What you should look for when evaluating in-memory NoSQL solutions?

• Automated scalability. The solution should be infinitely scalable in a fully-automated manner. This means you should not deal with nodes, clusters and scaling operations. All that should be automatically done for you.
• No data loss. Memory nodes often crash and when this happens you lose all data stored on them. Your solution must include persistent storage, auto-failover and backup capabilities. Moreover, these processes must be fully automated and guarantee data continuity on failure and zero data loss.
• Uncompromising performance. In some solutions performance is poor when your dataset is small. Make sure your solution uses the strongest servers and preferably processes your data using multiple servers (cores).
• Zero management. Besides creating your database, all other operations tasks (software upgrades, clustering, failure recovery) should be fully-automated and not require any action on your part.
• Pay-as you go model. Look for a solution that charges according to the actual GB/h used by your dataset. Many solutions charge by the number of cloud servers, which means you end up paying for more than you need.