Automate storage with data stratification technology

Network administration - Smart software can automatically distribute data on the most efficient forms of storage.

His difficulty was how to correctly identify the remaining 94% of unused files that he could transfer at any time to slower and less expensive Serial ATA drives. He finally got the software he needed, Dynamic Storage Technology (DST) - part of Novell Open Enterprise Server 2 - to create files and automatically implement migration policies. file on the basis when the last file has been accessed.

After a week of tweaking the configuration Glenn said: " It gave a great result, " he released at least a dozen Fiber Channel drives. By reducing the number of active files, he also cut his daily backup time from 14 hours to 47 minutes.

The installation is very simple, and the configuration requires just moving old LUNs to SATA hard drives, renaming LUN, creating a smaller LUN on Fiber Channel to replace it, specifying new LUNs to do. Primary volume ( primary ) and old LUN as a shadow . " Then I started setting up migration rules ," Glenn said. There is no additional cost for DST, he added, but it is estimated to save $ 140,000 thanks to reduced requirements for drives and power.

Glenn is one of the first beneficiaries of a new technology called automatic data stratification , which automates not only data movement but also the task of monitoring data being used like that. and determine which data needs to be included in the archive format. But automated stratification has not yet become popular because there are only a few distributions that offer this technology and have not been proven to be able to work in advanced high-end trading environments. It is usually only used in retail or file systems or supports only a limited number of storage protocols or topologies. But for organizations with simple needs, the current automated tiering tool is good enough.

How the stratification becomes automatic

" Tiering " means performing data migration between different types of storage media when the demand for it increases or decreases. Move old data, get less access to cheaper storage devices, slower speeds like SATA drives or even tape to reduce the cost of hardware while placing new data , accessed more often, more importantly on more expensive Fiber Channel drives, faster speeds even solid-state drives (SSDs) for increased performance. Finally, the automation process prevents it from bogging down in data classification and establishing policies that effectively influence "stratification" such as life-cycle management (SILM). ).

Preparation, Installation, Implementation?

Do you think your organization is ready to exploit the benefits of automated data tiering technologies? First consider these issues first:

- Does it provide the combined block and file level hierarchy you need?
- Is it possible to override the automated cascading process for performance reasons or data recovery?
- Does it support the features of backup or de-duplication if you are using them?
- Does it or will it support Sub-LUN stratification?
- Do distributors provide a path for further automation?

Storage administrators can move data between layers, but they must perform the process manually at least in classifying data and creating pre-stratified policies. Although policy creation is still needed, the latest series of automated products have been designed to reduce or eliminate the need for staff to monitor storage and search systems. Specific files, so the volumes and data blocks need to be re-leveled and moved by hand.

IT managers must first consider the criteria that software can perform (such as how often data is accessed) and whether it can evaluate and move blocks or files. independent instead of just moving larger volumes or LUNs. Since at least 10% of a block in a volume can be activated enough to justify the need to move to a more expensive, faster storage device, you will save money if possible. move these blocks, especially if you are moving to expensive SSDs.

See page 2

Other factors to consider include how software can quickly detect and respond to changes in data usage, and whether administrators can override automatic stratification if It affects application performance. Administrators can also use it to predict when certain data is needed (for example, accounting files for quarterly settlement), so tiered software can update it ahead of time. . Finally, administrators need to decide how they feel comfortable to give control to an automated tool.

While IT stores have struggled for years to implement ILM, some people who have used automated data stratification say they are still getting significant benefits with the software currently available.

Sandee Sprang, IT director for South Carolina's chief attorney's office, set up a regional storage network with automatic stratification using Compellent Technologies Inc's Data Progression software about five years ago, because she didn't have employees " to classify the type of records needed for the most efficient storage and quick access ". Setting up the Compellent system policies takes about four hours, and " the benefits it brings are completely sensible ," she said, saying that storage management time has decreased significantly from 24. hours a week down two hours a week.

Compellent's block level hierarchy also helps maximize disk usage, she added, and it " doesn't mean that the entire file is moved up and down the floors " - that " the move is just a copy." The resume you are visiting or a transcript from 15 years ago . "

Brian Nielsen, system technology architect at the Salk Institute's Computational Neurobiology Laboratory, works in a scientific computing environment with a large amount of variable workloads, thus needing real-time analysis and retiering provided by Network link storage devices of Avere Systems Inc. The first is just a trial, and finally buying specialized devices, he said, is a challenge for data transfer and the identification of what data to move.

Unlike previous ILM products, the data re-tiering is done sporadically and only based on the last access, the Avere system can " perform logging for multiple files. Calculate different I / O or dynamic [data] layers "when application requirements change, Nielsen said.

Brian Bosserman, network and network executive at Foster Pepper PLLC, a Seattle-based law firm, is experimenting with fully automated storage stratification technology (FAST) by EMC Corp. on the Celerra EMC NS-480 running at the company's offices in Spokane and Seattle. He estimates that it will save 10% of the time currently spent monitoring the server's storage needs, then planning and executing the re-stratification of virtual machines among them. With FAST, he said, he hopes to let EMC's Rainfinity File Management Appliance do monitoring and migration " based on the policies I provide it ".

Installing FAST "is very simple ," Bosserman said. " It's like a virtual [VMware] device. I just import FAST devices, start as a UNIX computer, then infiltrate it through the web interface and perform management, install it from there."

However, data tiering automatically requires some operations such as data classification and policy setting to determine when certain types of data need to be migrated (based on age of data, application performance, or legal and regulatory requirements. Some say that all of the above crippled previous stratification methods like ILM. But at least one major user - Intel Corp CIO Diane Bryant - is putting a formal ILM process in place where automated tiering is being studied before. Bryant began an effort with ILM late last year to cut Intel's annual 35% growth in essential storage needs, and about 40% of structured data and 30% of data The company's structure is currently managed by ILM.

Sanford Coker, head of the Unix division and senior Unix administrator at Weill Cornell College of Medicine, began using 3Par Inc's Policy Advisor in his test and development environment. Installation is easy, he said, and creating each policy takes about 30 minutes, although editing them to optimize performance takes a week or longer. He said that it is estimated that about 25% of Fiber Channel disk usage could be cut by moving data to higher-capacity, less expensive SATA disks.

What's next?

When complete, automated data tiering can help accept the use of SSD drives for data storage, because it will help administrators adjust the stratification enough to ensure that they are getting benefits. Maximum benefit with the highest performance from the most expensive storage media. But now, according to storage administrators, vendors and analysts, SSD drives are too expensive for most mainstream users.

Reichman says it is still cost effective in commercial space for performance thanks to "short-stroking" drives (this is the type of drive that only allows data to be written to external sectors of the disk surface. ) - as deliberately only using a portion of the disk space to improve their performance. Prices for free layering capabilities (for software already included in existing products) up to more than $ 50,000 for Avere FCN 2300 systems. Users, of course, are also a factor for the cost of classifying data and creating layered policies.

Major distributors such as EMC are also working to automatically classify data into "cognitive applications," meaning the software will understand I / O needs as well as other usage patterns of applications. Popular applications and automatically stratified to meet these needs. Therefore interoperability will require standards for information about the data being re-stratified. One of them is the metadata standard being developed by the Storage Networking Industry Association.