IT Trends - Oh Brother Where Art Thou?

Ok, it’s that time of year where we do a Midyear sanity check on Technology Trends… Basically, what is hot now, regardless of what Gartner predicted at the beginning of the year?  So what are the leading topics that CIOs are presented with as they forge forward toward the year end, or basically we ask, IT Trends, "Oh Brother Where Art Thou?"

We dug in and found the technologies being deployed were in fact the ones that were predicted. Hence, if you are considering new projects, you’d better be aware of these technologies, for not only are they “top of mind” with many CIOs, but may very well determine the success of your next venture.

• Big Data
• Cloud
• Data Protection
• SSD
• Virtualizatio 

Big Data

An integrated and centralized data store model that enables stakeholders from throughout an organization to harvest and analyze data on the same platform continues to be the goal of many organizations, even though it proves time and time again to be the absolute wrong answer.  Driven by social media, mobile computing devices and cloud computing, the volume of Electronically Stored Information (ESI) is increasing at an accelerating rate.  As a result, they find as the data starts accumulating into giant amounts, the simplest query could take an inordinate, if not unacceptable amount of time.   

In comes Big Data technology.  The rewards for successfully managing and leveraging what is now know as Big Data are well documented. At IBM's Smarter Analytics event in March 2012, clients and partners presented success stories about how organizations are driving business value out of big data, analytics, and IBM Watson technology.

Examples included:

• City of Dublin, Ireland, using thousands of data points from local transportation and traffic signals to optimize public transit and deliver information to riders.
• Seton Healthcare mining through vast amounts of unstructured data captured in notes and dictation to get a more complete view of patients. Seton currently uses this information to construct programs that target treatments to the right patients with a goal of minimizing hospitalizations in the way that most efficiently optimizes costs with benefits. The ability to mine unstructured data gives a much more complete view of patients, including factors such as their support system, their ability to have transportation to and from appointments, and whether or not they have a primary care physician.

Cloud: 31 Flavors

A convergence is happening in the cloud service provider space. More cloud-based archive and backup providers are evolving to account for transactional/production data while managed service providers want to extend their reach into the archival/backup space.  As a result, cloud service providers are evolving their offerings so they encompass transactional as well as more specific bulk/archival storage requirements.

Cloud service providers have the challenge of providing cloud services - compute and local storage - as well as meet client requirements to extricate their data from their local site for business continuity (BC) or disaster recovery (DR) purposes. There are a few things that play into in terms of pressures that enable that opportunity. Some are regulatory compliance for public companies. They need some sort of business continuity plan which requires they store their data offsite. In some cases, regulated industries like financial industry in California cannot ship their data outside of the state. You also see this in Europe. They have to keep all of their financial data within the borders of the EU. It is similar situation in California. Banks need to know where their data lives - that is not always simple - especially when dealing with a larger service provider like Amazon. Some cloud service providers, have cloud storage that vary in types. There is transactional storage which is extremely fast and local. Then there is bulk storage. This is network attached, and SATA storage that is not transactional but still very fast. Finally, there is archival storage. Archival storage is large amounts of data that need to be stored for extended periods of time or is driven by regulatory requirements to get it out of the area or out of the local data center.

Data Protection

No one disputes that enterprise backup has changed significantly in the last decade with the introduction of disk as a backup target and deduplication being the largest contributors to that shift. But in the last few years, array-based snapshots are emerging as the next big wave in how enterprise data protection will be done. As that shift occurs, it becomes more critical than ever for organizations to understand the right role for disk-based backup solutions to play in today's new world of array-based snapshots.

Array-based snapshots are becoming the first line of defense in a growing number of enterprise data protection schemes for five main reasons

• Server virtualization.
• Growing data stores
• Less application disruption as a result of shorter backup windows
• Data change rates
• Shorter recovery windows.

Reasons like this are why analyst firm Gartner recently forecast that by 2015 at least 25 percent of large enterprises will have given up on conventional backup/recovery software and have switched to snapshot/replication techniques. 

This trend would seem to suggest that the end of using disk as a backup target is nigh and that snapshots are poised to successfully replace disk-based backup targets in enterprise data protection schemes.

While an interesting hypothesis, this is unlikely to occur for a number of practical reasons. Just as snapshots have their place in future corporate data protection schemes so do disk-based backup targets. The reasons:

1. Snapshots may not meet regulatory retention requirements. Organizations may use backups to retain data needed to satisfy certain regulatory requirements. However laws like HIPAA require some data to be retained up to six (6) years to prove that it was not breached. This may be longer than the organization is able to retain a snapshot or even longer than the organization keeps the storage array itself.
2. Snapshots are 'risky' if used alone. The presumption is that snapshots are always recoverable should data loss on the production file system occur. However, anyone who is familiar with backup instances is also familiar with investable failure.  Hence, should the production file system be lost or corrupted any snapshots that reference that file system become useless
3. No geographic diversity. Array-based snapshots typically reside on the same disk as production data and reference the same data. So if the production storage system on which the snapshot(s) reside goes offline for whatever reason (catastrophic hardware failure, power outage, whatever) the snapshots on that array also become inaccessible. Buzzzzer…
4. No technology allegiance. Array-based snapshots are tied to a particular storage array so organizations may only restore a snapshot to that same array and, in some cases, Nuff sed.
5. Snapshots best suited for short retention periods every snapshot technique (allocate-on-write, copy-on-write, split-mirror, etc.) consumes varying amounts of production disk storage capacity. While some techniques consume far less than others, none easily facilitate the long term retention of snapshots due to the growing amount of disk space they consume over time. Solid state drives (SSDs) further aggravate the cost of retaining snapshots on production storage as they may cost anywhere from two to ten times as much as regular hard drives.
6. Array-based snapshots are inefficiently replicated. As a means to move data offsite, many storage arrays use replication to get a copy of data offsite. However most of these systems only replicate at the block level as a snapshot is a record of changed blocks. The problem this presents is that if only a single byte changed within a block, the storage array would need to transmit the entire block of data instead of just the changed byte.

Solid State Drives (SSD)

While a number of storage arrays already offer flash as either cache or solid state drives (SSDs), more storage vendors are looking to move flash closer to or into the server. Once there, they are then looking to manage the server flash in a manner akin to how they manage data placement and storage capacity on their arrays.

Flash memory arrays have already earned a reputation as being the most highly performing and energy efficient storage systems available. However the level of trust that enterprises have in the storage management software on these arrays is still being built which results in enterprises being reluctant to use flash memory arrays for hosting anything but a few performance-intensive applications.

The arrays further differentiate themselves from HDD-based storage arrays by better managing flash memory's idiosyncrasies. Garbage collection, wear-leveling, and write alignment are just some of the tasks that need to routinely run in the background to both ensure the integrity of the data stored on these flash memory arrays and extend the life of their flash memory media.

This combination of factors has resulted in a growing number of large enterprises using flash memory arrays in lieu of HDD-based storage arrays in support of specific applications needing high performance. Two such examples are:

• AOL replaced five shared high end storage arrays used to support a 50 TB database application with a Violin flash Memory Array to reduce response times from an average of 10-15 ms to under a millisecond.
• A healthcare organization replaced its Dell MD1000 storage array and its 15K RPM HDDs with a Violin flash Memory Array to support its database application. Using the Violin flash Memory Array it saw an increase in random read throughput from ~250 MB/sec to as high as 1050 MB/sec.

The improvements in application performance that resulted from each organization's use of flash memory arrays are clearly impressive. However the deployment of these flash memory arrays in these two cases underscores an important point: they were only deployed in support of a specific application. In order for enterprises to adopt flash memory arrays more broadly, flash memory arrays need to include software that optimizes their available capacity, protects their data and is viewed as reliable by enterprises.

Virtualization

Once the term "virtualization" overcame the stigma of being "evil", VMware was arguably the largest beneficiary of the corporate transition from the physical world to the virtual one. But as competing hypervisor platforms from Citrix, Microsoft and Red Hat mature, VMware finds its status as king of the virtualization hill under assault with storage vendors publicly sharing they are seeing more customers implementing multi-hypervisor environments. Never one to be put on the defensive, VMware has since moved beyond being a "virtualization-only" provider and is positioning itself in a broader context: A software management solutions provider that can help bring about an end to the schizophrenic state of corporations.

VMware is putting added emphasis on delivering an integrated management solution in the form of products like VMware vCenter that do more than just manage the corporate infrastructure - it is actually being architected and constructed to understand and interpret the data that it is collecting. While this sounds ambitious (and it is,) it is also absolutely necessary for VMware to execute upon if it is to maintain its perch in the hypervisor space.

This is driven by the fact that virtualization has made some tenets of data center administration that were once very simple to administer much more difficult. Case in point, it used to be very easy for administrator to know what physical resources were allocated and used by a specific application. That is no longer true

Granted, organizations can now accomplish tasks using virtualization in ways they never could before such as dynamically moving VMs from one server to another whether that server is in the next rack or the next country. However the tools that give corporations the ability to confidently understand how the VM is impacting the physical resources of the infrastructure on which it resides now or predict how it will impact the physical resources of the infrastructure to which it may move are still woefully inadequate.

Further, to move that VM from one site to another and then manage is still dependent on the hypervisor technology being the same at both locations. This has to concern VMware because, with multitude cloud service providers (CSPs), once you get out of the corporate data center and look to move a VM to a CSP, many of these providers do not use VMware. Instead CSPs are using Microsoft Hyper-V or, more likely, Red Hat Linux and its Kernel-based Virtual Machine (KVM).

Since more organizations - not less - are moving to the cloud and CSPs are likely where their data will end up, this story does not end well for VMware. It must either dramatically lower its pricing on vSphere so CSPs begin to adopt it in lieu of Hyper-V or KVM or deliver a new solution that facilitates the movement of VMs between different hypervisor platforms and can interpret how VMs will perform in each.

So VMware must instead develop and deliver technologies that simultaneously deliver dramatically lower costs with radically different experiences that, in effect, enable corporations to deploy virtualization that deliver the experience users wants without them becoming the management nightmare they are now.

Source(s):

• http://www.dcig.com/2012/08/virtual-federated-data-stores-are-attractive-alternative.html
• http://www.dcig.com/2012/06/bc-dr-compliance-driving-csp-convergence.html
• http://sepaton.dcig.com/2012/08/the-right-role-for-disk-based-backup.html
• http://symantec.dcig.com/2012/08/symantec-partnership-violin-memory.html
• http://www.dcig.com/2012/07/vmwares-new-vision-bring-end-schizophrenic-corp-state.html

So “Once more unto the breach, dear friends, once more;”

____________________________________________________________

About Rick Ricker

An IT professional with over 21 years experience in Information Security, wireless broadband, network and Infrastructure design, development, and support.

For more information, contact Rick at (800) 399-6085