Comments: Mycached: memcached protocol support for MySQL

Showing 1 - 9 of 9 comments.

While I enjoy that the actual analysis part of your smart cell phone market posts, you often sound for instance an Embittered Old Fart Apple haterboy.

I saw your blog awhile back and saved it to my bookmarks. Only recently have I got a chance to reading it and have to say great work.

Woh Everyone loves you , bookmarked ! My partner and i take issue in your last point.

it seems a very interesting approach but what's the difference from this one http://yoshinorimatsunobu.blogspot.com/2010/10/using-mysql-as-nosql-story-for.html ? they seem similar, so it would be interesting to see which one performs better on the same machine.

This is very interesting - thank you for doing this.
If you intend to access the same underlying table, it would be interesting to see a performance chart of equivalent operations using MySQL and their network protocol. I'm curious what the max concurrent requests are and the queries per second.

Thank you all for your comments.

Mark, I agree that a memcached-protocol-compatible server using embedded InnoDB as a backend would be an excellent durable KVS. However I cannot live without SQL and my current intention is to add a faster alternative access route to MySQL.

Yoshinori, Mycached is designed to return "structured" values as defined in the DDL of MySQL. And once I add support for "set" operation to mycached, it will only accept structured values. This "schema-first" approach is inevitable since I want to access the same table using both SQL and memcached protocol, but might be an obstacle for using mycached as a session store, etc.

BTW, I haven't yet done any benchmarks other than the chart shown in the entry.

This is great work. If you are willing to give up on SQL, then use embedded InnoDB with the memcached prototol.

i haven't learn about using udf,waiting for your good news !

This is very interesting.

Do you have benchmarking results on the following conditions?

1. get operation vs SELECT on disk i/o bound conditions (database size is much bigger than innodb_buffer_pool_size)

2. set operation vs INSERT/UPDATE (innodb_flush_log_at_trx_commit=2or0)

I assume this UDF + MySQL might work very well as a durable key-value store, for example storing session data, last access timestamp etc.


Post a comment