July 25, 2011 at 9:07 pm
Comments posted to this topic are about the item NewSQL Can Save the World
July 26, 2011 at 2:09 am
I read, "At least I hope.." and "perhaps.." in this editoral, and regain the memory of the funniest database parody I have ever heard:
www dot xtranormal dot com / watch / 6995033 / mongo-db-is-web-scale
Please enjoy!
July 26, 2011 at 3:48 am
Running in memory scares me, ... ... I still think a large scale power failure could be disastrous, so I assume these products do write to disk at some point.
Depends on what type of memory. There are many different technologies. Some are non-volatile memory which retains it's state even if the power is turned off. The newer SSD drives are memory and will retain it's information during a power failure as well.
July 26, 2011 at 5:19 am
So a man who runs a company which sells a competitor to "oldSQL" products has given an interview saying the "oldSQL" is bad, and his "newSQL" competitor is doubleplusgood?
OK, I'll trust every word he says then. :rolleyes:
July 26, 2011 at 8:48 am
cengland0 (7/26/2011)
Running in memory scares me, ... ... I still think a large scale power failure could be disastrous, so I assume these products do write to disk at some point.
Depends on what type of memory. There are many different technologies. Some are non-volatile memory which retains it's state even if the power is turned off. The newer SSD drives are memory and will retain it's information during a power failure as well.
Azure is indeed a "VM in the cloud"; however, it is stateless. Azure Storage Services offer you two choices of persisting your data: either in a fully relational SQL Azure database - SQLS with lots of restrictions -, or in Azure Tables. Azure Tables are a fast, non-relational lightweight service that typically stores entities and properties. Azure blobs are for arbitrary binary data - contents is whatever you put there.
July 26, 2011 at 9:04 am
Running solely in (ram) memory scares me as well, and almost dictates that you operate as a multi-node system.
I would hope that there is a "save to disk" operation that could be invoked (by default) when you choose to shutdown the server/service.
I briefly read through the C# sample and the commands seemed a little ambiguous. I guess I like structure too much
July 26, 2011 at 9:11 am
Efficient relational database design is about constraints, normalization, and discrete data types. The problem faced by companies like FaceBook is not that their databases are large or that they have a lot of users (other organizations like banks and governments handle similar data and user loads), but rather the type of content they contain in the database: nested hierarchies, entity-attribute-values, BLOBs, and click stream recording.
"Do not seek to follow in the footsteps of the wise. Instead, seek what they sought." - Matsuo Basho
July 26, 2011 at 9:25 am
We have been looking at SAP HANA - their in memory column store database. The demonstrated performance for a 6TB database was phenomenal - for reading... but we've a lot of questions re read/write and getting those log entries to disk just as quickly. This has been designed from the ground up for in-memory optimisation, including the application sitting on top.
This may be a way to go for reporting, analytics, planning and heavy business planning processes, but moving an ERP system into memory even on highly resilient IBM POWER7 architecture is a scary concept.
Another question for these systems is 'how long does it take to load 6TB+ into memory on restart' or DR?
Microsoft Research is also working on Trinity, an in-memory Hypergraph database which looks very nice as an in-memory, up to 100 node, object store - backed onto SQL for data integrity, storage and reporting. Its code is similar to MSAGL from some time back for visual graphing, so should be very accessible when stacked atop MS SQL.
July 26, 2011 at 9:41 am
If only I was a Linux person and had the time to play...
The little light bulb went on when I was discussing this with a co-worker. I think there may be a real potential for this when coupled with the LPS distro ( a Defense Dept Linux run from CD or Flash drive). It would kind of be like an industrial version of the Android/Sql-Lite combo
July 26, 2011 at 10:05 am
We evaluated an in-memory database a while back (Times Ten), and data was persisted to disk, just not as frequently. The database read all the data from disk into memory at startup, and wrote to disk as it changed.
An in-memory database usually means that, post-startup, the database doesn't go to disk to read.
I have a former co-worker who works for MySQL. She said that several former MySQL engineers work for Facebook, and MySQL has a team devoted just to supporting the social media company. The type of tickets Facebook tends to log with MySQL are,
"I was debugging mysql and this mutex in the whatsit is staying locked even though the whatsit should not be in use".
At Facebook, they have an understanding of MySQL that goes all the way down to the metal.
As a former MySQL and Oracle DBA, I say that you shouldn't knock it till you try it.
July 26, 2011 at 10:26 am
David In BC (7/26/2011)
. . . As a former MySQL and Oracle DBA, I say that you shouldn't knock it till you try it.
I think that some general principles can be taken for granted: it is great for read-only stuff such as cubes used for online analytics or reporting, and no better than traditional SQL database for frequent writes typical for transactional apps.
July 26, 2011 at 10:31 am
Does anyone have a link to the performance differences for all the different database products out there? I heard a rumor that the MS SQL license agreement prevents people from posting performance comparisons but I haven't confirmed that myself.
I'd like to see (using the same data set) the performance for:
* MS Sql
* MySQL
* Oracle
* Teradata
* DB2
* PostgreSQL
July 26, 2011 at 10:55 am
cengland0 (7/26/2011)
Does anyone have a link to the performance differences for all the different database products out there? I heard a rumor that the MS SQL license agreement prevents people from posting performance comparisons but I haven't confirmed that myself.I'd like to see (using the same data set) the performance for:
* MS Sql
* MySQL
* Oracle
* Teradata
* DB2
* PostgreSQL
That would be a MAJOR exercise: results may, and likely will, significantly vary with the hardware. Even if we agreed that the machine will have say 8 GB, and the same (or roughly same) clock speed, there will be differences betweeen Nehalem and Sandy Bridge and - I have not tried that - AMD.
July 26, 2011 at 11:03 am
Revenant (7/26/2011)
cengland0 (7/26/2011)
Does anyone have a link to the performance differences for all the different database products out there? I heard a rumor that the MS SQL license agreement prevents people from posting performance comparisons but I haven't confirmed that myself.I'd like to see (using the same data set) the performance for:
* MS Sql
* MySQL
* Oracle
* Teradata
* DB2
* PostgreSQL
That would be a MAJOR exercise: results may, and likely will, significantly vary with the hardware. Even if we agreed that the machine will have say 8 GB, and the same (or roughly same) clock speed, there will be differences betweeen Nehalem and Sandy Bridge and - I have not tried that - AMD.
Can't all of those run on the same hardware? In fact, if the hardware wasn't exactly the same, I wouldn't be interested in the results anyway.
You would have thought someone would have done this comparison already but I haven't been successful in finding one.
July 26, 2011 at 11:16 am
cengland0 (7/26/2011)
Revenant (7/26/2011)
cengland0 (7/26/2011)
Does anyone have a link to the performance differences for all the different database products out there? I heard a rumor that the MS SQL license agreement prevents people from posting performance comparisons but I haven't confirmed that myself.I'd like to see (using the same data set) the performance for:
* MS Sql
* MySQL
* Oracle
* Teradata
* DB2
* PostgreSQL
That would be a MAJOR exercise: results may, and likely will, significantly vary with the hardware. Even if we agreed that the machine will have say 8 GB, and the same (or roughly same) clock speed, there will be differences betweeen Nehalem and Sandy Bridge and - I have not tried that - AMD.
Can't all of those run on the same hardware? In fact, if the hardware wasn't exactly the same, I wouldn't be interested in the results anyway.
You would have thought someone would have done this comparison already but I haven't been successful in finding one.
TPC publishes database processing benchmarks for a handful of the big name database vendors running on specific hardware configurations.
http://www.tpc.org/tpcc/results/tpcc_results.asp?orderby=dbms
Of course some databases are optimized to perform very well under a narrow range of case usages, like querying EAV or BLOB data. Also, some of the more mainstream RDMS engines (I think that MySQL InnoDB is one example) can be configured for very fast data loads in a *staging* environment, because it's possible to totally disable tranaction logging. Of course that only useful in a staging environment or for populating write/once read/many tables referenced by a website or data mart.
"Do not seek to follow in the footsteps of the wise. Instead, seek what they sought." - Matsuo Basho
Viewing 15 posts - 1 through 15 (of 24 total)
You must be logged in to reply to this topic. Login to reply