BACKGROUND IMAGE: iSTOCK/GETTY IMAGES
All-flash storage arrays can provide much higher levels of performance than disk-based arrays, and with deduplication...
and compression, they can sometimes even approach the cost per gigabyte of disk-based arrays for certain applications.
They are not, however, the right fit for all storage needs. A small business only running file and print services that buys an all-flash array just bought a Ferrari for grocery shopping.
To get the most out of the expenditure, you need to determine which apps need the speed and power of all-flash. Are you going to start a new storage silo or do you plan to integrate all-flash storage into your existing infrastructure? Do you want all-flash from the same vendor that makes your existing storage?
You also need to determine your overall budget; look at hybrid scenarios and try to figure out the ongoing maintenance or warranty costs.
What apps do you want to accelerate?
There are many applications that can benefit from the high-speed (GBps), low-latency (sub-millisecond response times), high-IOPS and random seek capabilities of all-flash. These include real-time applications, video streaming, virtual desktop infrastructure (VDI), databases, big data apps and search engines. But not all of these apps have the same needs. For instance, VDI tends to consume a lot of IOPS, while video streaming must sustain multiple gigabit streams, and search engines need low random seek times.
In addition to characterizing the types of traffic, you will want to look at whether it is relatively steady state or burst-heavy traffic, which peaks at certain times of the day and generates relatively little use the rest of the time. For example, VDI applications typically burst to high levels of traffic at the beginning of each shift, when users start their virtual desktops, and also when users save their work before going on break. This can create storms of traffic, which can happen when updates to the operating system or client software are pushed out to each instance. By scheduling other types of jobs, such as database indexing when VDI demands are lower, you can accommodate multiple types of apps without hitting 100% utilization.
Backups normally generate their traffic during off hours, but jobs that take too long to complete can continue to generate traffic throughout the business day. Some businesses use all-flash array storage to ensure that backups complete during the off-peak window, even though backups are not usually thought of as the kind of high-priority traffic that requires all-flash storage arrays.
Integrate with existing infrastructure or do a forklift upgrade?
By retaining your old storage systems and using them for second- or third-tier applications, you can take your new all-flash array and, with the help of a data management system that supports tiering, automatically move data onto the most appropriate storage as needs change. Some all-flash storage arrays come with built-in automated tiering, while others may require the use of a third-party appliance or software package.
Do you know what you are getting for your money?
There are several factors that determine the cost of an all-flash array: the initial cost of the hardware; the cost of ongoing support; costs for added features such as tiering, deduplication, compression, snapshots and so on; and the cost of hardware monitoring and replacement. In many cases, the cost of ongoing support includes updates to the controller software and monitoring of the system, which may be billed as a percentage of the initial cost of the system.
Some companies include all the additional features in the initial cost of the system, while others have an a la carte purchasing method that allows you to buy only those features that you will use. Hardware support is usually an add-on with differing levels of cost; for instance, for 8/5 support with a 24-hour response time or 24/7 support with a three-hour response time.
Depending on the cost of the maintenance program, you might want to consider buying your own spare drives, controllers, cache cards and other parts, and make do with a lower level of support, rather than pay two or three times as much for support with a quick response time.
Not all budget items are easy to compare between products. For example, one company might have a software package that includes deduplication, while another might present the same capability as an add-on module. The add-on module might seem expensive, but given that it can add two to three times the effective capacity to a system, it might be worthwhile.
Snapshotting can produce big benefits in some backup strategies, allowing a backup to be run from a snapshot that takes only a few seconds to create, instead of requiring that a database be taken offline for the duration of the backup. Expensive add-on features can save you money if you are able to take advantage of the new capabilities.
In general, you get what you pay for. If you are willing to put a system together yourself, install your own software, debug any issues and do your own updates, you can use commodity hardware and off-the-shelf software to create your own array. If you have adequate staff with the right kind of expertise, you could save a lot of money over using a turnkey system from a major vendor. However, you will have to source your own spare SSDs, install the software updates yourself and possibly deal with several different companies in the event of a hardware failure. Either way, the initial cost of the hardware is only part of the total cost of ownership.
Considering a hybrid scenario?
In most enterprises, there are applications that will not benefit from all-flash storage. A user saving a Word document to a network share probably won't notice if the operation takes a millisecond or a microsecond.
While it may make sense to consolidate all storage on a new, all-flash array, you might want to consider using storage management software to combine existing HDD-based storage with the new system, or even buying a hybrid, rather than all-flash, system. This is another case where the manufacturer's storage management software might be a bargain, since it would allow you to prioritize traffic by type, user or application to ensure that the more expensive flash is used when necessary, and that low-priority traffic is sent to the slower, disk-based systems.
Third-party software, such as DataCore's SANsymphony, can also perform this function, and will allow you to integrate all-flash storage arrays from different manufacturers if you have existing storage from which you want to get value.
Warranties and maintenance subscriptions
The issue of warranties and maintenance subscriptions is a complex one. There are two issues: the expected or guaranteed lifespan of the SSDs and the period for which the product is actually warrantied.
SSD lifespans have been a concern for users since they hit the scene 15 years ago. However, lifespans have improved dramatically over the years, and the truth is that, in general, SSDs will last longer than the systems on which they are installed will be kept online. With new 2 TB to 16 TB drives available, it is not worth continuing to pay maintenance and support on the 128 GB SSDs from 10 years ago. Likewise, SSD performance continues to improve at a rate that is rendering old SSDs obsolete on a five year or shorter cycle.
Perpetual free replacement programs are about a guaranteed revenue stream for the manufacturer. Vendors lock in customers with a guaranteed upgrade policy and amortize the cost over three to five years of support fees, allowing them to give the new system to the customer. As long as the vendor does not have to replace the SSDs, due to failures, earlier than was agreed upon in the contract, it will make money on the deal.
Thin provisioning is an essential feature for all-flash storage arrays
How to vet an AFA storage vendor
You may need all-flash storage and not even know it