Why Build a Storage Array in the Cloud?

Multicloud for most enterprises is the reality today. Why? There's tons of reasons. Perhaps you're looking for best of breed for a particular solution. Perhaps you acquired a company using a different cloud provider. Or maybe you're looking to diversify your workload placement based on costs. Multicloud comes with benefits, but it also introduces challenges.

The biggest challenge of multicloud is the inconsistency of operations and services. While many of the primitives are the same: VMs are VMs are VMs; the management interface and nuances are completely different. Sometimes those nuances are the reason why one cloud is favored over another, but there's no doubt the inconsistencies make management and operations less efficient.

Wouldn't it be nice to have a consistent interface and solution that works across all the public cloud vendors and in your private datacenter too? Dell Technologies seems to think so, and at Cloud Field Day 19 they presented a series of storage solutions under the APEX brand.

First up, they talked about APEX Block Storage which is based on PowerFlex/ScaleIO storage. Some bold claims were made, stating 87% cost savings along with 100x performance improvement. Extraordinary claims require extraordinary proof, and allegedly they have a whitepaper that I have not been able to find a link for.

The APEX Block Storage product is essentially an aggregation of VMs in a given cloud- say a bunch of Azure VMs- running the APEX Block Storage software with a bunch of attached disks. Dell is create a storage array from VMs in the cloud. Each node in the cluster is called a ScaleIO Data Server (SDS) and VMs that will consume the the must be running the ScaleIO Data Client (SDC). Yes, you'll need to install a client on each VM that will consume this storage.

Leveraging the ScaleIO software allows the client to distribute traffic across the nodes and aggregate their throughput. The bottleneck will be the network bandwidth of the VM family. The latest Azure VMs with Boost can hit 200Gbps in bandwidth, so that might not be much of an issue. Since the I/O is being consumed over the network, you need to be mindful of egress costs between availability zones if you're running in a multi-AZ configuration in AWS.

It is the aggregation of bandwidth and custom software stack that enables Dell to claim the big 100x numbers. Their example is an APEX Block volume up to 1PB in size running at millions of IOPS per volume versus an Azure Ultra Disk with a max of 64TB running at at 160K.

If this idea sounds familiar, don't worry you're not crazy. Similar solutions have been implemented from vendors like Weka and Pure Storage. And in a similar vein, they also tout the improved performance and lower cost than using the native solution in each cloud.

It's not just the cost and performance that you get as a benefit. Going back to my previous point around consistency, APEX includes Navigator, which is their management plane for APEX Storage. So you can use the same management interface to control your storage across the public clouds and on-prem if you have local ScaleIO installed as well. However, APEX Navigator is not free, so you'll need to pay per TB of storage managed on top of the cost of your APEX Storage deployments.

I should say that this solution is not for the SMBs of the world. The minimum configuration has 10TB of usable storage, and the expectation is that customers will be deploying pools in the petabyte range. We're talking enterprise customers with a need for high-performance, low-latency storage across multiple clouds using Dell APEX.

If you have a dedicated storage team that wants a consistent solution across multiple clouds, Dell APEX Block Storage seems like a solid option.