Skip to main content

Lesson Learn - Vertical scalability is dead end, especialy when your app runs in cloud

This week I was involved in a migration of an existing project that was written on Java stack to Azure. The current solution is using different technologies like Oracle DB, HA Proxy, Apache Tomcat, Elastic Search, Load Balancer and so on. There are around 5-6 different out of the box applications that are running around the solution.

Objective
The main purpose of the workshop was to understand the current architecture, how different components communicate between each other and what are bottlenecks of the current solution. The expected output of this workshop was a first draft of migration plan from on-premises to Azure.
As usually, I’m done my homework before. I already had in mind different approaches using not only Azure VMs, but also Web/Worker Roles, Docker and Service Fabric. I didn’t know the current architecture, so I tried to sketch some possible solutions.

Current solution
The purpose of the solution is to increase the security of user identity by offering a additional verification layer. The idea is very nice and with a high potential, especially for large companies, this solution could improve their security level.
I was really impressed about the idea and the way how was implemented. A lot of systems integrated together. Each system was scalable by himself. So, I started to realize that migration to Azure will be pretty nice and easily.  

Proposed solution
The main limitation that we had was related to migration budget. They have budget for the infrastructure part, but zero budget for code changes.
Ups… this might be a little tricky, but It sad that this is possible. Because the current solution was composed from multiple system combined together, I saw a good opportunity to take each sub-system and deployed it on a different machine. Without being able to modify the code, running the code on Web Apps, Worker Roles and so on would be pretty impossible. But at least we have Azure VMs.
The proposed solution was to deploy each application (Oracle DB, HA Proxy, Elastic Search) on different machines. Based on the load we could group them together, under the same VMs – it doesn’t make sense to have a dedicated VMs instance for each application, if the load is not high and the resources that are consumed by the application are low.


In this way we could have a cluster of VMs that could scale at different levels based on the load. This would be the first step of the migration. After this, when they would have resources for code change, we could migrate a part of the system on top of Azure Services – until then, this solution is okay and can be used for a long period of time.

The big WALL
After drawing the solution on a whiteboard and presenting it, I saw that the technical people from the team were not happy at all. After a short discussion I discovered that in this moment the solution is running in only one VM and it is written in a such a way that there is no component that can be extracted outside, without changing the code.

I asked in my mind a big WTF and trying to understand what you could write such an application. This means that you have the configuration hardcoded inside your code (connection strings and settings that point to localhost). Configuration script of systems like HA Proxy that point to local host (without having the source and target IP extracted as variable) and many more.
Still after one day I’m shocked and I cannot understand what the hell you could do such a big mistake. There is ZERO scalability. The only way how you can scale a solution like this is by changing the hardware and allocate better and faster CPU, Memory, Disks. This kind of scalability is dead end – extremely expensive and clearly that the limit will be hit very soon. Once you have a lot of users that are using your system…. the nightmare will begin.

The disappointing solution
The discussion ended very fast and with the simplest solution ever. One Azure VM that is running Ubuntu where this guys have remote access. On this machine they have free power to deploy the entire system. We cannot even have two instances of the VM…. (I still don’t understand why they have a Load Balancer in front of their app, that is running on the same machine, if you have only one instance.

Lesson Learned 
Never and I mean never – not even in a PoC to hardcode configuration in the code. This will tied your hands for any further real scalability option. It is good to start with a simple system, but at least try to decuple it and manage in the right way all the dependencies.

Comments

  1. If somebody tries to use a cloud provider just as another hosting company, there is almost no advantage for going to the cloud..

    ReplyDelete
  2. But still, the aproach is still wrong, indifferent that you design a solution for Cloud or on-premises. You don't put all your eggs in only one VM Instance... it is a death end....

    ReplyDelete
    Replies
    1. Indeed - maybe all they can afford is only one VM instance that can't be split :)

      Delete

Post a Comment

Popular posts from this blog

Windows Docker Containers can make WIN32 API calls, use COM and ASP.NET WebForms

After the last post , I received two interesting questions related to Docker and Windows. People were interested if we do Win32 API calls from a Docker container and if there is support for COM. WIN32 Support To test calls to WIN32 API, let’s try to populate SYSTEM_INFO class. [StructLayout(LayoutKind.Sequential)] public struct SYSTEM_INFO { public uint dwOemId; public uint dwPageSize; public uint lpMinimumApplicationAddress; public uint lpMaximumApplicationAddress; public uint dwActiveProcessorMask; public uint dwNumberOfProcessors; public uint dwProcessorType; public uint dwAllocationGranularity; public uint dwProcessorLevel; public uint dwProcessorRevision; } ... [DllImport("kernel32")] static extern void GetSystemInfo(ref SYSTEM_INFO pSI); ... SYSTEM_INFO pSI = new SYSTEM_INFO(

Azure AD and AWS Cognito side-by-side

In the last few weeks, I was involved in multiple opportunities on Microsoft Azure and Amazon, where we had to analyse AWS Cognito, Azure AD and other solutions that are available on the market. I decided to consolidate in one post all features and differences that I identified for both of them that we should need to take into account. Take into account that Azure AD is an identity and access management services well integrated with Microsoft stack. In comparison, AWS Cognito is just a user sign-up, sign-in and access control and nothing more. The focus is not on the main features, is more on small things that can make a difference when you want to decide where we want to store and manage our users.  This information might be useful in the future when we need to decide where we want to keep and manage our users.  Feature Azure AD (B2C, B2C) AWS Cognito Access token lifetime Default 1h – the value is configurable 1h – cannot be modified

What to do when you hit the throughput limits of Azure Storage (Blobs)

In this post we will talk about how we can detect when we hit a throughput limit of Azure Storage and what we can do in that moment. Context If we take a look on Scalability Targets of Azure Storage ( https://azure.microsoft.com/en-us/documentation/articles/storage-scalability-targets/ ) we will observe that the limits are prety high. But, based on our business logic we can end up at this limits. If you create a system that is hitted by a high number of device, you can hit easily the total number of requests rate that can be done on a Storage Account. This limits on Azure is 20.000 IOPS (entities or messages per second) where (and this is very important) the size of the request is 1KB. Normally, if you make a load tests where 20.000 clients will hit different blobs storages from the same Azure Storage Account, this limits can be reached. How we can detect this problem? From client, we can detect that this limits was reached based on the HTTP error code that is returned by HTTP