Wednesday, January 7, 2009

Sizing your Geodatabase in a DBMS

One of the most unknown tools available for ESRI users is the System Design Strategies document published twice a year. (See link to the right.) This document is long and a little confusing to understand if you haven’t taken ESRI’s course on how to use this document. It’s a must read for any GIS System Architects out there but I’m always surprised at how many of the best GIS people I know don’t even know about this very useful document.

In this entry I’ll show how to size a Geodatabase. The first thing you need to know is ESRI technology, ArcGIS and ArcObjects, predominantly use integer based calculations (See page 9-6 of the PDF) . This is also the case with ArcSDE and the underlying DMBS like Oracle or SQL Server. The next thing you need to know is that the Standard Performance Evaluation Corporation tests a wide variety of hardware platforms for some 45 vendors based on integer calculations. ESRI has based its software off of this standard and the System Design Strategies document can be used to size your system for most of ESRI’s software.

The most important chapter is this document is chapter 9. It contains a series of graphs and detailed explanations for sizing your system. For example, if you wanted to handle little over 500 peak users that use the Geodatabase Direct Connect option then you would turn to page 9-17 or Figure 9-17 and you will see that a Xeon X5260 4 (2 chips) 3.33GHz configuration will support that many users. If you look on the left side of the graph you’ll see the SPECrate_int2006 for that particular CPU and then go to the SPEC site at this link and you’ll find that its base rate is 70.1. There are several computers that have that particular rate. Search on that page for 70.1 and you should find 4 computers that have 70.1 under the Base column: NovaScale T820, NovaScale T840, ProLiant Bl480C, ProLiant DL360 G5. Those are your purchase options.



Of course, it usually isn’t that simple. For most organizations you have a select list of vendors you can choose from. If that’s the case then you’ll have to make some compromise, have the machine configured differently, live with the limitation or oversize it. But, the good news is that most of the time you will find something that matches your requirements. Usually it’s harder to determine how many peak users you’ll be supporting.

One thing you should always do is make sure you’re handling the number of peak users out to some point in time. You don’t want to buy a new server and have it maxed out with the first 2 years. It’s always good to keep metrics about system usage and any good system administrator is doing just that.

One last thing to beware of…There is now a book that describes the very same process I just took you through and it comes with a spreadsheet that makes this process easier but you’ll have to pay ESRI some money. If you do a lot of system sizing and don’t like reading PDFs then this book may be the best option for you.

Happy system sizing.

No comments:

Post a Comment