TCO issues in disk technologies.
To calculate TCO, many people think that it's simply combining the purchase price and the cost of a multi-year support contract. Unfortunately, there is much more to it. So what does figuring TCO entail? There are several areas that should be included when determining TCO.
The initial cost of a storage purchase sounds pretty straightforward; what is the cost of the disk? But today's storage solutions are far more complex than just disk. Budgets need to include hardware, software, services and support, as well as "environmentals."
Hardware: Calculating the cost of the actual storage is usually the easiest part of the TCO equation. Each server will need at least one HBA, as well as cables, etc. These may seem minor, but all these extras can easily add 10 to 20% onto a quote (depending on the amount of servers involved).
When purchasing switches, plan ahead. You may only need 12 ports today; but buying two 8-port switches, while very cheap, is not a smart idea if there are any plans for growth. Consider 16-port switches, or investigate switches that allow growth by adding blades of ports (4 or 8 at a time). Buying more initially may lower the cost of the hardware, and it may also lower the overall TCO of the solution.
Software: Five years ago, most storage systems didn't come with much software. Typically, there was a management tool and maybe a monitoring piece of software. Today, options like snapshots (pointer and full copy), replication (asynchronous and synchronous), HBA failover, and host agents, among others, are very common. Some of these options may seem expensive at first, but when the business impact is considered, it's a small price to pay. For example, as I mentioned, snapshot technology may add 10 to 20% to the initial cost of the storage purchase. But if implementing snapshots can be used to delay or eliminate the need to replace an existing tape library, then that extra cost is well spent.
Also, understand how storage software licenses work: some are tiered based on capacity, and some are tiered based on the number of servers, while others are not even tiered.
Services: Storage services range from installation to data migration to database tuning. Be sure to understand what is being purchased. "Installation" means something different to everyone. To the storage provider, it may mean just plugging the unit in and verifying that it works. The buyer may expect the works: configuration, host setup, etc. The difference in price can be extreme.
Request a detailed statement of work as part of the proposal and verify that your requirements are met. A poorly implemented solution will have a negative impact on the business and potentially raise the TCO.
Support: Storage solutions' support and service can cost a lot of money. No matter how much is spent, it will need to be serviced. Be certain to purchase the level of support necessary to protect your information, as well as its availability.
If the storage is being purchased for a 24X7 production environment, it should have 24X7 support with at least a four-hour response time. Keep in mind that if your system fails overnight and all you have is an "8-to-5" service contract with a four-hour response time, contractually, the service rep doesn't have to arrive on site until noon. That means your production environment could be down for a long time.
Also verify what level of software support is purchased. Some storage providers supply 90 days of software support for free. Then, on day 91, a bill shows up to cover the remaining year of support. All negotiation leverage is gone. Hardware and software contracts should be purchased up front for the expected life of the solution (for example, a three-year lease would call for a three-year support contract).
Environmentals: Many storage devices have specialized environmental requirements. Sometimes these requirements are overlooked and end up being a major cost. Will the existing air conditioning system support the addition of the new storage system? Is the correct power in place? Will the UPS or generator need to be upgraded? Can the floor support the additional weight of the unit? If the unit is oversized, will it fit in the elevator and through the datacenter doors? There have been occasions where a roof/ceiling has been opened and a crane used to lower a storage unit into a datacenter.
Once the solution is installed, configured and running, the costs listed above may continue to rise.
Hardware: Adding capacity to a storage system usually consists of purchasing disks and a tray to hold them. Additional servers will require additional HBAs, cables and more switch ports. Once the initial switch ports are full, some switches allow for blades of ports to be added. When these are all used, a decision needs to be made on whether to add more switches, which creates a more complex environment (possibly requiring a SAN evaluation to determine how to proceed) or replace the existing switches with larger ones, which is more expensive.
Software: If more storage is added, the solution may enter a higher tier of support. As servers are added, solutions tiered on server may go to the next tier. These can increase costs dramatically. If server resident software is required for the solution to operate correctly, that should be accounted for as a host is added. If the decision was made to do some of the advanced functionality with host-based software (snapshots, replication, etc.) then those costs should be accounted for. When a new application is added that requires software that was not purchased with the original solution, the cost of the software and services to implement may easily cost more than the server itself.
Services: Is the staff trained on the software that was purchased? If not, professional services will be needed when additional storage is added to the solution. This cost is typically not high (to install and configure it), however, if advanced functionality is used on this new capacity (again, snapshot, replication, etc.) the cost of professional services can be significant.
If servers are added, the storage will need to be configured for the hosts. If the new servers need advanced functionality configured (with either host-based software or within the storage device), again, the cost of professional services can be excessive.
Support: Once the original support contracts run out (storage, switches, software, etc.), additional contracts will need to be negotiated. Sometimes it's more cost effective to replace a solution with new technology than to pay for new support contracts. It is very important to negotiate support contracts up front for the life of the equipment. If leasing, the support should cover the term of the lease. Not-to-exceed pricing should be obtained at the time of purchase to cover one or two extra years in case unforeseen problems should arise. These numbers can be used as part of the TCO calculation.
Environmentals: Many companies have to pay for floor space (either within the companies datacenter or at an offsite location). This cost is usually fixed and billed monthly and will include air conditioning and "power grid" (generators, UPS, etc.) expenses. As the solution grows, pricing should be very straightforward; however, if the storage purchaser is also responsible for the datacenter "environments," then they must calculate ongoing maintenance If the solution grows, the existing infrastructure may need to be upgraded, which is all part of the TCO.
The cost associated with the impact to the business of the storage solution is more difficult to calculate. Decisions made at purchase time may lead to a negative impact on the business.
Availability: Servers have one or two HBAs. A server with one HBA has a single point of failure. If that HBA or cable fails, the server is down and the productivity of the application on that server is lost. Switches can be purchased in redundant pairs or as a single switch. A single switch, however, is a single point of failure for all servers attached to it.
Some storage systems have only one RAID controller. Like a single switch, this is a single point of failure for the solution. The money saved by not buying the proper redundancy in the solution can quickly be lost with an extended outage.
Functionality: Initially, advanced features on a new solution are not always a must, but they can be added later. For example, implementing snapshot or replication software (for advanced backups or disaster recovery) can greatly increase the availability of the servers attached to the storage.
Scalability: Avoid purchasing a "maxed-out" solution. If there is no room for additional hosts (no available switch ports) or disks (the storage device is full) then the first time something is added to solution the cost will be extreme. Typically, buying the larger switch or storage device up front is far more cost effective than upgrading the system later.
Performance: It's much better to buy more performance than needed than to buy too little. A system that bogs down during peak processing times will bring applications to a crawl. When calculating capacity, remember that all disk capacities are not equal. 1.5TB of 146GB 10,000 RPM disks (approximately 10) will not perform nearly as well as 1.5 TB of 36GB 15,000 RPM disks (approximately 40). There is a balance between raw capacity, performance and physical space in the solution.
While a storage system can be a significant investment, proper planning can minimize the ongoing business costs. On the other hand, what may appear to be a cost-effective solution up front may end up costing the business a significant amount of money in the long run. Fortunately, much or all of the costs can be offset by the aggressive ROI that today's storage solutions can offer.
Jim McKinstry is senior systems engineer, Engenio Information Technologies, Inc. (Wichita, KS)
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||TCO: Disk Arrays; Total cost of ownership|
|Publication:||Computer Technology Review|
|Date:||Jul 1, 2004|
|Previous Article:||2nd annual TTC roundtable.|
|Next Article:||TCO analysis: where D2D fits--part 1.|