In September 2015, I spoke at the International Legal Technology Association (ILTA) conference about what’s new and innovative in data center technology (My presentation can be downloaded here.). To prepare for this event, I started a two-part blog series. My first post highlighted how Nutanix is using commoditized server hardware and redefining its capability into a full scale compute, network and storage environment.
Let’s take a deep dive and focus on defining how commodity server hardware is being deployed in the next generation data center. Effectively, we’ll venture from the organized world of block data, with its neat and tidy application structure, to the chaotic and unstructured world of file.
In 2015, one can say the technology industry has mastered application storage. The compute layer of the server storage network stack has become increasingly more powerful with each new generation of processors. Server manufacturers now offer platforms with multiple terabytes of RAM that will no doubt increase in future product releases. As a result, the current virtual machine (VM) density will only become denser as the future generation of servers hit the market.
As it pertains to the storage layer of the stack, the emergence of solid-state drive (SSD) technology has basically eliminated performance and response time concerns for nearly all applications. Software compression technologies have assisted in taking large data sets and fitting them comfortably into tight physical and logical spaces.
Most application workloads cannot be accelerated any faster than current capabilities of modern hardware. The bottleneck most often is the application itself rather than hardware. If you are setting up for argument, try to keep an industry-wide perspective and ask what the quantifiable difference is when applications are responding inside of one millisecond. Sure, if you write code for a high frequency trading firm, then you might be licking your chops on your response. But for 99 percent of companies and applications, the noticeable difference of performance inside of one millisecond is negligible. For this reason, I say technology manufacturers have provided the greatest amount of utility to the largest pool of customers around the block storage platform.
Backup and Block Storage
My last commentary around block storage centered around backup. The emergence of virtual integration by a number of backup providers has all but eliminated backup problems that plagued the industry for 25-plus years. Virtual Switching System (VSS) integration and Vstorage APIs for Data Protection (VADP) has enabled fast, snapshot-like backups of machines that used to require backup agents and painful push-pull data moving relationships between media servers and hosts. Sadly, virtual backup benefits don’t always benefit virtual file servers for a number of reasons, including VM data capacity limitations, large file sizes and immense quantity of files.
Alternatively, unstructured file data has not had the same luxury of being easy-to-platform and easy-to-backup. The development of fast and or capacious drive sizes doesn’t provide the same benefit for file that SSD technology does for block. To increase the pressure around the file data problem, this segment of data is the fastest growing piece of the data pie chart. Oftentimes, file data is growing at a rate that exceeds the capability of existing infrastructure. So, the result is that my beloved IT professional friends are constantly adding on storage, creating new file silos, migrating to new platforms and doing countless other activities that keep them aboard the bucking bronco that is file data growth.
Frequently Asked Questions
- What platform should I move my file data to that will be resilient to failure and provide business continuity?
- How can I scale my file footprint in a way that doesn’t also scale complexity and cost at an accelerated rate?
- Do I really need to have file data at each of my locations to maximize user productivity?
- Is there any way I can back up all of my file data without using NDMP?
- Can I back up data without using petabytes of tapes; without having to spend the GDP of a small nation on backup infrastructure and across my locations globally?
- Can I break free from having to manage snapshot schedules, corresponding capacity and performance lag on my production storage systems as a result of executing snapshots?
- Can I secure my file data at rest and in-flight in a way that allows me to pass my security audits?
So, who comes to the rescue to answer all of these questions at once? You guessed it: commoditized server hardware. But the server cannot do it alone because it is hardware without inherent intelligence. The first layer of intelligence is added by a cloud provider like Microsoft Azure, Amazon Web Services or Cleversafe, which combine hundreds of physical servers to appear as an object store or flat namespace.
The same could be said for on-premises object storage like EMC’s Elastic Cloud Storage. But a clear path between users and the object store is still missing if the intention is to utilize one of the aforementioned providers. The rug that ties the room together is a developed file system to run on the object store that remedies the pain around storing, migrating, managing, protecting and securing file data.
The Emergence of Cloud-Based Network-Attached Storage
Enter the biggest innovation in the history of file data: cloud-based network-attached storage (NAS). Software as a Service (SaaS) companies like Nasuni have made object storage providers valuable. Nasuni has thoroughly identified pain around day-to-day file operations and holistically remedied the suffering. Because the object store provider scales limitlessly and has multiple sites domestically (and/or internationally), scale and reliability is never a concern. Nasuni’s file system can also scale limitlessly, cruising past a billion files in lab tests without degradation or corruption. In those same lab tests, the Nasuni service was able to take unlimited, immutable and non-corruptible backups. I feel I would be doing you, as a reader, an injustice of epic proportions if I didn’t mention there is no additional charge for unlimited backup capacity. So, next time you talk to your backup hardware and software provider(s), ask them if they’ll let you do unlimited file backups at no cost. My guess is you’ll get one those cheese-ball fake laughs followed by long pause of awkward silence. If you appreciate schadenfreude like I do, this is must-try!
A common objection to cloud-based anything is security. It’s understandable to have paranoia about hackers and physical theft of disk drives when your IT team isn’t managing infrastructure, but Nasuni builds in appropriate security measures to minimize risk. Nasuni also includes 256-bit AES data at rest encryption with keys managed locally, as well as secure socket layer (SSL) protection of data in-flight. Again, this is a standard feature of the Nasuni file service.
I do not want to risk losing your attention as a reader, so we’re only going to solve one more business problem by using cloud-based NAS Nasuni file services: global multiuser read/write aces. In short, Nasuni’s file system running on object store allows global collaboration and versioning of a single file. Effectively, users no longer have to email, FTP or Box.com files all around world. Users don’t have to manage versions, as well as local versus network storage locations.
Having engaged in hundreds of conversations about cloud versus on-premises infrastructure, the most common feedback I receive is, “Why would I pay for someone else’s infrastructure when I can do it myself for less money?” Well, I’m always up for a challenge, so let’s create a fictitious customer example, “Customer X.”
Financial Benefits of Cloud-Based Storage for Customer X
This is an environment with 40TB of file data, one primary data center and another disaster-recovery data center. It also has two additional remote sites with local file serving, which replicates data back to the production data center for backup purposes.
The environmental pain this customer faces is around NDMP backups taking way too long and not being reliable upon restore from disk and/or tape. Physical backup infrastructure needs to be periodically assessed, upgraded and optimized. Production file serving infrastructure needs to be refreshed every three years because hardware and software renewal fees are astronomical. Migrations of data every three years require outages and disrupt IT personnel from projects that add value. I could go on and on about this subject, but I won’t; what I will do is show you how the cost compares between on-premises file architecture and cloud-based NAS in this handy-dandy table:
Of course, anyone could dispute costs of some component over another. Even the shrewdest individual would have a hard time bridging the entire $230,000 advantage of cloud-based NAS. In addition, this example doesn’t monetize security improvements, improved productivity based on multisite employee collaboration, IT productivity increases due to reduction in infrastructure management time and reliable backups with certain restoration.
In the next generation of data centers, I see traditional servers uniformly racked. For block, I see hyper-converged solutions effortlessly running their applications. On the file side, I see on-premises or cloud-based object storage comprised totally of server architecture. Software like Nasuni will define the value of the object store. An example like the one above demonstrates how financially advantageous it will be to move away from on-premises managed solutions and move file environments to the cloud object storage.