OK, so down to business. Here are my top 5 inflection points for storage technology in the next 3-5 years.
- Offline Storage (Tape) becomes extinct for most uses
- Flash becomes a viable Tier 1 storage option
- High Capacity/Low Cost Disk becomes the principle “bulk storage” medium
- FCoE (Fibre Channel over Ethernet) SANs become the FC evolution path for OLTP storage and enterprise data centers
- Web Storage Applications move away from SCSI and File System protocols and become connected principally via “Object” protocols (e.g. SOAP, REST).
Here are some deeper thoughts on each of these topics.
1. Offline Storage
becomes extinct for most uses
Yes, I know, I know – folks have been saying “tape is dead”
for so many years that it has become the Chicken Little story of IT (right next
to the death of the mainframe). One point to remember is, even when the change
is evident, existing technologies do not simply just fall off the planet. We
can still buy VCRs right? Yet, I would hardly consider VCRs relevant to the
video
There are so many factors here. The cost/availability of
network bandwidth, the cost of the “people” side of tape storage and handling,
disk cost declines, multi-site DR, data de-duplication and many more factors
will drive this change. Suffice to say I think this one is happening now and
will accelerate going into next year. Sure it will still take time to
transition but, just because it is a long trip, it doesn’t mean we don’t know
where we are headed.
Short of a few things, there is just not that much that we really need to store in caves. You could argue that some information is replicated so much that there is no need to backup at all.
2. Flash becomes a
viable Tier 1 storage option
Folks are starting to talk more and more about this one. Flash is fast (relative to disk), especially in terms of read latency. The technology is power efficient, and the costs are plummeting. As the total read/write cycle limits improve, expect to see flash begin to play a role as a Tier 1 storage technology especially for OLTP applications (see the previous blog if you don’t understand why).
3. High Capacity/Low
Cost Disk becomes the principle “bulk storage” medium
If you can imagine flash growing for the very low latency applications and offline (tape) shrinking, where does all that data go? I believe it goes to very basic low-cost disk. Performance is not an issue for most applications and the data can be spread to both utilize the capacity and reduce effect of individual disk performance limits. These drives will also be designed to facilitate power saving features.
With the technology and power costs/limitations of moving to
higher performance drives (power consumption grows as the cube of rotational
speed), I believe that we will achieve performance more effectively in two
ways. Flash will be used for OLTP applications, and simple replication will be
used for Web Storage Applications. This will make high-capacity power-optimized
disk drives an ideal fit for most data.
4, FCoE SANs become
the FC evolution path for OLTP storage
While there is always discussion about the future of SANs, I
believe they are alive and well and will carry forward, over time, with a
physical (but not protocol) consolidation to Ethernet. There have been a number
of contenders, namely iSCSI and Infiniband, which have tried to challenge SANs.
While each has made inroads, each also has significant other technical and
commercial limitations that will prevent mainstream adoption. FCoE allows the
consumer to build one physical infrastructure for both their IP and FC needs
but offers two additional key advantages. First, running on native Ethernet
allows FC to run “at speed” with all of the present capabilities maintained.
Second, it can be connected to an existing FC SAN without a complex gateway.
Data can travel without protocol change or significant latency across the two
mediums.
While there are several options out there to take a bite out of the
traditional SAN
5. Web Storage
Applications move away from SCSI and File System protocols and become connected
principally via “Object” protocols (e.g. SOAP, REST).
This is, by far, the most significant and long term inflection point but is also one that I am confident will happen. In fact, it is happening right now with Web 2.0 applications and with almost any application that uses an Enterprise Content Management (ECM) platform. When an application assess information though EMC Documentum or a service like Amazon S3, it is using this “object” principle.
A big part of moving forward with the concept of Web 2.0 or,
specifically, Information 2.0 is the simple need to decouple information from
applications. To do this, we must explicitly change how we package the
information so that it can be tiered, protected, and secured independently from
any single application service. Obviously, interacting at the block or even file
level is just not going to meet this need.
In the next part, I will give my take on the evolution of information protection and availability.
Mark…
Mark,
One thing you didn't mention in tape's favor is green computing. Tape is perhaps the greenest of IT products. Supporting your point, however is there appears to be a lack of leadership in the tape industry (with STK gone).
Some additional thoughts from "Can tape be eliminated?"
http://www.wikibon.org/Can_tape_be_eliminated%3F#Virtual_tape_without_tape:_Half_full_or_half_empty.3F
Posted by: Dave Vellante | August 07, 2007 at 01:09 PM