MySQL server config for table that has longblob field

I have a server that right now is pretty much just used for MySQL.  The DB has various tables but one of the main tables is one used for storing binary files in a longblob field.  The files can be anywhere from a few hunder KB in size to a few MB.  Another field is "program" and that is a always used in the query to get binary files just for the "program."  For the most common program the fields can be > 45 for a specific version and total 700+ records for that program.  A field "version" is also used to narrow down the results.  The "query" will just end up getting one record if any at all.

This is hosted on a Windows 2003 SBS server if that matters.

The problem is the "download" of the blob seems to be taking some time.  If the server is restarted, the number of records with the same program are lowered, or with "programs" that have fewer records and smaller binaries the download is at a normal speed.  What are configurations in the INI that would make the most difference in this type of DB and table?  What are status or server variables (in MySQL Admin) that will be especially useful in narrowing down problems that would cause this?  There are other uses for the server (e.g. web server) and DB (e.g. other tables with records that aren't blob fields) so let me know if those could be negatively impacted or factor into any of this.

I am familiar with MySQL but just not an expert in its set up, especially in a situation like this.  Let me know if you need any other info.  Thanks!

bol
LVL 54
b0lsc0ttIT ManagerAsked:
Who is Participating?
 
rydersaintConnect With a Mentor Commented:
hello b0lsc0tt

here is some ideas

longblob is a very slow data type
if you could get an native to ascii converter script or code on your files then it convert them on the fly to and from you DB from text to binary (like similar to what most mail server do).
Then you would be able to use the text datatype instead

If its only a max of a few MB then you could already switch it too mediumblob (this is faster too)

however if you are really stuck with longblob I sugest put in a really good raid5 or raid10
with lot of memory. ZFS is also a very good performance enhancer

Hope this helps
Ryder
0
 
b0lsc0ttIT ManagerAuthor Commented:
Ryder,
Thanks for the response.  I have made that change to that field (longblob to mediumblob).  Thanks for the input and recommendation on it.
Any other ideas or suggestions?  Let me know it some other details are needed.
bol
0
 
rydersaintCommented:
ok so mediumblob then

I would recommend having two tables

one like

TABLE1
keyID (INT) --primary key | filelink (INT) | file (mediumblob) --your medium blob

and then another table in memory

CREATE TABLE TABLE2 (blah blah blah) ENGINE = MEMORY;

TABLE2
keyID (INT) --primary key | filelink (INT)| fileName (VARCHAR) -- name of your file | and so on

then select with a join like

SELECT * FROM TABLE1
LEFT JOIN TABLE2 ON TABLE1.keyID=TABLE2.keyID
WHERE fileName = 'the requested filename';

(* Note - you will need another table for table2 to copy from as all data leaves the memory table when restarting the server. Basically one for storage and one for selecting from)
0
Cloud Class® Course: CompTIA Cloud+

The CompTIA Cloud+ Basic training course will teach you about cloud concepts and models, data storage, networking, and network infrastructure.

 
b0lsc0ttIT ManagerAuthor Commented:
Thanks for the response and additional suggestion.  I am interested in trying to use it but not real clear how the second table will help.  Is the "engine=memory" part the key to an improvement?  How or why does it help?
What is the purpose of the filelink field in each table?  Is that for some relationship or what?  The ID seems to be what you use for the JOIN.
Thanks for clarifying and the extra details.
bol
0
 
rydersaintConnect With a Mentor Commented:
I am sorry b0lsc0tt not enough coffee today

SELECT * FROM TABLE1
LEFT JOIN TABLE2 ON TABLE1.filelink=TABLE2.filelink
WHERE fileName = 'the requested filename';

is correct is the correct query

yes the memory engine loads into system memory so all queries are automaticly indexed and very fast
I would recommend this way for any blob queries
 
0
 
b0lsc0ttIT ManagerAuthor Commented:
rydersaint,
Thanks!  I have held off responding or closing this because I was hoping to test it.  That won't be able to happen right away though so I won't leave you hanging.  I do appreciate your help and info.
bol
0
 
b0lsc0ttIT ManagerAuthor Commented:
Thanks!
0
 
rydersaintCommented:
thank you for the points b0lsc0tt
let me know if you need any other info
0
Question has a verified solution.

Are you are experiencing a similar issue? Get a personalized answer when you ask a related question.

Have a better answer? Share it in a comment.

All Courses

From novice to tech pro — start learning today.