Last year, I attended Ravello Blogger Day 1 (#RBD1) which was about Oracle Cloud Infrastructure and relaunch of the Ravello on Oracle Cloud Infrastructure. This year, I was invited again to attend the second installment of it (#RBD2), which was more of an update on the past year, new things to come but also about gaining feedback from the community.
There is so much that we discussed and it can’t possibly fit in a blog but I will try to give you as much as possible here. I’ll have to break it down into two posts, this being the first one and slide heavy and a second one, with fewer slides but also a few thoughts from me. Please bear with me on the slide-fest as I am only including the important one that do cover a lot of information.
The day started with Rahul Patil (Vice President, Engineering – Core IaaS) talking about the vision, strategy and areas of focus for Oracle Cloud Infrastructure. Oracle’s vision is that customers should be able to run all their business applications on OCI.
It goes without saying that Oracle is focussing on their Enterprise customers. According to Rahul, they are talking to their customers and implementing what’s important to their Enterprise customers first.
A few things to note here are:
- Flexibility – Fundamentally, it’s about enabling you to move to OCI without modifying your applications
- Uncompromised Performance – There’s a lot of emphasis on the high performance, flexible virtual network. More on that below.
- SLAs – OCI offers SLAs on Performance and Manageability of storage and networking, in addition to the various Availability SLAs
- Pricing – The idea here is not only to make pricing simpler but also cheaper. Not a lot of details but in general, they say OCI will be cheaper than AWS.
The discussion then moved on to Oracle’s focus on providing their customers with performance guarantees for compute, storage and networking, backed with SLAs.
OCI provides a non-blocking Clos networked environment i.e. there is no over-subscription, which in turn, makes it adhere to their published SLAs. As seen in the slide above, it also makes use of “off-the-box” networking, meaning that the networking is virtualised outside of the hypervisor. That makes networking flexible and able to fully utilize the underlying network performance.
Last year was more about getting the infrastructure defined and creating that blueprint to replicate the same model across the various geographic regions. While that work continues, the focus is now shifting towards introducing/integrating more services living upwards in the stack.
As you can see from the slide above that a lot of effort has gone in since last year to make new services and features available, which shows the commitment that Oracle has in building an Enterprise level cloud infrastructure.
Even though we didn’t go into a lot of details about SLAs, I did mention above that in addition to the usual Availability SLAs, Oracle also provides Performance and Manageability SLAs. That was interesting and a major differentiating factor so I thought you might be interested. Here’s a slide that shows a comparison of what Oracle is offering as compared to AWS and Azure:
In my post last year, I said Oracle will need to expand their geographical coverage as soon as possible if they want to compete with other public cloud providers. I did ask this question again and not surprisingly, Oracle is aware and working hard at that. However, I’d say it needs to accelerate a lot if they don’t want to risk their customers start looking elsewhere. For information, below is a slide with the current status and upcoming plans:
Rahul then moved on to what are going to be OCI’s “Core Themes” going forward into the next year:
- Enterprise Lift & Shift
- Performance Intensive Workloads
- Cloud Native Workloads
By “Enterprise Lift and Shift”, Oracle does not just mean Oracle workloads to seamlessly move to their cloud but also non-Oracle workloads such as VMware (with the help of Ravello, of course), SAP and even Microsoft Applications. It also means that companies will be able to move L2 networking dependent workloads which typically are not allowed in other public clouds, due to the way they’re built. In addition to that, bare metal instances will allows BYOH (Bring Your Own Hypervisor) which is extremely useful in taking the complexity out of moving to a public cloud infrastructure, resulting in a quick move. Oracle is also working on Governance and Security aspects and aims to be compliant with all the leading compliance standards.
Oracle is also focussing on enabling workloads that demand large scale and high performance e.g. Big Data, HPC and Machine Learning etc. It aims to provide a platform that is not only high performance but also cost-effective, to make it attractive to more customers. There will be many instance sizes and pay-as-you-go options as well.
In terms of Cloud Native Workloads, Oracle will be working on integration with open standard tools e.g. Terrform, Puppet and Chef etc. but will also add support for container services, serverless infrastructure and common DevOps practices.
Next in line was Kash Iftikhar, who spoke a little more about the infrastructure offerings. You might have already noticed that a lot of work has gone into underpinning all the different services with a high-performance non-blocking network infrastructure so this conversation was more about how they’ve also increased the capacity for all core services that can talk via the same flexible virtual network.
In terms of network infrastructure components, they’re pretty much identical to AWS and Azure but with different names of course. However as I said before, the selling point here is the bandwidth and latency between the core services. Note that there’s less than 500 microseconds latency between availability domains. Granted it’s one-way but still pretty good for what should be different sites.
As for compute, you can see there are many instance sizes available for both regular VMs and bare metal. That said, it’s still not as many as what other cloud services offer. When discussed, Kash said that these are the most common instance sizes used and demanded by the customers so it makes sense to make those available as a priority.
For storage, all the different options that you would expect from a public cloud offering e.g. File, Block and Object storage are available. Points to note here are the same 25 Gbps network available in between core services but also the performance SLAs available for the I/O sensitive storage options.
Being primarily a database company, you would expect OCI to be strong in its offerings on that front. That is indeed the case and there are many options available terms of capacity, performance and availability. again backed by SLAs.
and then, of course, there’s Ravello, that enables running of VMware/KVM workloads without modification on OCI. As we’ve known for a while, the prime advantage of Ravello is the ability to abstract the underlying physical infrastructure in such a way (using HVX) that moving an entire application stack to OCI becomes just a matter of transferring a bunch of images and bringing them up.
Kash also went through what they are doing in the DevOps and applications space and that was his presentation done.
As expected, I am over 1200 words for this post so I am going to end this one here and start the other installment. Don’t worry, I aim to have just one more and about the same size so hopefully, I will see you there!
[Update 14/03/2018 22:35] I’ve also published Part 2 of this post now. Go have a look: Oracle Ravello Blogger Day (#RBD2) – Part 2