Cloud, Cache and Tool Talk
If you are planning to use cloud-based SSD resources, verify the effective performance, as well as any I/O limits and QoS optimizations, along with fees. Fees include not only capacity, but also I/O costs and how much capacity needed for a given level of performance.
Also with cloud SSDs, in addition to block, file and object services, there are also on-instance SSDs. For example, AWS EC2 (and other providers) has options for instances with on-board SSDs. These can be faster or have fewer limits than some of the storage services. However, there is catch: those on-board SSDs may not be persistent across instance start/stops. Double check your particular cloud provider’s details and best practices for using on-instance SSDs.
Tool trends include cache (read and read/write) along with micro-tiering that can be found in storage systems, appliances, hypervisors, virtual machines, operating systems and cloud instances. For example, VMware has cache capabilities in vSphere as well as vSAN. Also, Microsoft has added cache tiering with Windows Server 2016 combing Storage Spaces Direct (S2D) and ReFS cache. Want the functionality similar to Windows Server 2016, yet you are still on 2012 R2 or 2011? Then have a look at Enmotus FuzeDrive, which also runs on Linux and cloud instances.
Some general tips, recommendations and considerations include the following:
- Watch for near-term pricing fluctuations based on supply and demand.
- Work with your vendors to secure annual price reductions.
- Vendors will quote effective capacity as well as raw capacity.
- Effective capacity is what you get via data footprint reduction (DFR) or storage efficiency.
- Look for storage efficiency guarantees that will not impact your performance.
- Another technique is to use higher-capacity, lower-cost drives (SATA HDD dejavu?) .
- For lower-performing, less frequently updated data, higher-capacity, low-cost SSDs are ok.
- However for cache, tiering and primary update applications, use higher-performance SSDs.
SSDs and the Future
Will 2017 finally be the year of NVM and SSD with the all-flash data center? It is safe to say that as in the past, various types of NVM and SSDs will continue to become more common in your datacenter, as well as in everyday life. Moreover, yes, I am walking the talk using not only SSDs in laptops and tablets, as well as in my physical and Hyper-V OpenStack, as well as VMware servers with NVMe, SAS and SATA including devices Intel, Samsung and Seagate among other vendors.
NVM and SSDs of various types, form factors, interfaces, performance and capacity are in your future. The key questions are when, where to deploy them, what type to use, whose products to choose and how much can you afford (or afford not to) procure. Visit TheSSDPlace.com and TheNVMePlace.com to learn and view more about NVM, SSD and related technologies, topics, trends and performance and benchmark tools.
Greg Schulz is Founder of Server StorageIO, author of several books including Intel Recommended Reading List titles, and a new one in early 2017 “Software-Defined Data Infrastructure Essentials.” He is also a Microsoft MVP as well as VMware vSAN vExpert. Learn more at Storageio.com and Storageioblog.com. Follow him on Twitter @StorageIO.
Photo courtesy of Shutterstock.