I have been getting asked by IT customers, VAR's and even vendors how much solid state device (SSD) storage is needed or should be installed to address IO performance needs to which my standard answer is it depends.
I also am also being asked if there is rule of thumb (RUT) of how much SSD you should have either in terms of the number of devices or a percentage; IMHO, the answer is it depends. Sure, there are different RUTs floating around based on different environments, applications, workloads however are they applicable to your needs.
What I would recommend is instead of focusing on percentages, RUTs, or other SWAG estimate’s or PIROMA calculations, look at your current environment and decide where the activity or issues are. If you know how many fast hard disk drives (HDD) are needed to get to a certain performance level and amount of used capacity that is a good starting point.
If you do not have that information, use tools from your server, storage or third-party provider to gain insight into your activity to help size SSD. Also if you have a database environment and are not familiar with the tools, talk with your DBA's to have them run some reports that show performance information the two of you can discuss to zero in hot spots or opportunity for SSD.
Keep in mind when looking at SSD what is that you are trying to address by installing SSD. For example, is there a specific or known performance bottleneck resulting in poor response time or latency or is there a general problem or perceived opportunity?
Is there a lack of bandwidth for large data transfers or is there a constraint on how many IO operations per second (e.g. IOPS) or transaction or activity that can be done in a given amount of time. In other words the more you know where or what the bottleneck is including if you can trace it back to a single file, object, database, database table or other item the closer you are to answering how much SSD you will need.
As an example if using third-party tools or those provided by SSD vendors or via other sources you decide that your IO bottleneck are database transaction logs and system paging files, then having enough SSD space capacity to fit those in part of the solution. However, what happens when you remove the first set of bottlenecks, what new ones will appear and will you have enough space capacity on your SSD to accommodate the next in line hot spot?
Keep in mind that you may want more SSD however what can you get budget approval to buy now without having more proof and a business case. Get some extra SSD space capacity to use for what you are confident can address other bottlenecks, or, enable new capabilities.
On other hand if you can only afford enough SSD to get started, make sure you also protect it. If you decide that two SSD devices (PCIe cache or target cards, drives or appliances) will take care of your performance and capacity needs, make sure to keep availability in mind. This means having extra SSD devices for RAID 1 mirroring, replication or other form of data protection and availability. Keep in mind that while traditional hard disk drive (HDD) storage is often gauged on cost per capacity, or dollar per GByte or dollar per TByte, with SSD measure its value on cost to performance. For example, how many IOPS, or response time improvement or bandwidth are obtained to meet your specific needs per dollar spent.
What is the best kind of IO? The one you do not have to do
Is SSD dead? No, however some vendors might be
Speaking of speeding up business with SSD storage
Has SSD put Hard Disk Drives (HDD's) On Endangered Species List?
Why SSD based arrays and storage appliances can be a good idea (Part I)
EMC VFCache respinning SSD and intelligent caching (Part I)
SSD options for Virtual (and Physical) Environments Part I: Spinning up to speed on SSD
Ok, nuff said for now