Introduction
There are many types of cloud providers, public vs private, cheap vs expensive. You can read my article on which provider to pick and when. Today, we’re going to talk about the expensive clouds for enterprises, namely SoftLayer, AWS and Google.
For reference, IBM SoftLayer is used to order dedicated beefy servers. Including but not limited to terabytes of memory and a dozen of local SSDs. You build a server by selecting individual components (enclosure, CPU, RAM). It’s similar to ordering hardware from Dell but it’s leased to IBM and it stays in their data center.
Virtual servers from public clouds are mostly the same thing [1] except you select the amount of RAM and CPU directly, not branded hardware parts.
[1] A server is a commodity. It’s just a bunch of hardware put together in a box. The distinction whether ordering from Dell, HP or Lenovo is not meaningful in the current era. Everyone is running and reselling the same hardware.
Why SoftLayer?
Historically, the dedicated cloud allowed you to pick specialized hardware at a decent price. It wasn’t possible to do that with AWS.
IBM SoftLayer:
- High memory instance with fast storage
- Higher maximum performance and better performance/costs ratio
- “Dedicated”
- Many locations (including Asia and South America)
AWS/Google capped around 200 GB of memory and didn’t provide fast local storage. That was a problem if you needed it. Worse yet, their instance pricing doubles with the amount of RAM/CPU so the biggest instances were a total rip-off [*].
“Dedicated” means that the hardware is yours and has a well-defined physical presence, chosen among the many cities available. It could make it easier to deal with regulations and data management. Some business care about that and it was enough to tip the balance in favor of SoftLayer.
[*] For reference, buying an additional memory stick does not double the price of a physical server.
What Went Wrong With SoftLayer?
Let’s summarize the last paragraph. IBM SoftLayer had strong competitors and it differentiated by providing an unmatched product (beefy servers) at a reasonable price point.
What happened next? The competition caught up eventually. They broaden their offerings and ate their market from the bottom.
One day you wake up and Google is at your door.
Hardware (right to left):
- IBM. 56 vCPUS. Dual E5-2690 v4. 512 GB. 2 x 600 GB SAS. Dual 10 Gbps NIC.
- Google n1-highmem-64. 64 vCPUS. 416 GB.
- AWS r4.16xlarge. 64 vCPUS. 488 GB.
- AWS i3.16xlarge. 64 vCPUS. 488 GB. 8 x 1900 GB SSD.
- AWS i2.8xlarge. 32 vCPUS. 488 GB. 8 x 800 GB SSD. (previous generation)
Notice that all the providers are in the same ballpark.
Last year, Amazon released the new high memory instances [1][2] and high performance NVMe disks [2]. Google released higher memory instances [4] while they already had faster and more flexible disks.
None of these were available 18 months ago. The competition has been coming strong and aggressively.
For comparison, the i2.8xlarge is old generation hardware (year 2014; limited to 256 GB memory). It is half the capacity for double the price. I’ve commented before on them being a total rip off.
[1] https://aws.amazon.com/blogs/aws/new-next-generation-r4-memory-optimized-ec2-instances/
[2] https://aws.amazon.com/pt/blogs/aws/now-available-i3-instances-for-demanding-io-intensive-applications/
[3] https://venturebeat.com/2017/03/03/google-cloud-launches-vm-instances-with-64-virtual-cpu-cores/
[4] It would be nice of Google to make official announcements. Googling “n1-highmem-64” and “n1-highmem-128” found nothing. There is no reference I can link to.
What’s Left to Defend About SoftLayer?
The competition went from non existent and significantly more expensive, to suddenly providing the same thing at the same price. It hurts for SoftLayer and it will only get worse.
To be fair, Google cloud has been really disruptive to the market, they came out of nowhere and delivered an amazing product with great pricing. Why would any business gives lower pricing AND automatic discounts instead of milking their customers maximizing revenues? [1]
What about performance and specialized hardware? The major cloud providers are catching up on that. More memory, faster disks, stronger infrastructure. The comparison is over when they can all deliver a terabyte of memory with a good local disk, it’s enough to run anything.
Location is not an argument anymore. AWS and Google have been opening new data centers all around the world. They might not be everywhere yet but they will be soon.
What about being “dedicated“? It’s worthless at this point. Either you’re ignoring that aspect entirely and it doesn’t matter. Or you care about being HIPAA/PCI [3][4] compliant and being in the cloud is not an obstacle, AWS might actually be easier to certify.
It seems to me there is nothing left in favor of SoftLayer.
[1] See an introduction to market pricing and market segmentation, or how to make a profitable business more profitable. https://www.joelonsoftware.com/2004/12/15/camels-and-rubber-duckies/
[2] https://aws.amazon.com/compliance/hipaa-compliance/
[3] https://aws.amazon.com/compliance/pci-dss-level-1-faqs/
Conclusion
What’s the point of SoftLayer? I don’t know.
There is nothing they do particularly well that the competition isn’t already doing or soon to do. Their future is grim.
If I have to pick a third cloud for diversification, that should be Azure, not SoftLayer.
SoftLayer is so far away it’s playing fourth fiddle.
The HFT Guy 