On April 14th, Silk International CTO, Tom O’Neill, and Director of Solution Architects, Dwight Wallace, sat down to discuss untouchable SQL workloads: what they are, how you migrate them to the cloud, and examples of what other companies have done.
You can watch the full conversation here, but in the meantime, here are some highlights:
What is an Untouchable SQL Workload?
An untouchable SQL workload is a high-value workload that is customer-facing, revenue-generating, business-critical application. It often requires a high level of performance that is outside the boundaries of what native cloud can offer – at least 10GB/sec of data processing speeds. And if they aren’t performing at the level they should be, the impact on the business is significant.
These workloads tend to be legacy systems that were developed by a team that is no longer part of the company. For this reason, the current team in place often feels overwhelmed and tends to leave the system alone in a corner of the datacenter on its own dedicated infrastructure. If the team ever attempts to move these workloads to the cloud, they probably fail. Almost every company has some type of workload that has evolved into this legacy behemoth over the years.
Why Do the Cloud Vendors Want Untouchable SQL Workloads So Much?
Yet, despite their seemingly inability to be migrated to the cloud, the cloud vendors are clamoring for them. Why? Because untouchable workloads are extremely large and tend to be anchor workloads. Meaning that they have a number of applications, systems, and environments that depend on them. All of these dependents need to be on the same region and zone to leverage the database licenses and get the performance they need. So if you successfully move an untouchable anchor workload to the cloud, all of these dependent workloads will need to be moved with them.
To capture these workloads, the vendors offer a number of compelling reasons – most notably a significant cost model — for why you should move your workloads to their infrastructure. For example, if you’re looking to move Oracle databases to the cloud, you might see a significant commercial benefit to moving these databases to OCI. In that same vein, Microsoft has been making a major play for on-premises Microsoft SQL Server workloads. With SQL Server 2012 extended support coming to an end this summer, Microsoft is further extending this support for customers who plan to migrate these workloads to Azure.
How to Move Untouchable SQL Workloads to the Cloud
So customers continue attempting to move their untouchable workloads to the cloud. After all, the cloud generally has the fastest CPU, offers the best networking, and provides the latest and greatest technologies and security protocols. But if their workloads require over 1.5-2GB/sec of throughput – more than what native cloud can offer – issues arise.
That’s where Silk comes in. The Silk Cloud Platform is a virtualization layer that lives between customer’s untouchable workloads and the cloud infrastructure. It aggregates the large amounts of VMs these workloads demand, increasing the native performance of the cloud while increasing reliability and resiliency. Silk offers customers up to 10x faster performance compared to native cloud alone. And since Silk aggregates IaaS resources with RAID, the system gets increased resiliency that protects against hiccups like VM reboots or hardware failure in native IaaS offerings.
Who’s Successfully Made the Move?
A number of Silk customers have already successfully migrated their untouchable workloads to the cloud. One large e-commerce company had an aggressive deadline to be out of their datacenter by the end of 2022. They had moved 85-90% of their workloads to a native IaaS cloud offering. But the last 10-15% just couldn’t seem to get to the cloud. The performance requirements for these workloads were, at a minimum, 2GB/sec of throughput which the cloud just could not handle natively. The customer made a number of attempts to move their databases into native cloud, trying out every solution that the cloud provider offered. Yet nothing was successful. The team was forced to make some hard decisions. Their datacenter was coming up for renewal and they were considering if they should make the large financial investment to keep it up and running for their untouchable workloads.
With the introduction of Silk to their stack, the e-commerce company has now been able to completely move all workloads into the cloud and sweep the datacenter floor – way ahead of their deadline.
Another Silk customer has been working since December to move their untouchable MS SQL workloads to the cloud. These workloads need at least 2GB of throughput and everything that the team tested on the cloud – including ultra-fast high end SSDs – did not meet this basic requirement. That’s when the cloud vendor’s sales team brought in Silk. At first, the customer was skeptical that Silk could actually help them. Not only did Silk successfully move the customer’s MS SQL workloads to the cloud, but it decreased the workloads’ latency by 0.2ms compared to their previous on-prem infrastructure. This might sound incremental, but it adds up. The customer is now happily on the cloud and getting faster performance than they had ever seen.
You can listen to the full conversation between Dwight and Tom to learn more about how you can begin getting your own untouchable SQL workloads out of the datacenter and into the cloud.