Last month, the SNIA Networking Storage Forum (NSF) hosted a live webcast, “Object Storage: What, How and Why.” As the title suggests, our NSF members and invited guest experts delivered foundational knowledge on object storage, explaining how object storage works, use cases, and standards. They even shared a little history on how object storage originated. If you missed the live event, you can watch the on-demand webcast or find it on our SNIAVideo YouTube Channel.
We received some great questions from our live audience. As promised, here are the answers to them all.
Q. How can you get insights into object storage on premises? E.g. quota, who consumes what, auditing the data for data leaks, etc. Is there a tool for understanding and managing object data solutions?
A.Yes, on-premises storage systems have quota management including enforcement options (and even that can be a hard or soft enforcement). As for data leaks, object storage access and consumption are always logged but even on-premises, the security model of the Internet should be used there as well (take security seriously). If you’re unsure about where to start with security, may we suggest our Storage Security Series.
Q. Are object sizes a consideration if it makes sense to use object storage?
A. Yes, both the size of the objects and the overall amount of data are important, especially around egress from the public cloud. This is where modeling and testing a solution will provide valuable feedback to the performance and economics of object storage for a use case.
Q. I hear that object storage equals SLOW storage i.e. for backup or archives, but can object storage have high performance and if so, what use cases are there for high performance object storage?
A. Object storage is not necessarily slow; in some cases it can be quite fast. It depends on how the application writes and reads data from object storage and on the media used for the object storage. A single monolithic file simply put in object storage will have a different behavior characteristic than a more data-specific placement in object storage. Different public cloud service levels (Hot, Cool, Archive for example on Azure) make a difference in performance as well. Lastly, public cloud throttling can come into play.
Q. You mentioned that on-premises-deployed S3 compatible object storage solutions support object locking or enforced retention (like Amazon’s Object Lock feature). Cloudian’s HyperStore supports a fully compliant and SEC17a-4/FINRA Object Lock WORM solution. Does NetApp StorageGrid support Object Lock?
A.Some on-premises can do object lock as well. We recommend that you take a specific look at each vendor’s support specifications for more details.
Q. What is the minimum size of data below which object storage becomes inefficient, and other types of block and file-system storage are more efficient?
A. There is no single number that would answer this question for every application.
Q. How do current customers define the common meta-data format, when you have a variety of data which are hard to group?
A. This would really be determined at the application level; specifically, what application is reading and writing data from object storage.
Q. Can Object storage be enabled with versioning capabilities? Is there any limit on the total number of versions for an object?
A. In some cases (for instance, at least for S3 storage), the version is a copy of an object. Each copy has a version. It will depend on the specific vendor’s solution as to whether there are limits of the versions, and it is best to always check the vendor’s available information for their implementation details.
Q. At a high-level, how would you differentiate on premises-based object storage solutions from public, cloud offerings
A. The main differences are that with on-premises solutions the customers have full control whereas the public cloud and service provider offerings are more globally accessible. Additionally, public offerings may have additional features, functionality, and service levels available.
Q. Fundamentally, object storage eventually lives on block storage. For things like erasure coding for geo distributed access and protection, does the object storage engine handle that replication of data blocks on SSD/HDD storage up at the application layer?
A. Each object storage provider will ensure the availability of the data in their own way at the hardware control plane level. The public cloud providers intentionally abstract the details of the hardware in many cases; the shared responsibility model however puts the ultimate control of the data on the tenant.
Q. What are the leading object storage solutions in the Gartner benchmark?
A.As recently as 2019, Gartner did issue a Magic Quadrant for Distributed File Systems and Object Storage which showcases the industry solutions for non-hyperscale object storage implementations. Many of the vendors allow reprints and we recommend you read the full report for those implementations.