byrmol Blog

Garbage

First post

Hello all and big thankyou to Graz et al for this handy site.

Although probably a bit off topic for a SQL log, but I'll be focusing on .NET as I am in the middle (read Deep End) of a large implementation of it.

The real goal of the project is to kill off the mainframes. Slowly but surely more and more SQL Server boxes are being implemented to ease the burden and reliance of our mainframes. I didn't make this decision but I will be following orders.

We currently stand at 16 physically servers separated into 4 clusters (4 boxes per cluster). We use the term "street" to describe a collection of clusters.

The Middle tier is 4 times the size the off the Data Tier and Web Tier is the same size as the Data Tier.

Probably the first thing that jumps out at you is why the middle tier is big.

The answer is two fold. First the entire DAL and BAL are in this tier for both mainframe and SQL Server access. And secondly and most damaging is XML.

Besides Middle Tier to Data Tier communications, every other tier relies on Web Services. The XML generated is not shy on size and each server requires a lot of CPU for the parsing of this information both in an out.

So after a 40 years of computer technology, the database is now one of our fastest tiers……

Legacy Comments


robvolk
2003-10-02
re: First post
Well, now I'm wondering how many mainframes you had to replace with this brand spanking new tech? :)

DavidM
2003-10-02
re: First post
It sounds impressive but here is the breakdown for the database "street"

1 Cluster is for redunancy..
2 Cluster are in use and running at about 10%.

The other cluster is for the application I am working on. We will have a minimum of 2000 clients.

At present the Mainframe is running at 80% and has hit its maximum MIPS limit. At the moment the work is split about 60-40 in favour of DB2 on the mainframe but we are slowly pegging it back.

The SQL Server "street" has been way over specced and they hardly know they are alive.

The workload on the SQL Server boxes ranges from simply OLTP to heavy Full text searches and OLAP. In affect they have decided to forgo a datawarehouse and just use big iron.

It works amazingly well and as I initally said the database tier is our fastest tier!!!