insideflash

Performance more interestingly DBAs are very interested in capacity savings too: deduplication and compression. More databases on faster storage in my opinion.

consistent performance is a critical component of db infrastructure http://wikibon.org/w...

from our conversations with customers, performance tuning is a top task many DBAs spend a lot of time on, AFAs not only bring extremely low latency for performance, but remove a lot of the steps needed in tuning storage for Oracle databases

Significant performance improvements resulting in better end user experience.

DBAs have too much to think about as it is. It's nice to be able to *finally* be able to provision capacity. A device like XtremIO requires to concern for RAID, etc. Just provision capacity.

So he or she can focus on working to improve database design and application functionality, rather than trouble shooting storage problems

DBAs cannot spend time identifying individual database hot spots and reacting ..each database has its hot spot. Consolidate any number of those and you must have an all flash array #insideflash

The performance problems that take the most time are usually storage performance problems - flash can eliminate those. The performance problems are more challenging are design problems.

Tiered storage is too complex to manage. Just provision some LUNS and forgedaboudit.

But "performance" is easy- consistent predictable performance requires the right architecture

and of course not all of the All-Flash Arrays are created equal so please consider this paper http://www.emc.com/c...

Flash is not just about fast storage - it is an enabler for new functionality and better application design; the old DBA rules need to be rewritten for flash only environments.

let's face it, and http://xtremio.com/d...

And with consistent performance and copy services, DBAs can consolidate entire Oracle lifecycle and related app landscapes

That's the real value- the agility and consolidation for total Oracle workflows!

DBAs understand flash is fast. Its a short discussion. More important to DBAs is deduplication and compression. DBAs are very interested in how these space savings features work with Oracle databases.

a heavily consolidated environment will have too many hot spots to identify and react to (tier reaction) so tiering efficacy can be neutered quickly

The DBA can extend his job by developing space efficient snapshots procedures to create instant latest copy and publish them for developers, QA and testing

Developers, QA and testers use 40% of their time to enable the right data to be available for themselves. Providing and sharing data on flash avoids copy sprawl, reduces storage costs and improves overall development productivity and time to value.

XIO delivers database snapshots or clones in-memory w/dedupe allowing new Oracle instances to be created in min, without impacting prod.

price of flash is relative to the value re: transactional data vs cold storage

storage cost for Oracle database environments can be looked at from a lot of dimensions, one factor to consider is how many copies of production you will have?

most Oracle lanscapes include a lot of copies of data. In-line data reduction and zero-overhead snap/clone can mean fewer "silos" upping the effective capacity of an array

Considering your total Oracle database landscape of production and copies lets you factor in things like compression and deduplication which can have a big effect on the ultimate end cost.

Another angle on cost/price is if you spend a bit more on flash you can reduce Orcl lic & maint costs http://wikibon.org/w...

What this chart shows is that with flash you can reduce the number of cores required - orcl prices DB based on cores

Oracle is going to take over this space

Not only can you reduce license cost, or do more what what you have, the way applications are tested can be dramatically changed... copies with full production performance - http://www.youtube.c...

The true cost is $/effective capacity. That should include compression, dedupe, space-efficient copies, thin provisioning impact.

Another aspect is the fact that XtremIO storage provisioning is literally 3-mouse clicks. This device is as close to self-service and suitable for app-owners as can be. There is cost savings in the model itself I think

Using deduplication, compression and space efficient snapshots means DBAs can have more databases on AFA arrays but the key is performance remains consistently fast too.

seems like just XtremIO is being mentioned...IBM FlashSystem is best performing AFA on the market with MicroSecond Latencty...with DBs...its all about Latency...

Price is no longer an issue if flash is managed correctly. Compression and deduce work better with flash because flash has much higher access density - helps to reduce cost difference.

@dfloyer so you're saying flash is same cost?

However, real benefit of flash comes from organizing the original data to be shared without having to make endless copies. Avoiding sprawl makes flash cheaper than disk to deploy and manage; very different operational management processes needed

Great question, in the past storage guy = storage buyer, DBAs were pretty indifferent...but today we're seeing more DBAs actively involved in infrastructure and even storage decisions.

More and More the buyer is both, the value prop and the conversation is different. best meeting we have are with both in the room at the same time!

More and more these days the DBA has a greater role in the decision around storage and acceleration technologies. I believe that has triggered the trend to enable DBAs via plug-ins & other apps to work with storage.

Database is extension of storag,e so both.

It's a bit of a purchase by committee I think - can't ignore either constituency imo - oh and don't forget about the apps guys

Here's a question- how much self-service should we provide DBAs with array copy services out of OEM? If the array has no capacity or SLA risk, let the DBAs control their destiny?

@fenselau I love that concept - absolutely agree - let the DBAs have self-service - they don't want to talk to the storage guys anyway!

it all depends on whether the all flash array has data services or not. A JBOF offers LUNS. That's a long way from provisioning a database. On the contrary a zero-overhead snap/clone is an immediate *database* provisioning feature

That's the biggest missing piece w/AFAs is the so-called "stack" - it takes a long time to build a robust stack but there's a ton of investment going into flash stacks

That's a good point! There is a new study by VMware on DBaaS and it used XtremIO! 50 by 50GB databases were provisioned in very fast. It took 47 seconds to provision each database. http://www.principle...

snapshot and cloning technology within storage can go nicely with database provisioning. In memory storage snapshots for example to not only create copies fast, but without exponentially consuming storage as you go through dedupe

IMHO that should be independent of storage. APIs are a great middle ground.

that 47s would be time spend in the host not in XtremIO array operations!

Flash is a key enabler of orchestration and automation of database provisioning; consistent low latency is the biggest performance determinant and allows fastest recovery. Self-provisioning of most Oracle databases is possible!

Now that AFA takes storage of assumed culprit list, the bigger issue is how to manage database application SLAs for the system as a whole; great job for good DBAs

Right on! And with the right array services and consistent scale-out performance, it should be an easy job