Cloud Storage: Ready for Primetime?

Panel: Cloud Storage: Ready for Primetime? @ Structure Conference 2011

Vanessa Alvarez, Analyst, Forrester Research

Val Bercovici, Sr. Director, Office of CTO, NetApp

Dheeraj Pandey, CEO, Nutanix

Andres Rodriguez, Founder & CEO, Nasuni

Dave Wright, CEO, SolidFire

Summary. At Structure 2011 one of the better panels I attended tackled cloud storage and what it will look like in the coming years. At one point in the panel the conversation centered in on the next generation of cloud storage being driven by applications written to take advantage of evolving architectures. However, the use of the term “Cloud storage” is a bit confusing in this conversation because we are really talking about different types of data management applications that exploit cloud. In these applications the storage is integral, yet practically invisible, to the experience. To help make my point, the example applications cited by the panel as “cloud storage” were Dropbox and Box.net. (note: I would add Evernote to the mix).

These applications are really SaaS companies running on Amazon S3 or something like it. But the cloud is an important component of their value proposition. The cool thing here is that these initial applications are running on commodity data buckets like S3. Looking forward, the cloud storage model is going to evolve as new companies build new systems that take advantage of new technologies that enable new applications. What next generation applications will be enabled by storage systems that fully exploit the latest innovations in databases, processing, networking, SSDs, etc.? Analytics, and all of the activity around the NoSQL movement, would seem to be the obvious answer here, but what are the other opportunities we are not thinking of?

My take from all this is that while likely not a tactical driver of business, it is important for vendors to find ways to evangelize the value proposition of application-enablement in the cloud as a result of new architectures. It is in vendor’s best interest to promote the application development ecosystem to write applications to exploit these new architectures. In a slightly different context, this is what VMware did with SpringSource to ensure applications were written for cloud, and they were able to optimize for them at IaaS level. This type of influence should in turn further magnify the relative value proposition of the architecture.  Panelist Val Bercovici from NetApp put it best when he said, said “it is crystal clear that cloud storage will only become real as new applications are developed with the knowledge and awareness of the cloud”. I couldn’t agree more, the application development ecosystem will be a key determinant in the adoption and impact of some of these new cloud storage architectures.

Included below is some of the other more relevant commentary from each of the panel companies:

Nasuni

  • Business is about putting the cloud in the datacenter, have built enterprise class NAS that can use the cloud natively, unlimited capacity, doesn’t need to be backed up, can do DR
  • Nasuni has all you would expect from NetApp like box in the data center
  • Adding cloud into primary storage provides inherent benefits, storage and compute need to be together in data center, never going to move home directories into the cloud
  • Business issues are holding back cloud adoption more than technical at this point
  • Cloud has created new requirements on the technology, unlike anything we have ever seen in enterprise storage

SolidFire

  • Next gen all-SSD scale out primary storage system for cloud computing providers
  • Challenges very different in cloud
    • Performance – need guaranteed performance across VMs
    • Efficiency – driving for high utilization, dedupe, thin provisioning, compression in real time across cluster
    • Management – requires cloud scale mgmt, API based mgmt, moving away from GUI
  • Primary storage in cloud
    • Important to differentiate where servers reside
    • Physics prohibit separating servers and storage for primary storage
    • Primary storage in cloud is more than just VMs in the cloud
    • Challenge is shared disk based infrastructure with a lot of contention, lots of capacity here but not a lot of performance to go around
  • SSD perception
    • Still perception on cost
    • SLC flash in enterprise arrays is very expensive the way it is used today
    • Build system from ground up using MLC and aligned with flash allows price point that is competitive on per GB basis
    • Key benefit of flash is performance, specifically guaranteed performance
    • AWS gives you VMs but when you talk about disk the only number you get is capacity…no guarantees around latency, bandwidth or IOPS
    • Key package is SSD + system-wide efficiency + guaranteed performance

Nutanix

  • Pioneering storage for datacenters that can live w/o SAN, can put compute, storage and VMs in same boxes
  • Data center boxes, allow you to scale compute and storage together
  • Team comes from google file system v1 & 2, also Aster, VMware and XenSource
  • Building all NetApp features into a box, will have flash and all features of enterprise grade network storage
  • Shipping boxes today, launching company at VMworld
  • Architecture is compute, gateway and cloud storage service
    • Have to understand where metadata and file system exists
    • Can’t be putting block to cloud
    • Pushing file system to application to gateway (like Nasuni)…ok for file mgmt, but dead for running exchange server or DB, not sure where data resides, can’t have latency
    • File system and metadata in cloud (box.net, drop box) is seeing huge acceptance, latency is ok in this model
    • “We are not a storage company we are data center in a box”
    • Box.net/Dropbox model is here to stay, SMBs like it
  • Tiering
    • Block can’t go over the WAN, blocks belong as close to the app as you can put them
  • SSDs
    • Massive space savings from SSDs, it is a game changer as long as you can deliver enterprise feature set, otherwise it is just race cars on race track
  • Environmentals
    • Data centers need to shrink, putting compute and storage together can trigger next generation of consolidation
    • IT doesn’t own power in its budget, this is handled by utilities
    • Most people in IT push this problem off to utilities

NetApp

  • Focusing a lot on clouds impact on existing infrastructures, application transformations and end user mobility
  • very interested in the business model changes that cloud is forcing on the entire supply chain
  • Enterprise IT departments very hesitant to move primary storage to cloud
  • Backup is easiest first step, cloud actually adds value here versus adding risk…DR, global access
  • Clearly need new taxonomy when talking about cloud storage, over cloud you have “link of uncertainty”
  • Crystal clear that cloud storage will only become real as new apps are developed with the knowledge and awareness of the cloud
    • Dropbox, photobucket, box.net, Evernote
    • These are the models going forward that will survive and thrive
  • Will save money when you implement SSD correctly “without fail” for performance centric apps, driven by reducing capacity tier versus performance tier disk
  • Question some of the new app architectures, forces us into building too many data centers…5 copies of data is expensive
  • Cloud storage infrastructure at scale and facilities all need to be considered together

 

Share
  • stu

    Unfortunately the video from this panel wasn’t recorded. For those that might want to see related video:
    Val Bercovici of NetApp in The Cube at SNW: http://siliconangle.tv/video/netapps-cloud-czar-transformations-storage-industry
    Dave Wright of SolidFire in The Cube at EMC World: http://siliconangle.tv/video/dave-wright-emc-world-2011
    Andres Rodriguez of Nasuni talking with Stephen Foskett: http://foskettservices.com/2011/01/episode-1-cloud-storage-gateways/