Impact of Too Many Tables?

[SageDog]

Hi,

I have a MySQL database that's been running well for a few years but is undergoing huge growth right now.

In this DB, I have one big table of clients (name, contact info, etc). When a client signs up, a table is created specifically for their oder history, and occasionally a second table may be created for them to track other info. These tables are both very small (20-40 rows, 100 chars each row).

What this means is that I currently have almost 3000 tables in the database, with the possibility of 6000 new tables added each year... Am I going to run into trouble when the database grows toward 10,000+ tables? Is there a limit of tables and can it be expanded? And, should I expect any issues around how fast MySQL can access the data?

Any info would be appriecated. Thanks!

[lavazza]

should I expect any issues around how fast MySQL can access the data?

If, by tables, you mean records (a.k.a. tuples):
No :)

However, if you really do mean tables, then I suggest you redesign your database

[ogletree]

As long as you mean records and not tables that is not a problem. When you do start to get big a lot of companies use sharding.

[SageDog]

I do mean tables. A new small table is created for each new client, each table only having about 20-40 rows...

What kind of issues can I encounter with thousands of tables? I can't find any documentation that outlines risks and issues.

Thanks!

[lavazza]

I do mean tables

Without seeing a model of your dB, I can only guess... and my guess is that your design will prevent you from benefiting from the Relational aspects of a Relational Database...

Before commenting further, I'd be keen to see the syntax of one (or more) of your queries (perhaps a query that returns data from more than one client)

[blaketar]

Question: Why do you create a new "Table" for each client? Why not create 1 table (table:clients) then insert customer data into this table, row-by-row.

If you need to breakup your client data (like order history) create a table (table:orderhistory) then store that history row-by-row with an id that points back to the clients table?

EXAMPLE:
TABLE:clients (id¦clientname¦address¦phone¦email)
1¦Barrys Home Repair¦433 Main St.¦555-9887¦[e-mail]
2¦Tops Painting¦PO Box 93¦555-3320¦[e-mail]
3¦Fitz Plumbing¦210 C. Street¦987-0755¦[e-mail]

TABLE:orderhistory (id¦clientid¦partid¦invoiceid)
1¦2¦834¦9
2¦2¦2211¦9
3¦2¦0093¦9
4¦2¦832¦9
5¦1¦733¦10

This way you can use an easy query to see that Tops Painting ordered 4 products (you can use a JOIN to get the part name using the PARTID) and they are on INVOICEID 9 (again use a JOIN to get INVOICE info).

In this model you can store millions upon millions of rows.