Home
 
contact us | search
products & services | download | support | order | partners | about issel

Managing Hit List databases, database growth, and some speed tips.

By default Hit List creates Access97 databases from your logfile data, with a filename extension of .mwd. For those using Hit List Pro or Hit List Commerce the following tips will help you maintain, protect and enhance performance of your Access97 databases.

Enterprise and Live users have the option of using SQL Server or Oracle databases instead of Access97, tips included below.

General Speed and Maintenance Tips, applicable to ALL Hit List databases (regardless of format).

A) Control what logfile data you load into the database.

Make sure you are excluding from the database all "irrelevant data", namely GIFs, JPEGs, JPGs, internal IPs, User-Agent strings (browser types you don't care about), HTTP Status codes you don't care about, etc.under Options/Updates. Keeping such data out of the database altogether will make reports run faster, slow down database growth rates, and record the data you want for the content you want to track the most efficiently. For more information on using the Options/Updates tab including the QuickList Extreme optimizations mentioned in the next paragraph, check our Options or our Filtering tech support documents.

Enterprise and Live users also can choose from several QuickList Extreme optimizations under Options/updates. See Help after hitting the "quicklist extreme" button in the upper right for more details on these features and how they can help you maintain very large databases.

B) Controling Database growth.

1) Update the database daily, optimize frequently. Make sure you are updating/optimizing your database on a regular and frequent basis. A "typical" schedule (depending on your volume, of course) is usually nightly database updates with optimizing the database once or twice weekly. Optimizing the database is done in the Database Manager/Tools tab.

If your webserver doesn't rollover its logs but once a month or once a week, investigate the possibility of switching to daily rollovers. Database updates will take less time, you will be better equipped to back up the database if something goes wrong, and you will be able to put in the most data possible.

2) Use Summary-only data. If you aren't doing high-level filtering against your database, and have no need for reports such as Path Analysis or Marketing Report (which require detail data - you can see this by highlighting the report's icon and checking the gray Status bar at the bottom of the main window) you can use Summary-Only Data (see the Options tech document, Updates tab discussion for more details). This will drastically reduce the size of the database as compared to the logfile, grow the database MUCH slower, and still allow further Optimization as well.

For an extended discussion of the differences between Summary and Detail data, read our Summary vs. Detail tech document.

In some internal tests, we've loaded 800 mb of a customer's logs as Summary-Only data, excluding GIFs, JPEGs, etc. with a resultant 100 mb database. Optimizing the database afterward reduced the same databsae to 60 mb in size. So you can see that Summary-Only data can make a drastic difference.

Many Professional and Commerce customers use a "Summary-only" database for long-term tracking while using a "Summary and Detail" database for short-term Marketing and Path Analysis reporting. Check out the Multiple Databases tech document for instructions on setting this up.

C) Back your database and logs regularly -"MOST IMPORTANT"

As with any other mission critical data, regular backup of essential files is VERY IMPORTANT.

  • Logs. You should back up your webserver logs (so if you need them for additional database work you have them ready) onto a CD-ROM, Jaz drive or similar on a frequent basis.

  • Databases. You should back up your Hit List logfile database DAILY. PROTECT YOUR DATA! We at Marketwave Tech Support talk to customers on a frequent basis that don't back up their databases, and then have to rebuild them in the event of a hard drive failure, computer crash, etc. If you want to preserve the data and don't want to rebuild your database from what might be many megabytes of logs, back it up!

    If using Professional or Commerce, this file is usually called Hitlist.mwd. If using Enterprise or Live, this file would likely be the same if using Access97 databases. If using SQL Server or Oracle databases with Enterprise or Live, use whatever scheme you have within SQL Server or Oracle to backup and protect the databases accordingly. If none, consult your SQL Server or Oracle dba, who may already be backing things up.

  • Reports database. You should back up your reports database (filename = hlrpt40x.mdb, the "x" representing which version of Hit List you have, p for Pro, c for Commerce, e for Enterprise and l for Live). on a regular basis to preserve custom reports you make, more often if you build new reports fairly often or have many reports already built.
  • D) Reporting speed tips

  • Remove elements you don't care about.
    With Hit List reporting, the easiest way to speed up reports is to remove/disable elements you don't care about from the reports. Thus if you're using Complete Analysis (or a copy of same) and you don't care about "most popular browsers", simply remove this element after highlighting the report, clicking Design and scrolling down to this element on the Outline tab. Once you delete this element (or disable it from within its Properties) there is one less calculation for Hit List to perform when running this report, and it will complete quicker.

  • Scale back certain elements.
    Many Hit List elements such as "most popular pages" allow you to set the number of records returned to "Top 50 Items" or Top 100 percent" - if this returns more records than you care about (ie you only want the Top 10 Items or Top 10 "most popular pages") then scale this element back by adjusting the "limit results to" setting in the element at issue (highlight the report you're using, click "Design", then go to the element's Properties/Definition tab and make the change).

  • Make your Filters more efficient.
    When you impose Filters on a Hit List report, either at the Report or Element level (see our Filtering tech document for more details) you can potentially slow this report down, especially if you put in multiple filters at both the report and element levels. So, after deciding what results you need or want to see from a given report, take the time to plan your filtering to get the results you want more efficiently. Remember that the "URLs within the Visit" filter can take a long time to run in some cases.

  • Reduce the Time Period for the report when possible.
    Remember that the more time a report covers, the more data it must report against and thus you can speed up reports by setting "time period" to "yesterday" or "last week", etc. instead of "since inception" in many cases. Certainly, there will be times when you need a complete year-to-date report or something similar, but in many cases you do not, and thus can reduce the Time Period against which the report runs, and speed up reporting as a result.

  • Managing Access97 databases specifically (always applies to Professional and Commerce users, Enterprise and Live as necessary).

    1) If the database gets over 1 GB in size, it's not clear what Access97 (or Hit List) will do. Access97 databases have an effective size limitation of 1 GB. Sometimes everything will work, but in MOST situations it will "konk out" and despite attempts at repair in Access97 or Hit List, etc. you won't be able to use the database and will have to rebuild.

    So keep a close watch on the database growth rate, and when the database has grown to 900-950 mb, begin a new one or "cycle" as discussed below. This size limit is why the importance of a backup CANNOT BE OVERSTATED. In the simplest scheme, you could simply zip up the existing logfile database (usually hitlist.mwd) on a daily basis and keep replacing the zip daily.

    2) When you get up around 850-900 mb in database size, it's usually a good idea to "cycle" the database in the Database Manager/Tools. This retains your IP lookup data, the information about which logs you've already loaded, and creates a new database to import more data into while renaming the old. It also allows you the flexibility of retaining summary data (provided you're using Summary and Detail) to allow certain types of longer-term reporting while still doing short-term reports. See Help in the Database Manager/Tools tab for more information.

    SQL Server databases

    Enterprise and Live users have the option of using SQL Server as a back-end database instead of Access97. For these users, the following may help them keep their databases running smoothly.

    1) Set up SQL Server "correctly".

    In many cases, users don't consult Hit List Help under the "Using Oracle and SQL Server" Help topic or our FAQ document on setting up their SQL Server and then have problems (common SQL Server/Oracle problems are outlined in the Database in Use by another user document).
    So if you're setting up SQL Server (7.0 or 6.5) for use with Hit List for the first time or are having problems, read this Help topic in addition to the document noted above.

    2) Configuration of the machine.

    Here's a few configuration options you may adjust as your requirements vary, your website grows or you upgrade your SQL Server/Hit List machine:

    Memory - in both SQL Server 6.5 and 7.0, you can adjust how much RAM SQL Server gets to exclusively use. The usual recommendation in most setups is slightly less than half your RAM should be allocated to SQL Server (i.e. 225 mb RAM to SServer on a 512 mb RAM machine). Updating will go quicker if you give more memory to SQL Server, but then reporting might be slower. And the converse is true, that if you take away RAM from SQL Server, thus leaving it for Hit List and NT, then reporting will go quicker but updating slower. So, a bit of experimentation might be in order if you have specific needs - for example, if you have plenty of time available for updating (overnight) but you don't for reporting, you may want to scale back the RAM SQL Server gets to allow reporting to run quicker.

    NOTE: There probably will be a point where either Hit List or SQL Server will end up with too little RAM to be effective (i.e. don't give SQL Server 12 mb RAM and Hit List/NT 500 mb RAM) so these adjustments should not be too extreme.

    Hard Drives and RAM - SQL Server (6.5 or 7.0, especially 7.0) will ultimately run fastest on either very fast SCSI drives or a RAID configuration, and if you are processing several hundred mb of logs nightly, you should consider such a setup. In a similar vein, the more RAM, the better. There have been a few cases of people trying to run several hundred mb of logs into a 64 mb RAM, IDE SQL Server setup - this will not suffice!


    3) Evaluate the "general" database tips at the top of this document and USE them.

    Many customers think that no adjustment needs to be done or no regular "maintenance" of the database files is important - THIS IS NOT TRUE. These maintenance items were added to the product to enhance use and provide flexibility whenever possible - they are designed to help you!

    Oracle Server databases

    As of this writing, the only recommendations beyond the above suggestions (that apply to either all databases or SQL Server databases) would be to make sure that your network connectivity is stable and constant (can sometimes be a problem for Oracle Unix users) and that your Oracle dba has allocated sufficient table space for the log(s) you want to load into the Oracle database. Make sure that person or persons know that you want to process a lot of data and thus usually will require a lot of space in the Oracle setup as a result.

    To examine possible connectivity issues with Oracle and Hit List, see our Database in Use by another user document.


    p +44-(0)870-166-2435, f +44-(0)870-054-8795, e info@issel.co.uk
    © 1996-2004 Intranet Software Solutions (Europe) Limited. All rights reserved.