http://ayende.comAyende @ RahienCopyright (C) Ayende Rahien 2004 - 2021 (c) 202660I mean under the Real Application Cluster. I can't wait to take a look at it. Thanks very much! http://ayende.com/3422/distributed-in-memory-cache-storage#comment8http://ayende.com/3422/distributed-in-memory-cache-storage#comment8Tue, 02 Sep 2008 05:06:47 GMT RAC ? I am not familiar with the term. I recently implemented a couple of distributed caches, you can find the references in the blog (NMemcache) http://ayende.com/3422/distributed-in-memory-cache-storage#comment7http://ayende.com/3422/distributed-in-memory-cache-storage#comment7Tue, 02 Sep 2008 03:16:31 GMThi genius, Great topic! Is there any reference to the implementation of Memory cache, especially under the RAC environment? Thanks Best regards http://ayende.com/3422/distributed-in-memory-cache-storage#comment6http://ayende.com/3422/distributed-in-memory-cache-storage#comment6Tue, 02 Sep 2008 03:08:09 GMTIt is bull, basically. What he is talking about is an approach for concurrency that requires several versions of the row to exists at one time. As a simple example, I read the customer row in one transaction, and you modify it in another transaction. I can re-read the same row in my transaction and get the values that I got when I first touched that row. This is a base feature in Oracle, in SQL Server it is called Snapshot Isolation. This require to keep several versions of the same row in the DB. In general, unless the DB is implemented by idiots, there isn't any way that you will get to that point. http://ayende.com/3422/distributed-in-memory-cache-storage#comment5http://ayende.com/3422/distributed-in-memory-cache-storage#comment5Sun, 20 Jul 2008 23:50:04 GMTHi again. Finally I found the post but unfortunately is not exactly like I said. He says something about multi-versioning that I sincerely don't know what is. Here is the link. http://asktom.oracle.com/pls/asktom/f?p=100:11:0::::P11_QUESTION_ID:821088900346364898 But Oracle has a product called Times Ten that is a in memory database. Do you know anything about it? Thank you. http://ayende.com/3422/distributed-in-memory-cache-storage#comment4http://ayende.com/3422/distributed-in-memory-cache-storage#comment4Sun, 20 Jul 2008 23:41:54 GMTI would like to say where that consultant said that, it doesn't make sense. A 1 GB of data on HD would consume 1 GB of data in memory. There are issues of the model used. Key/value vs. RDBMS. In general, it is the old issue of memory utilization vs. processing time. You can trade one for the other. When using the disk, you tend to prefer to minimize size, because it end up more efficient this way, but two orders of magnitude bigger is way beyond what I would expect. http://ayende.com/3422/distributed-in-memory-cache-storage#comment3http://ayende.com/3422/distributed-in-memory-cache-storage#comment3Sun, 20 Jul 2008 20:29:51 GMTHi Ayende. Do you know why in memory cache can spent much more memory than the database itself? I read on oracle consultant post where he says that 1GB database could use more than 100gb in memory. What do you think? Where can I read more about cache and distributed cache? I already searched in Google but I didn't find what I was exactly looking for. Thank you. http://ayende.com/3422/distributed-in-memory-cache-storage#comment2http://ayende.com/3422/distributed-in-memory-cache-storage#comment2Sun, 20 Jul 2008 20:24:09 GMTThe way I've been dealing with memcache to get around this is that memcache is just a first place to look for readonly data. I have a single service that uses DB transactions for writes and it populates memcache on writes and on cache miss requests, Yeah, that means that the DB is once again the write bottleneck, so this is more useful for scenarios of reads greatly outweighing writes. And yes there's still stale cache issues of the single process writing the new data out in a non-transactional manner. Staying tuned to your approach (my vote would be to build something on mnesia) http://ayende.com/3422/distributed-in-memory-cache-storage#comment1http://ayende.com/3422/distributed-in-memory-cache-storage#comment1Sun, 20 Jul 2008 15:04:14 GMT