Directly following my opening keynote on The Future of Information Infrastructure at Implementing Information Infrastructure Symposium was Steve Duplessie, who is recognized as one of the top people in the world on information infrastructure.
Here are some notes taken from his excellent keynote:
One of the key values for tagging data is to differentiate between the value of information. Currently all data is stored equally. Having metadata allows you to take different approaches to how you store and backup data depending on its value.
EMC and NetApp are direct competitors, and have both been runaway success stories. For IBM and HP, while storage is important, it will never dominate. Dell is now serious about storage.
Disruptive trends in storage include flash (solid-state) memory. There never used to be I/O problems, but server virtualization exposes them. Everything in IT is solid state except for storage. We’re not that far away from no-one buying disk drives any more.
Flash can be implemented in different ways:
– Server-based. Expensive but won’t survive if it is non-networked and shared.
– Network-based. We will work out how to do network caches in front of legacy storage.
– Array-based. Adding flash to an array is a short-term fix – it doesn’t address the cause of the problem.
The only reason to use a mechanical disk in the data center is cost. Within a year we will have 100% flash-based arrays at a comparable cost to disk-based arrays.
Optimizing storage by deduping needs to take place on primary capacity. Deduping in backup treats the symptom not the cause. The closer to the point of origin the less has to be done down the wire.
“Scale-out” storage – scaling by adding nodes connected by an external network that work together as a single system – will become the norm.
Tape isn’t dead. But it will eventually only be used as a last-resort backup: Armageddon.
Backup is broken. Data growth is killing traditional methodologies. Mobile devices are where the enterprise is going, so you have to deal with the reality of backup beyond the server.