Cobb Electric Membership Corporation Makes the Transition to All Flash for Primary Storage

  • IDC has predicted one of the highest growth rates in enterprise storage for the AFA market through 2019.
  • Bland knew that additional server consolidation would place increasingly stringent demands on storage performance.
  • The Kaminario K2 is built around a storage architecture that combines elements of both scale-up and scale-out designs.

All-flash arrays (AFAs) have rapidly grown to become a major segment of the external enterprise storage market, generating $2.55 billion in revenue in 2015. Flash storage brings many benefits, including very low latencies and high throughput that is consistent with heavy and/or highly burstable workloads, reduced energy costs, increased storage density for effective storage capacity, and improved server efficiencies (due to its ability to drive significantly higher CPU utilization in application and database servers).

Because of the many benefits AFAs bring in terms of not only IT infrastructure but also business transformation, IDC has predicted one of the highest growth rates in enterprise storage for the AFA market through 2019, at which point AFA revenue will clearly dominate primary storage spend in the enterprise. In detail:

  • Customers like Cobb Electric Membership Corporation (more commonly known as Cobb EMC), a utility company whose legacy storage was reaching the end of its useful life, are increasingly looking at next-generation, flash-optimized storage solutions as replacements for their legacy infrastructure. The deployment model for AFAs has shifted in the past year from a dedicated storage platform for high-performance applications to a more general-purpose primary storage platform for mixed workload consolidation.
  • Kaminario, an AFA vendor, offers a platform well suited to dense, mixed workload consolidation, and customers of the vendor’s K2 array have used the system in this manner earlier than many other successful AFA vendors. Initially attracted to Kaminario by an aggressive cost structure that allows all flash to be deployed at an acquisition cost roughly comparable with that of hard disk drive (HDD)-based systems, Cobb EMC found that Kaminario’s next-generation storage architecture offered other benefits important to Cobb EMC in the areas of performance, scalability, ease of use, and footprint. Cobb EMC’s view of the ultimate inevitability of flash taking over primary storage is shared by close to 80% of enterprises, based on primary research conducted by IDC in the summer of 2015, although Cobb EMC is ahead of most enterprises in that its entire primary storage workload has already been successfully migrated to all flash.

This IDC Buyer Case Study reviews how Cobb EMC, a regional not-for-profit electric utility company serving the greater Atlanta metropolitan area, addressed evolving IT infrastructure requirements with flash storage technology to improve performance and substantially improve IT efficiencies. This Buyer Case Study explores what drove Cobb EMC’s initial interest in flash, how flash deployment has evolved in the company’s environments, how the deployment has impacted its business, and what future plans exist for AFAs at Cobb EMC.

Cobb EMC’s workloads are primarily customer care and billing, energy use analytics, enterprise resource planning (ERP) applications, and general office, facility, security, and environmental support applications — all of which are mission critical.

Challenges and Solution

Cobb EMC’s workloads are primarily customer care and billing, energy use analytics, enterprise resource planning (ERP) applications, and general office, facility, security, and environmental support applications — all of which are mission critical. These workloads include 40 Oracle databases, close to 90 SQL Server databases, and Microsoft Exchange and SharePoint, used internally for messaging and collaboration across approximately 500 end users.

An upgrade of its IBM AIX-based Power servers and replacement of its legacy Lenovo Windows Servers with HP Windows Servers provided significantly more compute power, helping clear the way for further consolidation and virtualization. Bland knew that additional server consolidation would place increasingly stringent demands on storage performance and wanted to complete the storage infrastructure upgrade with a solution that would keep pace with and complement the more robust servers.

For disaster recovery (DR) purposes, Cobb EMC’s workload is split across two active/active datacenters, both located in Marietta a couple of miles apart. Bidirectional replication is currently being tested in Cobb EMC’s environment, which, when fully implemented, will be used to maintain full data sets in both locations for very rapid recovery without any data loss, closing the loop on Cobb EMC’s disaster recovery and business continuity plans.

Cobb EMC’s legacy storage systems offered the broad set of mature data services needed for the type of dense workload consolidation necessary on general-purpose storage platforms. Snapshot and clone technology is used for data protection as well as test and development (test/dev) (which is colocated with Cobb EMC’s primary storage workloads), and Bland also plans to use synchronous replication and encryption in the future.

With the combined production, test/dev, and DR environments, Bland knew that Cobb EMC would initially need approximately 150TB of effective storage capacity, split across both sites, with the ability to accommodate an annual growth rate of 15–20% for the life of the new storage solution. In addition, Cobb EMC runs a lean IT shop, and storage management tasks are performed by personnel that have other administrative IT responsibilities also.

For this reason, Bland was as concerned about the ease of management as he was about the ease of expansion with the new system. The newer scale-out storage architectures offered advantages in these areas that he found very attractive.

Although his colleague’s comments about cost competitiveness are what initially prompted Bland to look more closely at Kaminario, other features of the solution provided a good fit with Cobb EMC’s solution requirements as well.

Results

Although his colleague’s comments about cost competitiveness are what initially prompted Bland to look more closely at Kaminario, other features of the solution provided a good fit with Cobb EMC’s solution requirements as well. First, the Kaminario K2 is built around a storage architecture that combines elements of both scale-up and scale-out designs, offering the best of both approaches.

While a single K2 node (called a K-Block) could provide in the range of 250,000 IOPS and as much as 92TB of raw storage capacity, the system can scale up to eight K-Blocks, all connected over a redundant, dedicated 10GbE network, to provide as much as 2 million IOPS and 737TB of raw storage capacity. This provides ample performance and capacity, particularly when taking Kaminario’s inline compression and deduplication into account, to meet Cobb EMC’s growth requirements over the next five years.

Both data and metadata are fully distributed across all system resources, a feature that allows K2 to deliver very predictable storage latencies even under heavy and burstable I/O workloads. Systems can be expanded online (either within a K-Block or through the inclusion of additional KBlocks), and the workload will be automatically redistributed across all resources, making expansion significantly faster and much easier than with legacy storage solutions.

Second, the K2 incorporates many features designed to maximize storage efficiencies. Powerful controllers anchor very efficient packaging in terms of both performance and storage density in only 6U. The 1.92TB solid state disks (SSDs) enable a storage density in the 16TB/U raw capacity range.

Kaminario’s inline data reduction doubles that density in “database only” environments and can more than quadruple density in mixed virtual workload environments. Additional storage efficiency features, such as thin provisioning, space-efficient snapshots and clones, and byte-aligned compression, help make even more efficient use of available storage capacity. Global adaptive block algorithms vary the block sizes used for I/O to more closely match the requirements of a mixed application environment, reducing the IOPS required to service any given workload.

Kaminario’s flash-optimized data protection algorithm, K-RAID, provides better protection than RAID 6, with a nominal capacity overhead of only 12.5%, and Kaminario’s consistently low latencies significantly increase CPU utilization on the server side. Increased CPU utilization, coupled with the server upgrades, allowed Cobb EMC to reduce its Oracle processor core count by 75% while still driving needed performance.

The reduction in core count allowed “substantial real dollar savings” in the Oracle Database license support costs. Cobb EMC has moved all its primary workloads to Kaminario, completely retiring its former legacy storage and reducing the rack space requirements by over 50% and energy consumption by almost 40%.

As a result, Cobb EMC’s storage infrastructure is “physically smaller and less complex” while providing much higher, more consistent storage performance and much more cost-effective expansion, with much lower operational and support costs.

Third, Bland and his team are finding they like many aspects of “storage 2.0” platforms, which include:

  • Improved reliability. Because AFAs have no moving parts, they deliver much better reliability than HDD-based systems.
  • Rapid provisioning. AFA platforms provide rapid storage provisioning that does not require an experienced administrator.
  • Better performance. AFA performance resulted in significant improvements in batch processing times. Month-end processing for its ERP system is now more than 30% faster. Applications and services boot up so quickly that Bland’s team had to field a few calls early on from end users stating that an application service couldn’t possibly be working right since it came up so quickly!
  • Higher availability. Systems are designed from the ground up to provide five-nines of availability, and telemetrics built into the management GUIs of many of these systems track installed base availability as a proof point. Kaminario can perform all microcode upgrades and physical expansion in a nondisruptive manner, and its telemetrics data indicates overall uptime across its entire installed base between five-nines and six-nines.
  • No “hidden” costs. All software is included with the initial cost of the system, making initial purchase and system expansion less expensive and easier.
  • Operational flexibility. The ability to apply certain data services, like encryption, at the volume level provides additional flexibility in consolidating mixed workloads.

Fourth, the Kaminario “cost” value proposition has proved itself in practice. Kaminario guaranteed 3:1 data reduction against Cobb EMC’s mixed workload. Bland tracks this through Kaminario’s GUI and is seeing data reduction ratios that are consistent with the Kaminario guarantee and his own expectations. The amount of money that Cobb EMC saved in Oracle licenses (due to the higher CPU utilization and migration to the IBM POWER8 systems) pretty much paid for the K2s.

Flash latencies enable the use of inline data reduction against very performance-sensitive primary applications, one of the key factors allowing Kaminario to deliver these systems at a price point competitive with that of 15K RPM HDD-based systems configured to provide similar performance and capacity. Cobb EMC initially installed two of these systems in production in January 2015 and in January 2016 purchased a substantial (100%) capacity upgrade utilizing the 1.92TB SSDs, with plans to implement replication once testing (which is targeted by the end of 1Q16) is completed.

ESSENTIAL GUIDANCE

Since the Kaminario systems were deployed to production in January 2015, Cobb EMC has migrated its entire primary storage workload, including production and test/dev environments, to Kaminario. As a result of the new IBM and HP servers and the Kaminario storage deployment, more than 90% of Cobb EMC’s workload has been consolidated and virtualized. Kaminario’s scale-out design has been important in providing what Bland sees as the key benefits of the Kaminario AFAs:

  • Consistently high performance, including both low latencies and high throughput
  • Easy and nondisruptive horizontal and vertical scalability
  • Lower cooling and energy costs

It is IDC’s view that an all-flash strategy for primary storage environments will happen for most enterprises within the next four to five years. Many enterprises like Cobb EMC are pretty much already there. Our market forecasts for the enterprise storage market predict that by 2019, 60–70% of all primary storage spend will be driven by AFAs. IDC also sees the industry moving more broadly toward the use of scale-out designs across both primary and secondary environments, although this migration occurs over a much longer time horizon. Vendors like Kaminario that already have a highly flashoptimized storage platform built around a scale-out architecture are offering what IDC feels will be the storage environment of the future.

More Details Visit TECH CLOUD LINK

Only $1/click

Submit Your Ad Here

techcloud link

Tech Cloud Link is the place to get free technology whitepapers downloads in a variety of formats, including PDF versions of popular articles tech briefs, tech whitepapers, and research articles into profoundly diverse spectrum within IT landscape. Here you will resolve trending IT concerns on topics like – Network Communication – Storage – Data Center – Server – Network Security. The whitepapers accurately address convergence between industrial and enterprise networks and collections of Articles, Features, Slide Shows and Analysis on Enterprise IT, Business and Leadership strategies that focus on critical
https://techcloudlink.com/

Leave a Reply