A common mistake I have seen, and continue to hear about, is someone forcing an application to run on a slower and less than desirable storage configuration to meet cost requirements. Buying storage on a dollar-per-capacity basis, instead of looking at capacity, performance, protection and other service characteristics, can result in sub-par performance and misconfigured storage. When making a purchase, look at the level of service (availability, security, protection) and performance required to meet your needs. Effective storage acquisition requires a clear understanding of your applications'
It is also important to consider the following characteristics:
- I/O performance (I/O vs. bandwidth, read vs. write, random vs. sequential)
- Items such as large or small I/Os, online or batch and daily vs. seasonal workloads
- Data protection (RAID level, local and remote replication, high availability, business continuance and disaster recovery)
- Storage capacity (raw, formatted, usable), availability and response time needs
- Data retention (archive), recovery time objective (RTO) and recovery point objective (RPO) along with compliance requirements
Storage costs include acquisition, installation, maintenance, configuration, allocation, environmental (power, cooling, ventilation, floor space), security and data protection. Demands on storage resources include capacity (total raw, total usable), performance, data retention and availability. Consider what percentage of your storage capacity you want or need to allow for free space to meet different service and performance requirements, as well as for on-demand allocation and unplanned growth.
Storage systems differ in their ability to support dynamic storage optimization and configuration. Some vendors such as Pillar support quality of service policy-based placement of data on storage to meet different performance, capacity and cost objectives. Other vendors, including 3PAR, leverage thin provisioning and capacity-on-demand dynamic allocation to optimize costs and storage service delivery.
Many vendors including Engenio and HP support some form of dynamic volume or LUN expansion, and most NAS vendors supporting some form of dynamic file system expansion. Vendors differ in their implementation and functionality when it comes to dynamic allocation and optimization, with some vendors supporting "on-the-fly" RAID level migration. This relatively new feature gives users the ability to migrate data from one type of LUN or volume to another while in use. An example is migrating data from RAID-0+1 to RAID-5 while the data is being accessed. The advantage of migrating from one RAID level to another is the ability to lower your storage cost for a given amount of usable storage capacity.
Some of the clustered and peer-based storage vendors (or what some call grid-based storage) products load balance data across available storage capacity on different nodes. Examples include iSCSI vendors EqualLogic and Lefthand Networks, among others.
You can avoid surprises like not having enough storage, or the wrong type of storage, after trying to buy a product that is more than your budget allows. You risk compromising capacity by buying higher capacity disk drives instead of high-performance disk drives. Another example would be configuring storage for RAID-5 to reduce the number of disk drives purchased when your applications needs RAID-0+1 for performance. Figure out what your needs for point-in-time (PIT) copies, snapshots, remote mirroring and backup will be. Avoid being surprised with PIT copies and snapshots by understanding if a particular product is a full- or pointer-based copy and how that meets your RTO and RPO requirements.
If you have not yet prepared a storage capacity forecast, take the time and put one together. You can learn about storage performance and capacity planning in Chapter 10, "Performance and Capacity Planning for Storage Networks" in the book "Resilient Storage Networks" (Elsevier). A storage capacity forecast is important to help balance costs vs. proper allocation in order to avoid being caught off guard and having to allocate storage in a less than effective manner.
Do you know...
About the author: Greg Schulz is founder and senior analyst with the IT infrastructure analyst and consulting firm StorageIO. Greg is also the author and illustrator of "Resilient Storage Networks" (Elsevier) and has contributed material to "Storage" magazine and other TechTarget venues.
This was first published in March 2006