This yields a more "accurate" document, but it's not as fast as
returning asis entries.
- If you have more then one set of DELIMiters, or you
have turned the DO_SEND_PIECE parameter on, then SSI-caching
will not occur. Thus, we recommend that you do not
use multiple DELIMiters, nor should you use the send pieces
as they become available option.
- Cached files are stored in your TEMPDATA_DIR directory (typically,
GOSERVE\TEMP). They have names of _CS?????.80 (or .nnn, if you are
using port .nnn).
- Once the size of the cached files exceeds SSI_CACHE_SIZE
kilobytes, the oldest
entries will be deleted (where oldest is measured in terms of
"last request").
- If you use the
asis
or notrigger
CACHE options,
the quickest way to reset the cache entry
is by touching the document's file (say, by modifying its creation time).
- Cache entries are host-specific -- the same file delivered
by different hosts will have seperate cache entries.
Some SSI-Cache Controls
The following links (and forms) can be used to view and control the
SSI-Cache. However, you must have SUPERUSER privileges
to use these controls.
A very brief status report will be returned to your browser.
Hint: To suppress use of the SSI-Cache, you can use SRE-Http's
!FORCE/a_url special directive.