Skip to main content


Tomasz Klim

Comparison of Public Cloud Platforms

When choosing the right platform for your next project, don’t worry about management tools. Instead, worry about capabilities and restrictions of each platform, because that’s what can stall your business or allow it to spread its wings. Here are our opinions, based on our experience.

 

cl-aws-tPricing: lowest prices per hour for small instances (even lower prices for reserved instances when you pay upfront). Quite expensive for bigger instances. Billing per hour.

VPC-based internal networks: great for implementing privilege separation eg. as a part of PCI DSS certification, and helpful for iptables-disabled people. However, it works differently from the ordinary LAN that you can set up in your office. Additionally, it results in highest network latencies in most scenarios.

The best known platform. For many people cloud = AWS.

Best APIs, SDKs for many programming languages, documentation, examples and other materials for developers, administrators and IT managers. Particular APIs (eg. EC2, S3) are de-facto standards.

Has lots of complementary services. Most of them are vendor-specific, but some are fully generic (RDS, SES, Route 53 etc.).

cl-azure-tVirtual networks work almost just like ordinary LAN. This includes the possibility of setting up your own DHCP/DNS services. This means you can develop complex solutions using low-cost office-grade network equipment and deploy everything without reinventing all of the network part.

“Resource groups” concept – you can export/import/clone your entire infrastructure (or chosen parts) using JSON template files. I’ve judged it to be great for working with complex per-developer development environments.

99.9% SLA for single instances (with premium storage). I think it’s very good for maintaining old applications written for a single server.

In general, the best place for Windows-based solutions. You’ll find many production-ready complex templates, not just for Windows Server and SQL Server, like everywhere else.

Has lots of complementary services. These are mostly vendor-specific, Windows-specific or big data-specific, but also generic ones eg. DNS or Redis.

Minute billing, no exceptions.

Less tolerance for sending mailings, especially “cheap” mailings (but still not spam). It’s easier to get blocked in response to falsified, exaggerated or inadequate-to-laws-applicable-in-customer-country-or-cloud-instance-country abuse reports.

cl-gce-tLowest prices per hour for bigger instances + even lower prices (80% discount!) for preemptible instances. These can last up to 24h, and can be terminated by Google at any time, giving you 30 seconds to finish).

Minute billing. But you always pay for the first 10 minutes, and there are other exceptions.

Virtual networks work almost just like ordinary LAN, including the possibility of setting your own DHCP/DNS services. This means you can develop complex solutions using low-cost office-grade network equipment and deploy everything without reinventing all of the network part.

Lowest network latencies in Europe (both internal and external; at least from our observations).

Has a variety of complementary services with good price-quality ratio. However, everything is vendor-specific, and you won’t find any equivalents with at least a similar ratio.

Best load balancing (real scalability, fully automatic, capable of handling any amount of traffic). On the other hand, it’s expensive: you also pay for load balancing rules, not just for traffic.

cl-e24cloud-tLocated in Poland (Poznań and Warsaw) – lowest possible network latencies from/to Poland, Polish support, Polish invoices etc. Helpful if your business is located in Poland.

Overall, I’ve judged its quality to be a bit lower, but it’s enough to host less important solutions (e.g. development environments). They had a major failure in 2013, but published materials suggest that they have learnt a lot since then.

cl-rackspace-t2Based on OpenStack, with many behaviours directly exposed to the customer. A cheap way to learn OpenStack before attempting to run your own private/hybrid cloud.

Occasionally, the instance setup phase can be unreliable (I’ve seen it fail quite often).

Its major advantage could be that it suits high-traffic solutions targeted to Australia.

 

Amazon AWS is our frequent choice, but there are many pros to other solutions. Different experiences? Please share!

Share:Share on FacebookTweet about this on TwitterShare on LinkedInShare on Google+Pin on Pinterest

Like what you see?

Get in touch! We'll respond quickly, and we'll keep your data confidential.