Measuring for Change(with Storage)
How do we prepare for change in our IT infrastructure?
In my previous article I discussed how IT professionals must constantly be prepared to change what works for what will work better. There is no avoiding progress. You can only ignore it, and then pay the high consequences for doing so. Change is inevitable.
But acknowledging the possibility of change, while crucial, is just the first step.You have to prepare for change in the right way, by proactively making beneficial changes yourself instead of just waiting for changes to happen to you. You have to know how you want to change and how to measure your success (or failure) following a change. If you don’t, you could end up badly misleading yourself by falling into a trap that I’ll call the “Measurement Shell Game,” where you focus on the wrong details to measure and miss answers that may be right in front of you.
I could create an example for you, but the immortal classic Monty Python & the Holy Grail beat me to it (but only by about 40 years). During one scene the villagers bring a woman accused of being a witch to Sir Bedevere who takes the villagers through this perplexing and hilarious maze of pseudologic:
- Witches are burned, and wood is burned, so witches are made of wood.
- Wood floats in water, and ducks float in water, thus wood and ducks weigh the same.
- Since witches are made of wood, and wood weighs the same as a duck, then a witch also weighs the same as a duck.
- Which means that if a woman weighs the same as a duck she is a witch!
Of course in the film all of this silliness actually works when the witch is found to weigh exactly the same as a duck and the witch concedes, “it’s a fair cop.”
Unfortunately in IT, when we make these fundamental measuring errors, things don’t end up just working out. Projects fail, money is lost, and careers suffer (though at least we’re not burned at the stake–hopefully).
IT & the Measurement Shell Game
In IT everything comes down to the numbers. Is the performance fast enough? Was the SLA met? Without metrics we have no way of knowing if our IT solutions are meeting the needs of our clients and organizations. You cannot argue with numbers. Numbers do not lie after all.
Yet while numbers are always honest the way that humans interpret them is sometimes questionable. The raw capacity of a storage system may be accurate, but there are many needs that must be accounted for that will reduce the actual usable capacity. You have to consider other factors such as:
- how many drives will be used for RAID parity purposes
- how much capacity will the storage OS require
- how will data protection be accounted for
None of these are frivolous components, but each one may reduce the amount of usable capacity and performance that the storage solution provides. When you acquire storage you need to know exactly how it is going to meet your infrastructure’s needs. This is why customers need to be very precise in stating their requirements for a storage solution. A simple misunderstanding of what a measurement means coupled with how a solution is being described may lead to disastrous results.
To be blunt, I and most of my customers don’t have the time or energy to waste on deciphering how the solution is being measured. We want the solution measured and described in the terms of our needs. Don’t tell us how many disks we are getting and what the potential performance and capacity of all of those disks would be independent of the solution. Tell us exactly what the solution will provide to the infrastructure once it is is fully installed and configured.
The shell game with storage
The funny thing though is that the storage measurement shell game only works when the customer is willing to play along. There are ways to avoid the storage measurement shell game altogether. Start by having all quotes for storage explicitly state what the usable storage capacity is for the purchase. This means that all RAID, spares, and system capacity requirements are accounted for, and the higher raw capacity numbers are not the only numbers quoted.
If a storage capacity reduction technology is part of the solution, be sure to have a guarantee included in the purchase agreement. This guarantee should state in clear terms that a minimum usable capacity will be provided by the solution for a fixed amount of time based upon your current storage profile and expected growth. Do not get burned by “expected capacity” promises that fall short when it comes to data deduplication and compression. Deduplication and compression are both great technologies, but you and your vendor should be conservative when estimating how effective either technology will be with your data.
Finally, make sure that you know what your performance needs are. This is not just the number of IOPs that you need, but also what kind of IOPs you must account for, the latency associated with accessing and retrieving your data, and the bandwidth that you need for your storage workloads. A system that can generate a million IOPs might sound impressive in a sales brochure, but that same system is anything but impressive when applications suffer due to a storage network bottleneck despite having more than enough performance in the form of IOPs.
Coho Data does the measuring for you
Coho Data’s engineers understand the risks involved with getting a storage purchase wrong. That is why the Coho Data storage solution was designed to be purchased and expanded upon as needed using measurable components. Each Datastream 1000h chassis provides 26TB of usable capacity. You do not need to worry about RAID parity requirements, or the storage system’s operating system. Each DataStream chassis delivers 180K IOPs, this is based upon an 80% read/20% write random workload with 4K block size. The Coho solution leverages software defined networking implemented via an Arista 10Gbe switch for the storage network with has as 40Gbps connectivity to consistently deliver storage with less than 1 ms of latency.
All of these variables and more are measurable by using our built-in analytics. Through the intuitive GUI you can access these metrics to know exactly how any virtual machine is performing at the storage level. Some of the metrics available are:
- read and write IOPs produced
- latency of accessing the storage system
- storage capacity consumed
- throughput used on the storage network
Storage for the enterprise is changing, and at Coho Data we want it to keep changing for the better. Let’s stop playing shell games and instead focus on delivering measurable results. If you are interested in learning more contact us to arrange for a demo of the Coho Data solution. We will work with you to make sure that all of the variables are accounted for with your next storage purchase, so that you can continue to measure positive results.
8,124 total views, 2 views today