Storage Economics

Save it the smart way

As storage requirements grow and more and more organizations look at consolidating storage, there is now a need to look at some of the finer and more sensitive points around storage and storage economics. by Sanjay Lulla

There has been a paradigm shift in the storage landscape. Customers have moved from Direct Attached Storage (DAS) environments to SAN/NAS environments, from direct-attached tape to centralized automated tape libraries, and from native OS backup commands to intelligent backup software.

Many have moved from local clusters to continental clusters, and from tape shipping to intelligent storage based data replication solutions. With this kind of exponential growth in storage requirements, it is no longer surprising if a customer needs storage in multiple TBs.

Therefore, it becomes essential for us to look at a very important aspect of storage deployment—storage economics. We will now look at some of the best practices by which we can reduce our storage cost and improve ROI.

Storage decision

Here are some factors that need to be considered when you take a storage decision:

• Storage requirements in usable GB's/TB's after a particular RAID level
• Servers deployed on SAN
• Availability-clusters, no single point of failure
• Need for point in time copies

Storage requirements

An enterprise must consider its storage requirements based on usable GBs/TBs after a particular RAID level. This factor defines the basic storage requirement for a customer.

However, the calculation of storage requirements should be done carefully and accurately. Storage solutions available now have the ability to perform storage subsystems-based virtualization.

Virtualization at the storage system level is the disassociation of capacity from its underlying physical disk restrictions to create large pools of storage available to multi-vendor hosts. These pools of virtual capacity can be configured as virtual disks and presented to any or all connected hosts.

This implementation of virtualization allows more efficient use of storage capacity, simplified management, and an overall reduction in costs. A good thumb rule is to configure a mix of RAID 5 and RAID 1 for storage requirements. Thus instead of configuring everything in RAID 0+1, it would be a good idea to mix and match based on performance requirements. This further optimizes on the required usable capacity.

Servers deployed on SAN

SANs are usually capable of serving a lot of servers. For each additional server deployed on a SAN however, we have the added cost of dual HBAs, multi-pathing software, FC cables, and FC switch ports. So the question really is how many servers are actually required on a SAN? And answering it requires some careful planning. Ideally you could just deploy the most critical, performance-hungry and data-intensive servers on SAN.

Availability

Availability is high on mind recall for any customer, and so must never be sacrificed. It's always a good idea to deploy a SAN/NAS architecture in a No Single Point of Failure (NSPOF) configuration to the extent possible.

If there are requirements to necessitate an NSPOF you can always look at an architecture with a single switch, single HBA on each server, and a RAID5 on storage array. More customers today look at a mix of the above. So the basic storage infrastructure is NSPOF with some servers, which are not critical, and have been configured with a single HBA.

Point in time copies

Point In Time (PIT) copies have almost become an integral part of the storage landscape. These copies can be full copies—clones or partial copies like snapshots. While planning for PIT copies it is important to calculate the amount of primary data that needs to be copied.

So, it's not necessary that the total usable capacity required is twice the primary data. There are efficient implementations of PIT copies available which need up to 30 percent of reserved space for PIT copies. In fact we can also have near virtually capacity-free snapshots.

Fibre Attached Technology Adapted Disks

The Fibre Attached Technology Adapted (FATA) drives have dual ports, 2 Gb/s FC interfaces with 250GB of capacity. The dual port interface provides a level of reliability and availability that a serial interface cannot provide.

Today users can deploy an ATA disk for data protection, using ATA arrays as a target for backups in order to reduce the backup window and decrease recovery times. In addition, there is a trend to use ATA disk in tiered storage schemas, where more important data is stored on an FC disk, and 'less critical' data is stored on ATA drives.

The basic premise is that more important data should be stored on higher performing, highly reliable resources. Now vendors have begun to support both higher performing FC disks and lower cost ATA disks in the same array. In that way, the ATA drives can benefit from the performance, reliability, and operational features of the system, but still drive down overall costs.

The new hybrid FATA drives can be deployed and since the new drives will use the same FC front end, the drives can reside next to existing FC drives. Customers can buy the lower cost hybrid drives, place them in the spare slots, and begin to implement a tiered storage infrastructure.

Of course the FC front end is not the only difference between FATA drives and ATA drives. These hybrid drives have better performance and better data availability than ATA. The drive opens up new levels of flexibility for tiered storage schemas.

Who says there should only be two choices, FC or ATA? A third possibility, a FC hybrid disk, gives users a possible 'mid-tier' level of tiered storage.

Information Lifecycle Management (ILM)

ILM is a process that recognizes that information has a life cycle. It is the next step for the storage industry, and may change the paradigm again. ILM now links the storage utility to applications-and enables the control and intelligent management of information.

ILM controls the movement of data across storage technologies from high performing disks, to midrange disks systems, to tape, to optical. It enables intelligent index, and search and retrieval of relevant information capabilities to comply with business policies and government regulations.

The result is a combination of processes and technologies that determines how data flows through an environment at each stage in its lifecycle.

Typically these stages are:

• Create and modify
• Replicate and distribute
• Archive and recall
• Protect and recover
• Remove
  

Challenges

ILM addresses three key information management challenges for customers: Retention management, data management, and reference information management.

• Retention management - Corporate governance and increased government regulations in healthcare, financial services, and life sciences require strict retention policies and practices for achieving compliant records.
• Data management - Information growth is on the rise. The need is to manage data as cost-effectively as possible while maintaining service level objectives.
• Reference information management - IT needs to enable businesses to tap that pool of static or reference information to increase its value to the business value.

There are products available with the help of which data automatically gets migrated from expensive enterprise disks to more affordable disk-based reference information storage, thus reducing cost, while maintaining full data accessibility.

Conclusion

Today's business is operating in a very challenging environment. Companies are looking to quickly access and leverage information to make quick and accurate decisions.

This puts a lot of pressure on IT to be operational 24x7. Time is critical. Businesses want information faster—whether it is something a few days old, a few months old or even older. In addition, IT is not getting any additional resources—financial or human resources.

Many IT organizations are being asked to do more but with the same budget as last year. And surprisingly, a large number of IT organizations are being told to do more with low budgets and less resources. Storage economics is the inevitable one-stop-shop for such needs.

The author, Sanjay Lulla, is the Solutions Architect, Customer Solutions Group, Hewlett Packard India.