One of the things I like about Spiceworks is the website’s user forums. They give a good sense of what is going on in the trenches of IT. Recently a user asked a series of questions about Object Storage. It makes a great baseline for an Object Storage FAQ.
What amount of storage do you need to have to consider the switch to Object Storage?
Switching to object storage should not be about capacity. You can build very inexpensive, and open source (if that’s important to you), NAS systems. The advantages of object storage are high file count, rich metadata, geo-dispersion, and typically sophisticated search capabilities. The capabilities of object that most people focus on scale-out, high capacity, commodity hardware are not unique to object storage. Once you have a data set or two that can really use object storage then pulling the rest of the organization’s unstructured data into the object store probably makes sense. As we discuss in our article “NAS vs. Object: Performance – Thinly Provisioned NAS” its ideal use is as a vehicle to keep NAS “thin” so that only the most active data resides there.
Are there robust systems that are affordable for an SMB customer?
As usual, it depends on what you mean by SMB (small to medium sized business). The term covers such a wide range of businesses. To make things simple let’s assume we are talking about an SMB that is large enough to have a dedicated data center. If so then there are robust object storage systems that are ideal for this type of customer. There are several turnkey systems that are easy to set up and are accessible from a variety of protocols. The do-it-yourself (DIY) solutions do require some patience and expertise. The decision to use object storage is not an SMB vs. enterprise discussion, either can qualify. The decision to move to object storage is more about do you have data sets that make sense in an object storage environment – high file count, data generated from machines that can provide rich metadata (serial number, date, GPS location, weather condition, etc.
Should I use cloud or local object storage?
If you are storing data in the cloud you are probably storing it on object storage. The cloud brings another wrinkle though. You are renting your storage, which means a low entry price point, easier scaling (because you don’t own it) and upgrades are no longer your concern. The downside is you are renting capacity, which just like any rental, it gets more expensive than owning over the course of time. My colleague Joseph Ortiz compared cloud pricing to DIY pricing in his article “Tape vs Cloud for Archive and Cold Data“.
Pitfalls of object storage compared to traditional filesystem/block storage (DAS or SAN)?
Most if not all object storage systems provide NFS and SMB access and we think object storage can and will replace an organization’s NAS. Some object storage systems will allow object and NFS/SMB access to the same volume or bucket. This is a critical capability. It means you can feed an object store via a legacy system and then process that data through a modern application like Hadoop, Splunk or Spark. While object stores are not typically options for high performance storage, we are seeing a handful of vendors deliver all-flash versions of their object storage systems. But these systems are all-flash to speed analytics processing more so than to host legacy database architectures. One of the keys is to understand that all these gateway type of accesses are not created equal. Listen to our “StorageShort: The NFS to Object Storage Problem“.
How does this fit into one’s backup and DR plans?
Almost all object storage systems provide replication and/or erasure coding (parity protection across nodes). You can tune that protection by data center row and by site. For example, you could use erasure coding within the data center and then replicate to one or two others. They can also maintain versions of data. While this should be enough, the truly paranoid, like me, should consider creating another copy on tape. There are interfaces that allow a tape library to be addressed like an object store. My colleague Curtis Preston covered this subject in his article “Do you Always Need to Backup?“.