Having just recovered from a crippling recessionary year, 2010 seems to be a year of hope for India Inc. Today, with the lessons of 2009 still fresh in memory, Indian CIOs are demanding technologies that deliver better utilization levels, are more energy efficient and come with a lower CAPEX and TCO.
Over provisioning of storage is a problem faced by perhaps every Indian enterprise. It was the recession last year that brought into sharp focus the massive underutilization of storage infrastructure.
This has accelerated the push towards storage-efficient technologies such as deduplication, storage virtualization, cloud-based storage and thin provisioning. Alternate storage-efficient media such as Solid State Disks too are gaining prominence.
Let’s now look at why vendors and storage experts believe that each of the above technologies will be technologies to watch out for in 2010.

Trend #1
Solid State Drives (SSDs): Durability and performance make SSDs stand out
SSDs have been the latest entrants on the storage scenario. With their ability to provide high I/O performance, these storage devices can help organizations resolve storage performance challenges. For example, while capacities per spindle have been going up, the performance from a single drive has remained the same. This means that irrespective of whether it is a 36 GB drive or a 450 GB drive, the number of I/O generated operations remains the same. SSDs can help in correcting this anomaly due to their ability to handle higher I/O requests.
However, given their high performance, they would be required only for applications which require storage with high I/O processing capabilities. Their nature of use could be for dynamic storage rather than static storage.

“While a typical SAS or a SATA drive would offer up to 300 I/O operations per second, an SSD can provide up to 30,000 I/O operations per second. Hence its use would be currently limited to real-time, mission-critical applications,” says Narayanan B, Project Manager - Storage, American Megatrends India.
Till their price decreases, one can expect SSDs to be confined to transactions that require high performance. “While the technology issues related to SSD have more or less been resolved, the price point is still an issue for customers when it comes to large-scale storage. SSD adoption will be initially limited to the level 0 of a tiered architecture, with resolution of latency issues being the primary driver rather than power savings,” says Ramachandran Narayanaswamy, VP and Head, Network and Storage, MindTree.

Given this fact, SSDs can be expected to find a place in the level 1 of a tiered infrastructure which uses a high performance storage media for I/O intensive applications. Verticals such as BFSI, Telecom, Media, Research and Defense, GIS—where there is a high requirement for real-time, mission-critical, compute-intensive data—stand to benefit from the use of SSD. SSD could be used in core banking environments in banks, stock exchanges, for online transactions, and in telecom billing environments.
However, as SSDs gain acceptance in the enterprise, analysts expect their price to come down significantly. “Initially when SSDs were launched, a single SSD drive was 40 times more expensive than an equivalent fiber channel drive. However in terms of performance, it was 30 times faster. Today, we have found adequate volume of adoption for our SSD offerings, and have been able to bring down the costs from a 40x to an 8x level,” explains Sanjay Lulla, Director, Technology Solutions - India & SAARC at EMC Data Storage Systems.

Surajit Sen, Director Channels Marketing & Alliances NetApp is of the view that solid state storage is not just about SSDs. The company has developed an offering called PAM or Performance Acceleration Module. The storage device uses solid state technology, and is designed to offer solid state functionality at one tenth the cost of an SSD. The module acts as a cache between the permanent storage and the controller and can process I/O intensive data requests from applications.

“The module is a deduplication-aware cache. Assuming a 50 percent duplication of data, the module can store 8TB of data in an otherwise 4TB capacity with duplicate data. However, the module is not a permanent storage and would eventually be used in a tiered architecture for the most I/O-intensive applications. The module has algorithms built in, to automatically direct the high-performance application I/O towards itself,” explains Sen.

Going forward, while SSDs will be considered actively for high I/O-intensive transactions, they still have a long way to go before they can start replacing traditional enterprise storage options. “One should not deploy flash drives for every application. For some applications the benefit could be 30 times more than a fiber channel while for others it could be only 4 times higher. For applications where the benefit is only 4x, that cost will be prohibitive for the deployment of flash,” explains Lulla of EMC.

Trend #2
Tiered Storage: Aligning storage to suit information
As high-end storage is expensive, organizations are increasingly looking at adopting different storage options for different kinds of information. For example, important data which is frequently accessed may be stored on Fibre Channel or SSDs, while less important or less accessed data may be stored on tape or SATA drives. The objective of tiered storage is to intelligently align storage to suit information needs, rather than having all applications on a single type of storage media.

To handle this issue, applications with high performance but less storage requirements are mapped to a faster storage media located at the first level of the tiered architecture. A comparatively slower storage media is allocated at the second level for less I/O-intensive but higher data storage-intensive applications. Finally, a much slower but high capacity media for very high data storage applications such as backup and recovery is assigned at the third level of the architecture.
“Rather than buying a storage infrastructure, all of which is either on SSD, fiber channel or SATA, we can look at a tiered architecture where 10 percent of the storage environment is SSD, 70 percent is on fiber channel and the remaining on SATA,” says Lulla. With such an architecture, Lulla claims that customers will be able to acquire storage at costs up to 17 percent less than that of a full fiber channel based storage deployment.

However managing tiered storage manually is a fairly difficult process. Allocating different type of storage to different applications at different times can be a very complex, time-consuming and a cost-intensive task. To counter this issue, most vendors today are offering dynamic or automated tiering capabilities for tiered storage infrastructures. They believe that the adoption of these tools will be a major trend this year owing to the growing move towards tiered storage architectures.
“With dynamic provisioning and storage virtualization, we are able to implement dynamic tiering of storage. Dynamic tiering provides the capability to align storage resources with application data, and also facilitates non destructive mobility of data within the storage tiers,” says Vivekanand Venugopal, VP & GM, India, Hitachi Data Systems.

Venugopal provides the example of an ERP application. If the requirement of the application reaches 1TB in six months, it would not require the same amount of storage on day one. While the organization allocates 1TB logically, it has a pool of 1TB storage shared across 10 applications. This ensures maximum utilization.
An application may also require data copies for backup, recovery, archival etc. However a backup operation might not require high service levels and can be allocated a service level wherein the data is copied onto a slower storage media of the tiered storage infrastructure. That said, service levels for an application can change overtime. If an application requires data recovery—but only in situations where there is complete loss of data—the storage infrastructure must be able to move a copy of that application data from servers with standard service levels to higher service levels. This can be achieved with dynamic provisioning and tiering.
With dynamic tiering and dynamic provisioning, organizations can save on CAPEX and OPEX costs. Venugopal says that organizations can achieve up to 25 percent reduction in storage costs, using these technologies. He refers to the case of HDFC Bank which was able to reduce application migration time from 32 days. The bank was also able to increase its uptime by 372 hours using dynamic tiering and dynamic provisioning.