Wednesday Sep 25, 2013

ZS3 vs DataDomain for Database Backups

For those of you attending OOW13 this year, I am sure you are wondering what this ZS3 (ZFS Storage Appliance) platform is and what all the buzz about these phenomenal speeds that we can get out of it are all about.  It excels at IOP's, low latencies and high throughput.  It's primary storage and also a great target for RMAN backups.  You probably also see our friends from EMC displaying the merits of the DataDomain box.

While I am a fan of deduplication technology, it's only good when its used in the right context and for the right workload / data type. We offer it in the ZS3 but rarely do we turn it on.  Compression now that's another story, sure, turn it on and at least use LZJB compression, it's inline, virtually overhead free and it actually speeds up data transfers in most cases.

DataDomain; however, is not the right place to send your RMAN data.  Don't get me wrong, DataDomain is a great box for disk based backups of unstructured data types and all the 20:1 things you hear are very relevant for this kind of data.  Unfortunately, RMAN data streams are not dedup friendly. (Unless you, to quote a friend of mine, "de-optimize upstream in order to improve your backup performance").

What I am saying is, follow all the best practices to make your primary database go as fast as possible and use compression techniques like ACO and HCC and in 12c, use ADO (BTW, you need Engineered systems or Oracle storage for HCC).  Run incremental or differential backups in addition to your weekly fulls.  We want you to get the best possible query speeds etc.  In turn, we also would like you to take a good look at what the current ZFS Storage Appliance line and now the next generation ZS3 platform can do for your RMAN backups.  More importantly, what they could do for your RESTORES! 

So backup that Exadata at full speed, restore at better speeds than any vendor can provide.  Don't have an Exadata, fear not, we can backup/restore any Oracle Database faster than anyone else (provided you have a decent pipe like 10G or even better IB).  

Leave all those performance related settings on and go ahead and multiplex like crazy.  Don't send 4 like they recommended in their session this week, send 16 RMAN streams at a time.  Parallelism is the best way to drive up the aggregate throughput.  The ZS3 platform, ZBA or the current ZFS Storage Appliance line can all deliver results up to 20TB/hr plus for backups and up to 15TB/hr restores.  These numbers are repeatable and we, your local Oracle Storage Teams, are ready to demonstrate them for you any time in our solution centers.

I'll leave you with this thought, typical day to day operations - back up your current database, restore it to an alternate location, clone it, prepare it to be mounted on a different host, how long does that take you today with your current gear?  This is something most of you do regularly so that you can mine that data, do some test and dev work on it etc.  For a large database, this process will take several days with a DataDomain, our customers have told us this many times.  Oracle ZFS based Appliances can reduce the time to complete that entire process to mere hours.  I ask you, what can that ability do for your business?

Oracle on Oracle - it's an awesome story - Hardware and Software engineered together to provide extreme performance !!!

Sunday Sep 22, 2013

Side by Side Comparison 7320 vs ZS3-2

So now that you have either attended the launch or watched the replay, you probably have some questions about how the new box fits into the portfolio and how it compares to its predecessor the 7320.  In the chart below, I will compare them on a contents per controller basis side by side for you.

 ZFS Appliance Model Name 7320 ZS3-2
 Cluster capable (recommended)  Yes  Yes
 Rack Units  1U  2U
 CPU's  2x4 core per controller 2x8 core per controller
 Memory 144G 256G
 IO slots 3 but really 1 is available for IO options 6 and 4 are available for IO options
 Onboard Ports 4x10/100/1000 (RJ45) 4x1G/10GbE (RJ45)
 Read Cache
2TB (4x512G SSD) 6.4TB (4x1.6T SSD)
 Write Cache 24x73G SSD 32x73G SSD
 Expansion Tray Count 6 8
 Drive Types 300G/600G 15k, 3TB 7.2k 300G/900G 10k, 3TB or 4TB 7.2k *

* The drive industry is moving to the 4TB drive so the 3TB will EOL this fall as supplies will diminish quickly.

So from what you see above, the hardware has changed a lot and what you can now do with the box has dramatically changed ... but the biggest improvement is all the software enhancements in OS8. The public benchmarks speak for themselves, industry leading results, almost double the IOPS and almost half of the latency (compared to the previous generations used for the benchmark tests). Oracle has gone above and beyond with this new code base (OS8) and the future is bright for all storage consumers when it comes to cost effective high performance storage. We all get asked to do more with less and this is the solution to that problem.

Stay tuned for more comments as Oracle OpenWorld starts on Sunday September 22nd and the ground breaking combination of Oracle Software and Oracle Hardware will be on full display. 

Tuesday Sep 10, 2013

Oracle Unveils ZS3 Storage

Oracle Press Release: 

Oracle Unveils ZS3 Storage, Delivers World Record Performance and Unrivaled Efficiency for Oracle Applications

New Oracle ZS3 Storage Accelerates High-throughput Business Analytics and Database Queries; Helps Customers Reduce Oracle Database Manual Tuning by 65 Percent; Outperforms NetApp, EMC, IBM and HP

Read more:

Coming Up Next:: My bits and bytes about the new ZS3

Monday Sep 09, 2013

Oracle's Next-Generation Application Engineered Storage Announcement

Webcast Details:

Date: September 10, 2013
Time: 10am PDT/1pm EDT

Register NOW:

Featured Speakers:

Scott Tracy
Vice President, Software Engineering, Oracle

Steve Zivanic
Vice President, Storage Business Group, Oracle 

Jason Schaffer
Senior Director, Disk Storage, Product Management, Oracle


Mike Triska, Storage Solutions Architect, blogs creative thoughts to help you get the most out of your Oracle storage


« July 2016