Webcast June 19th: Jonathan Lewis – Expert Look at Delphix
Live webcast: Jonathan Lewis – An Oracle Expert’s Look at the Delphix Technology
Date: Wednesday, June 19 @ 9am PT
Click Here to Register
Jonathan Lewis joins us for the 2nd webcast of 3 in his series on Delphix technology and his experiences. Jonathan came out to the Delphix offices in California in March and kicked the tires on the product for a few days and had a chance to talk to some of the creators of Delphix, ZFS, DTrace and Active Dataguard at the Delphix offices.
The first webcast was an informal discussion of his experiences. This second webcast will also be informal but more technical.
Please send us questions by commenting on this blog post and we will try to incorporate the questions into the webcast
Some of the areas we may talk about are
Introduction
- motivation for virtualizing databases covering some use cases that database virtualization solves.
Ease of Use of Delphix
- link to a source database
- provision a virtual database clone
- eadministration automates of change collection and purging of old data.
Delphix on a laptop
- Jonathan and I have set up Delphix on our laptops and we will discuss a bit on how this has worked for us.
- Demo of Delphix (either on laptop or on a “real” machine”)
Technical explanations
ZFS
- ZFS “Keeps a copy of every new version of a block”
- ZFS equivalent of the read-consistent index that gives you the file as at any point in time
- Awareness of Oracle block size – setting logical ZFS block size to Oracle block size
- Being able to compress the ZFS logical block to a smaller number of sectors
- Special case of empty blocks compressing to fit the ZFS meta-data entry
Source database linking
- RMAN full backup run by delphix
- Delphix uses RMAN tape library APIs
- RMAN backup from SCN
- Allows consistent versions of database for EVERY incremental backup taken, forever
- Enable ctwr (block change tracking) to minimise backup times
- Ability to copy all redo logs, and keep up with online redo logs so able to start from any level 1 and roll forward to any point before next level 1
- How old versions of data blocks from the backup history can be dropped when there are no snapshots that depend on them so that the backup keeps rolling forward in time without needing a whole new starting snapshot to be taken
- The complexity of making this task efficient
- The importance and benefit of being able to do it.
Virtual Database
- Creating a vDB and how the new data is created without a history trail and unchanged datablocks come from a snapshot and are mostly the original backups.
Related blog posts by Jonathan’s blog
- http://jonathanlewis.wordpress.com/2013/02/06/delphix/ – intro to visit at Delphix
- http://jonathanlewis.wordpress.com/2013/03/22/delphix-debrief/ – debrief from visit Delphix
- http://jonathanlewis.wordpress.com/2013/03/09/virtual-db/ – EMC vs NetApp
- http://jonathanlewis.wordpress.com/2013/04/04/delphix-overview/ – write up on experiences with Delphix
- http://jonathanlewis.wordpress.com/2013/06/18/delphix-2/ – annouces 2nd webcast
Oracle L
Trackbacks
Comments
the second webcast (with JL) clarified things a lot for me – from the first webcast I had the impression, that Delphix would be sitting *directly in* the production storage access path – which was the reason for my questions on HA for Delphix (as on outage of Delphix would, in such a case, mean an outage for Production).
From the second webcast with JL I understood, that my original impression was entirely wrong, and Delphix would be in fact sitting *outside* of the Production storage access path, merely sending delta changes (via custom SBT library rman incremental archivelog backups) to the Delphix server, and processing them there to create a virtual image of Production as of any distinct version of the DxFS root block, and the only load on top of the actual Production, would be the added CPU+IO load of the frequent custom incremental rman backups (would be good if these could be directly used as Production backups – i.e. made accessible as rman backupsets if Production would need a (partial) rman restore – so that there would not be a need for a double rman-backup done on Production – once for Prod backup purposes, and once by/for Delphix).
Great to hear the second webinar with Jonathan cleared things up.
I like your idea of integrating Delphix more tightly with RMAN. One strategy to tie Delphix more tightly in with RMAN is to make Delphix a backup device which some customers are doing already. Some customers use Delphix as a backup for their databases. Delphix provides awesome logical recover options, for example if a someone drop a table on production or somehow damaged the data on production, then one can spin up a virtual database in minutes and get the correct data back.
Delphix can also be used for physical recovery in the case of physical corruption. When there is physical corruption one can copy the uncorrupted file from Delphix on to the source database and recover.
Delphix can be used as a fail over if the source goes down.
Delphix can restore the source completely with an operation called “virtual to physical” (V2P) where the whole database is written from Delphix back to the source.
Is a recorded Webinar link available for playback ? I couldn’t attend the live Webinar.
Hi Tobias, Yes the recording will be available. I’ll try to get it up on my blog in the next day or two.