storagesummit

Wikibon expects tape cartridge storage density will grow faster than disk storage density, and bandwidth for ingestion and retrieval from tape will grow faster.

as long as flash+deduplication doesn't come near enough to the cost of tape, it will remain a viable medium for long term retention (last resort kind)

@oraclehardware See more here: The Rise of the Machines: Rebirth of Tape http://wikibon.org/w...

A huge amount of Internet-of-things data will be steady state. Will that ever go on disk or will it be captured on tape close to the machine where it is generated?

Tape definitely has a place in the hierarchy. In large object file use cases it will perform better than disk if combined w flash

Several well-known giant internet companies are leveraging Oracle StrageTek tape systems to backup their data. Apparently, we have enough drives to satisfy demand.

Tape adoption increasing in cloud as well: Tape in the cloud
http://www.oracle.co...

that depends and what you call "satisfy" - if your SLA is "we will retrieve your files probably within 6hrs to 1 day" than yes

@HansDeLeenheer agree. Tape will continue to have the economic advantage

@HansDeLeenheer But for some uses retrieval in a day or two is adequate -- for instance analysis of bridge live loads to support design of a replacement bridge.

Now wait a sec... Max retrieval from tape is just a couple/few minutes and it is often faster than that

And I am talking about offline tape contained in a large (PBs or even EBs) tape library

I think you're referring to Amazon's Glacier SLAs.

@oraclehardware I think that's a lock

That is Glacier SLAs. That's why they call it Glacier. And Amazon says it is a disk-based archive, no tape storage

And online tape, that is tape already on active drives, or in robotic libraries, can provide access that is much faster than an hour. Not fast enough for near real time, but a lot of analysis does not have that short a fuse.

@cmosoares you may have spotted my intent there ;)

It used to be about data life cycle management. Still is but now with flash it's done at a way faster life cycle

@dvellante Agreed - we went from three tiers (performance disk, capacity disk, tape) to four tiers (flash, performance disk, capacity disk, tape).

the big difference for me was when we started seeing sublun-tiering. so a multi TB database could reside for 90% on cold disks.

David Floyer has advocated a two-tier system with flash and online tape as a future tiered environment.

that will never work. the basic problem of tape in any form is that it is transactional data that never will have the possibility to follow random IO

there will always be a certain point where there are not enough tape drives available to supply the demand

it's the exact same reason why tape doesn't work for today's backup: because you can't run a VM from it for example - tape is for retrieve usage such as long term retention

Several well-known giant internet companies are leveraging Oracle StrageTek tape systems to backup their data. If we can satisfy their demand, I think we're OK in the tape drive supply.

there's no disagreeing on that note :)

Definitely, I think the average is ~80% is inactive in most organizations. Making archive and long-term, cost effective retention critical.

As much as 80% of an organization's data is not accessed after the first 30 days of being created. Tape is the most economical way to store cold data.

-> Will that change over time as:
1) more data is created &
2) keeping it active costs go down &
3) organizations desire to implement "big data" strategies to use data it in new creative ways

and the most expensive one to retrieve from

Clipper Group study showed disk-only long-term storage costs 26x more than tape. http://www.oracle.co...

@JeffFrick good point Jeff! we need to learn not to keep every single bit forever in a dozen copies

@HansDeLeenheer There are also legal reasons for eliminating some data at the end of retention periods dictated by regulations.

oh, I know. I've seen some multi clearance level types of data that couldn't be touched by retention systems at all

I'd say well over 2/3rds of data is inactive. Well over.

New high-capacity, high-performance tape drives, such as Oracle's T10KD, make the retrieval process fast and cost-effective.

@cmosoares Good info on that here: T10000D Infographic
http://www.oracle-do...

Reduce IT costs, increase performance, manage growth, gain increased scale...

aligning the business needs of certain data to it's economic relevance

Promary benefit is optimizing the performance and cost of storage with the changing value of data over time. One of Oracle's customers has done just that

Navis Takes Multi-Tiered Storage Approach with Oracle (Video)

http://bit.ly/1lGdvM...

Depends on applications and use cases in effect. Could be perf. disk and capacity optimized disk in a single subsytem or could be independent hardware tiers, i.e. flash, spinning disk, tape

Managed by an HSM or archive manager software

The latter, that is

I think it's about aligning data to the best storage so that you optimize performance, capacity and cost simultaneously

he who still needs to explain tiered-storage should not be in a storage discussion :)

Can you please define HSM?

Tiering started with different speeds of disks - has flash made a dramatic change in how the technology is used?

Also, we should discuss issues such as long term data retention for compliance, archiving or BI/BA purposes. as well as how to manage tiered storage.

Not in a long term retention, i.e. archiving environment. Flash has less value here.

i think of tiered storage as aligning different capabilities to different sets of data to optimize cost.

@KerstinWoods -> tough to optimize all three legs of that stool.

But a great HSM with robust policies might do the trick

The key here is that not all data is equal yet so often we end up treating it all the same - putting it all on the same storage systems - such a waste!

@JeffFrick But that's the beauty of tiered storage :)

-> Does the media demand bifurcate? Flash (speed) and Tape (cost). Not much left for spinning disc in the middle

there are 2 types of storage tiering: the first one is multiple capabilities in a single platform/system. This could be that 1 file/volume resides on one of the tiers 1/2

How about matching the economic value of data to storage device? High-value, high access demand to fast spinning disk or flash, lower value to SATA; dormant to tape?

@KerstinWoods -> like designing an airplane, constant tradeoffs until you get an optimal solution for the objective at hand.

the second type is where a single file/volume is automatically distributed over different storage types

HSM = Heirarchical Storage Management. Been around for a long time. Also, sometimes referred to as ILM. But there are probably some distinctions between HSM and ILM that not everyone is aware of or agrees to

i see flash generally used to avoid tiering.

but tiering as such has nothing to to with archiving. maybe about retention but not archiving.

Let's say about the storage infrastructure behind archiving

Flash instead of tiering - Wow - that's an awfully expensive solution for a lot of inactive data.

no, archiving as such is the act of removing data from a production location

Archiving includes search, content awareness, compliance and deletion/scrape policies

tiered storage does not necessarily define or help with any of the above

Yes, removing from production environment, but still may be the primary copy of the data

if you are removing data from production and it's the only copy, you got to fire your data architects :)

In other words, its not necessarily data protection, though archiving may include BC/DR capabilities to support the archive

A good archive manager will manage multiple copies of the file across multiple tiers of storage.

Automatically based on established policies, of course.

This actually can reduce dramatically both production storage requirements and data protection (secondary storage) requirements. A lot of people don't recognize these added benefits of archiving less active data from production

"Performance Disk" is becoming an oxymoron - if it spins it's slow