Today’s Backup And Recovery Menu: A Recipe For Disaster

By: Steve Wexler at Silver-Peak Systems

Recipe BoxWhen people talk about the skyrocketing amount of data being created — and consumed — they typically focus on the storage, bandwidth, or analysis aspects, but there is another critical element that has major implications for organizations: backup and recovery, or business continuity. Issues like productivity, service level agreements — i.e. recovery point objectives (RPOs) and recovery time objectives (RTOs) — backup windows, disaster recovery, and governance are all affected by this data deluge.

In the recent Best Practices for Repairing the Broken State of Backup, Gartner Analyst Dave Russell said the biggest problem areas for backup include capability, cost, and complexity. Legacy backup approaches don’t scale well, as they cannot handle increased demands for protecting today’s data sources and volumes to meet the backup window.

Clients’ top capability complaints include not meeting the backup window, not being able to restore data fast enough to meet SLAs, point solutions not being able to fully protect all data types or locations, daunting administrative interfaces, and incomplete reporting. They also feel that too much time, expertise, and effort are spent keeping the current recovery systems afloat. Organizations would like backup to be a process that requires far less supervision and administrative attention, and for higher backup and restore success rates to be easier to achieve.

In a recent Dell whitepaper, the vendor listed a number of backup and recovery challenges, including:

  • lack of operational oversight — backup and data integrity verification are difficult to impossible, and there is little confidence in reaching RTOs and RPOs, while continuous data protection (CDP) with integrity verification provides more confidence but is resource-hungry and expensive to run;
  • exponential data growth results in poor backup and restore performance and requires frequent provisioning;
  • virtualization growth has a serious impact on backup resources;
  • inefficient data silos have been around for years, and are more costly and inefficient than ever, and they also impact WAN transport costs;
  • SLAs are crucial to maintaining application availability but many data protection products offer mediocre insight at best into application recoverability, while data silos running backup point products worsen the problem, making it very difficult to test and remediate recoverability; and,
  • heavy management overhead is epidemic in traditional backup environments.

The fact that backup is broken was made blatantly clear when I spoke with Silver Peak recently about their latest replication acceleration software upgrade. A customer said they were taking 27 hours to complete their daily backups and full replication was taking 95% of the week to complete. One reason I became a journalist was my lack of math skills, but even I can figure out that you can’t fit very many 27-hour backups into a 24-hour period.

However, backup and recover is not only broken, it’s expensive in terms of downtime and data loss when it doesn’t work.  On an industry/hourly cost, these are the consequences of downtime:

  • brokerage services — $6.8 million;
  • energy — $2.8 million;
  • telecom — $2 million;
  • manufacturing — $1.6 million;
  • retail — $1.1 million;
  • health care — $636 thousand; and
  • media — $90 thousand.

The costs of data loss are even higher:

  • 93% of companies that lost their data center for 10 days or more filed for bankruptcy within one year of the disaster;
  • 94% of companies suffering from a catastrophic data loss do not survive — 43% never reopen; 51% close within two years;
  • 30% of all businesses that have a major fire go out of business within a year; 70% fail within 5 years;
  • 77% of companies who test their tape backups found backup failures;
  • 7 out of 10 small businesses that experience a major data loss go out of business within a year;
  • 96% of all business workstations are not being backed up; and,
  • 50% of all tape backups fail to restore.

The bottom line is that while IT must focus on overcoming their growing data storage, bandwidth and analysis requirements, the cost of not providing equal remediation to backup and recovery can be dear indeed.

Image credit: joy garnett (flickr)

Networks Are There To Serve The Applications

By: Marc Goodman at Silver-Peak Systems

Waiter_TrayWhen I was younger I waited tables at a neighborhood restaurant. When I first met with the general manager, I promised to work hard to make the customer’s dining experience exceptional. I knew the difference between a good dining experience and a bad one.

Little did I know how difficult it would be to deliver on that promise. The reasons for a bad experience were endless. Maybe I forgot to bring something to the table. It could have been the meal was undercooked, or the customer was just in a bad mood.   Every time a customer left unhappy, even when it wasn’t my fault, I always felt responsible.

I’m sure IT administrators feel the same way when they get complaints about slow and unreliable data center and network performance. Well, there is new technology that shows signs of offering relief.

Building networks that truly serve the applications has always been a dream of data center managers. And now, with the elasticity and open architecture that Software-Defined Networking (SDN) promises, network resources can be controlled and managed specifically with the applications in mind. Provisioning and managing today’s network services is a hefty undertaking, and plenty of human errors are made when manually managing groups of diverse equipment. One objective in the alleviation of these problems is to have physical network infrastructure become less of a management challenge and become more of a seamless, integrated building block within virtual infrastructure adaptable to changing workloads.

The promise of SDN for enterprises, cloud operators, and service providers is to build networks that effectively serve their applications, allow them to manage their networks to support diverse workloads, and to ensure compute, storage, and network resources are always available for all their applications.

For the most part, applications are transferred over general-purpose, one-size-fits-all networks that are hard to control and manage. Unfortunately, one-size-fits-all doesn’t work very well, as some applications are mission-critical, while others are not; and some applications are highly sensitive to latency, and others aren’t, and so on.

Rather than deploying applications over traditional networks, with no ability to automatically address dynamically changing workloads, SDN programmability promises to enable the automatic and intelligent appropriation of virtual and physical IT infrastructure resources to meet the needs of all applications.

Having a network architecture that has an understanding of application workload needs, and dynamically allocating IT infrastructure capacity to meet the workload requirements is a huge leap in network service. This capability can truly enable IT administrators to deliver a positive and productive user experience for any and all of their dynamic and ever-changing applications.

“Big Data” or just “a lot of data”?

By: Clive Longbottom at Silver-Peak Systems

petabyteOver the past few years, there have been the bandwagons – full of shiny promises, covered in good words and often totally confused when it comes to how they have been messaged.  From virtualization through cloud and now onto Big Data – end users just begin to start to understand one when the next one rolls into town.

And Big Data certainly seems to be confusing many – the two words themselves may not be helping, because it is not really about “data”, as it is more about information and intellectual property which come in many different formats, and it certainly isn’t about “big”, as analyzing even small amounts of the right sort of information needs a change in approach in how the underlying assets are dealt with.

Don’t get me wrong.  Big data is an issue – but not necessarily as many commentators see it.  For many, big data is all about how much data there is to be dealt with.  This is not (necessarily) a big data issue – it is a lot of data issue.  OK – I can see the brows furrowing as you read this – what is this guy up to arguing semantics?

Look at this way – if we take an organization such as CERN in Switzerland, they have databases containing petabytes of data collected from massive experiments like the large hadron collider [that seeks to unravel the origins of the universe].  Note the key here – these are databases, not datasets (semantics again).

A database can be dealt with through simple evolution of database technology combined with data analytics and reporting tools – maybe using in-memory or flash-disk based approaches to speed things up, and using more visual tools to make the analytics easier for the end user.  By contrast, a data set is a collection of all the different data sources that I may need to investigate, aggregate, collate, analyze and report on to provide the information I need.

Let’s take a company like British Pathé as an example – it has massive archives of text, image, sound and video.  Sure, it could push all of these into a standard database as binary large objects (BLOBs) and apply SQL queries against them, but it wouldn’t be able to effectively analyze what it has got here.

Enter Big Data in its true form.  This variety of data is a key aspect, requiring a different approach to how the data is dealt with.  For example, although much of the data will be perceived to be unstructured, it will actually be underpinned by structure, such as XML or even just as a comma or tab delimited file.  Voice can be machine translated into text, making analytics so much easier.  Video can be parsed using color, shape and texture analysis so that metadata identifying scenes can be added to the files.

The use of emerging systems, such as Hadoop and noSQL databases such as MungoDB or Cassandra can also help.  Hadoop’s MapReduce can help in dealing with a large data problem in the un- or semi-structured space by helping to reduce the volumes of data under examination, and can also help in deciding whether a particular piece of data is best dealt with under a noSQL or a real SQL database, each having its own strengths and weaknesses.  But, once the multiple different types of data are under control, analysis and reporting become so much easier.

Big data can make a world of difference in how an organization deals with all the sources of information available to it.  Just don’t fall for the glib sales talk from those vendors with too much vested interest – if a SQL vendor says they are all you need, or you hear the same from a noSQL vendor, show them the door.  A hybrid solution; one that includes Hadoop to help in refining the less structured stuff is the way forwards.

Image source: flickr (scjody)

Confusion over Consolidation of Storage and Consolidation of Data

By: Hu Yoshida at Hitachi Data Systems

Randy Kerns, a blogger on Storage Soup and also a Senior Strategist at Evaluator group, recently blogged about Confusion Over Storage Consolidation.

Randy says that “the confusion comes about because of some vendor messaging and what IT storage professionals actually view as storage consolidation. This leads to miscommunication and different sets of expectations about storage optimization projects.”

While he agrees that the simplest form of consolidation is to reduce the number of boxes on the floor, he does not believe that storage consolidation means one storage system for all purposes. He believes that there are legitimate reasons why IT operations end up with multiple systems over time.

These reasons include:

  • Projects come with budget specifically to purchase new storage for that project
  • Mergers and acquisitions bring in disparate storage systems
  • Purchase of additional systems instead of expanding existing storage systems due to impact on overall performance or access density
  • Impractical to reset depreciation schedules when adding to capacity
  • Less valuable data should be stored on less expensive, lower performance storage systems
  • Storage systems are transient and need to be replaced after four or five years.

While Randy acknowledges that tiered storage allows for greater consolidation by managing the variations in storage requirements, he does not believe that the “single box for everything” concept is practical. I would agree with Randy on all these points if you were not using storage virtualization.

Hitachi Data Systems strategy for storage has been “one platform for all data” since the introduction of USP in 2004, which was the first storage system that could virtualize external storage from different vendors. Note that this is not one platform for all storage. We expect that customers will have a mix of storage systems for all the reasons that Randy has stated.  With storage virtualization IT operations can use different storage systems but the data can be consolidated into a common pool of storage resources. And when attached to VSP or its predecessors, USP and USP V, all the resources are immediately upgraded with the enterprise function and performance of an enterprise storage system. With storage virtualization there is no need to rip and replace storage systems and no need to reset depreciation schedules in order to consolidate storage.

The current VSP virtualization engine is a powerful multiprocessor, which can nondisruptively scale-up by adding cache, processor and port modules to increase and balance performance. Unlike systems which can only tier storage across internal pools of storage, VSP can dynamically tier across external storage systems and extend the life of existing storage assets. Storage systems are transient and do need to be replaced, but storage virtualization from Hitachi makes this seamless. Even VSP itself or its predecessors eventually need to be refreshed, but this can be done without disruption to the application with nondisruptive migration services.

The confusion over consolidation of storage versus consolidation of data comes about when one fails to recognize the difference that comes with storage control unit based virtualization.

Virtualization, Software, and Cloud Driving 2013 WANOp Market

By: Steve Wexler at Silver-Peak Systems

Having performed better than otheChalkboard Chart SP Blog Postr networking segments during the ongoing economic crisis, the prospects for the WAN Optimization market continue to be bright, according to a new report from TechNavio Insights. The WANop market is growing rapidly, and is estimated to continue with the same growth in the near future, albeit at a slower rate, as enterprise IT organizations may have restraints on IT budgets and expenditures.

The report says one of the major drivers is the transferring of high-memory content — i.e. real-time video conferencing, video surveillance, and high-definition content — over the network. Another driver is the need to ensure smooth flow of such data without any glitches.

Another recent report, from TRAC Research, found that new adoption of WAN optimization technology is being driven by its effectiveness in supporting major IT projects, such as delivery of applications to mobile users, cloud computing, and virtualization. In addition, organizations are increasingly using the quality of user experience as the key metric for evaluating their WAN performance.

The TRAC Research report provides a current snapshot of just how WANop is deployed, and a glimpse of plans for the future, confirming that even though WAN optimization solutions are currently being predominantly delivered as hardware appliances, organizations are becoming more interested in new deployment methods such as software-based or cloud-based services. According to the report, 64% are currently using hardware appliances, and 26% plan to deploy them. Only 25% currently use virtual appliances, but 42% plan to do so, while 26% use software clients, and 39% plan to deploy them. From a services perspective, 19% use a managed service provided by a telecom carrier, while 17% plan to deploy, and 8% currently use a cloud service, but 38% plan to take to the clouds.

“Moving infrastructure towards the hosted environment of the Cloud and looking towards Software-Defined Networking is resulting in a battle for supremacy in data center. It is also causing unique requirements and solutions across the customers,” said Alan Weckel, Vice President at Dell’Oro Group.

Although virtual appliances in the data center only accounted for 7% of total sales, they delivered 37% of the revenue growth in 2Q12, according to Dell’Oro. SDN is also a relative newcomer to the enterprise, with only 16% of the market planning to implement some form during the next 12 months, but that number will grow to 40% over the next 24 months.

It would appear that 2013 will see the continuing evolution from hardware-based WANop to less-expensive but more flexible and agile solutions.

Image credit: anthonycz (123RF Stock Photo)