Jeremy Nees is the CTO and Product Officer @ TheInstillery, a company born in the cloud and automation era. He helps enterprises develop and release new products & services faster with a cloud-agnostic, data-driven approach.
In this Q&A session, Chandra Reddy @Actifio shares Jeremy’s perspective and expertise in multi-cloud snapshot management.
Chandra: Jeremy, How are you? Thank you for contributing to this hot topic of cloud snapshot management and its benefits. Before we get started with Cloud Snapshot Management, perhaps you can set some context with on-premises storage snapshot management that most enterprises are very familiar with.
Jeremy: VMware snapshots have been the default way to protect virtual machines for over a decade. It is fast, cost effective and simple to implement. While it does not cater for every use case, it is a great starting point for protecting your everyday workloads. The maturity of VMware’s VADP API means capturing and restoring a VM is highly reliable and efficient.
Chandra: All the cloud vendors like AWS, Azure, GCP offer cloud-native snapshots from their console. Is there a need for a 3rd party Cloud Snapshot Management? What are your thoughts?
Jeremy: For some businesses exclusively cloud-native snapshots may suffice, however we come across more businesses who have requirements to protect either multi-cloud, hybrid-cloud or use application aware backup. It very quickly becomes burdensome to operate multiple backup platforms, with different functionality and reporting. This is particularly true if there are compliance requirements your business has that are much easier to manage in a single place.
Chandra: An alternate approach to native cloud snapshots is to use agents or connectors in the cloud VMs for backup and recovery. What are the pain points and disadvantages of such an approach?
Jeremy: Snapshot backups are popular for a reason – they are very convenient, and while they don’t offer the capability required in some circumstances, in a lot of cases they are more than adequate. Installing and maintaining connectors or agents adds additional tasks for backup administrators, therefore you really only want to do this where there are clear benefits to doing so.
Chandra: Managing cloud-native snapshots without any connectors seems to be a lightweight operation. Are there any security benefits of this approach?
Jeremy: Anytime you can reduce the amount of software you have to operate and maintain, there is a benefit. Reducing the attack surface is beneficial.
Chandra: Are there any other advantages of cloud-native snapshots?
Jeremy: The restoration process for a snapshot is extremely straightforward and can lower the recovery time for a workload. The downside of course is that it is crash consistent and therefore not appropropriate for some applications such as databases.
Chandra: The snapshots of Cloud VMs are stored in object storage. Do you have any concerns about the performance of recoveries from snapshots?
Jeremy: It depends….and that is the issue. A lot of the time object storage performs to an adequate level, but if you need certainty around recovery times of particular workloads, you may want to consider other forms of backups that can deliver rapid instant mount and recoveries from backups stored in block storage. For example, this could be AWS EBS, Azure Managed Disk or GCP Persistent Disk. Actifio now also has the ability to use block storage as a cache, accelerating access to object storage, but without having to pay for all your data to be stored in block storage.
Chandra: Here is a great write up on EBS Fast Snapshot Restore (FSR). What are your thoughts on this feature?
Jeremy: It’s a really good feature to have available, however it is not necessarily something you would want to have enabled on all your snapshots due to the cost over time. You can enable the feature only for the snapshots just before recovering from the snapshot. I’d use this feature primarily for cloning workloads in auto-scaling groups.
Chandra: The same article on EBS FSR also mentions that the best restore performance you can get is a TB per hour. Generally, large-sized VMs tend to be databases. They also tend to be mission-critical and demand RTO in minutes. Thus, for non-DB Vms, we recommend snapshot management. And for DB VMs, which tend to be fewer, we recommend connector based backup and recovery to achieve database consistency, log management, low RTO, and rapid DB cloning. What are your thoughts? Would you agree with the hybrid approach which gives users the best of both worlds?
Jeremy: I think it is key to have a grasp on what your business requirement is for recovery of a workload. You can then very simply work back from that and calculate if your RTO is achievable. You should also test this, to confirm that performance aligns with expectations. I think often a hybrid backup approach, which includes the use of connectors, allows you to meet business requirements for DB VMs and other workloads with very specific requirements.
Chandra: All the major cloud vendors like AWS, Azure, GCP offer Database-as-a-service PaaS offerings. Yet I see many customers still run databases in Cloud VMs. What do you see is the % adoption of the two models?
Jeremy: There is definitely a desire to move towards Database-as-a-service PaaS offering and we promote these to customers where they are suitable for their applications. It is worth noting that there is not always feature parity or other limitations may exist with PaaS offerings. It is hard to put a % on the adoption due, however an estimate may see 30% of customers using some level of PaaS DB services with adoption growing steadily.
Chandra: When should an organization run databases in Cloud VMs vs PaaS instances? What are your guidelines or recommendations?
Jeremy: That’s an incredibly difficult question to answer due to the disparity of PaaS offerings. One observation I would make is that using a cloud-native PaaS DB services for applications developed to run in cloud is typically an easier fit then using a Common off the Shelf (COTS) application with a PaaS DB service that is essentially just a managed DB service. The requirements of the COTS software can be quite specific and they were not necessarily designed with PaaS offerings in mind.
Chandra: What does Multi-Cloud mean to you? And how important is Multi-Cloud Snapshot Management?Jeremy: Multi-cloud is the use of more than one cloud platform, and in my opinion, can include hybrid cloud. Personally I’ve never had the appetite to operate more than one backup solution across a business, therefore being able to manage multi-cloud snapshots in a single system is a must have for me.
Chandra: And my last question…In the next 12 months, what are the top 5 services that enterprises will look to consume in the cloud as they plan to migrate or re-write their applications to/in the cloud?
Jeremy: A top five may be difficult but some of the most common I see are DB PaaS services, containerisation (in particular using Kubernetes), serverless code execution such as AWS Lambda, cloud native network security and load balancing and basic storage services. The combination of these services allows a lot to be achieved with an application written for cloud. On top of that there are a lot of more advanced services around data analytics and transformation, cognitive services and IoT management that companies are adopting, however these are not yet as prevalent.
Chandra: Thank you so much for sharing your deep expertise in this space.