At last I have been able to start working on yet another exiting project – SAP ASE IMDB for custom application, with ~1 TB DB as an IMDB template…

I agree, working with IMDB is pretty much fun – most of the operations are cut in time drastically.  E.g., for the DBA-driven system activities, you may expect reducing the time spent on your operation to 1/6-th.  Cool. 

A note to the engineers, though.  On start up, IMDB reads the template DB from the disk before releasing the DB to general use and as it looks, it does it – single  threaded.  Since IMDB is always built from scratch, one must have though about the start-up time with a more critical eye, I think.  It takes about half an hour to recreate IMDB DB of ~1 TB from its DRDB template – which sound reasonable time to wait for the in-memory replica to become available.  But imagine that the start up has been done reading the DRDB DB devices in parallel (DB is not accessible anyway to the DB users until the IMDB DB for which DRDB is the template is brought online).  One might have cut the start up time drastically – 5-15 minutes?. 

Anyone from the SAP ASE engineering team here can pick up the challenge and speed up the start up process?   Seems like something not so hard to implement… 

Guys, I love your product.  It beats all the competitors so far.  But since you at SAP love to be challenged and make the product shine even more – here is an FR for you…

If you need client-side cooperation, be my guest…

ATM.

To report this post you need to login first.

2 Comments

You must be Logged on to comment or reply to a post.

  1. Jon Egerton-Kemp

    Hi Andrew,

    It is always good to hear your feedback.

    I did a quick search and found internal CR 593111 raised for:

    Parallelize reads to improve create inmemory DB performance


    I will pass you comments on the engineer to owns this – it would be good to see it done.


    All the best,

    – jon

    (0) 

Leave a Reply