Building for high availability: Measuring success

Although highly available is easy to grasp conceptually it can be quite hard to define in practice. To be able to strive for higher and higher availability you will need to figure out how to measure it. To measure it you will need to define exactly how to calculate it.

A typical API service from client to service and back looks something like this. With the request starting at a client, traversing the internet before it hits the boundary of your service and then the response flows back the same way.

It is important to realize that any part of this chain can fail, and if it does, it will lead to a drop in availability as perceived by your clients. A large part of this you have no control over, and it is also fiendishly hard to even measure. If you only measure availability for requests in your service, you are missing a lot of potential failure modes. If one of your hosts goes bad it might not be able to report the metrics of failing requests or incoming network traffic stop all together.

It is often sufficient to measure availability from the first system that you have access to consistent logs from. This usually means either the gateway or if you are not using that a load balancer. If you are using Amazon API Gateway it can give you excellent request logs that are very useful for measuring availability and latency among other things. It will also emit Amazon CloudWatch Metrics that can measure availability directly both for the entire API and for individual methods.

How do you define availability?

The first thing you need to do is to separate out errors and faults in your metrics. An error is a request that could not be processed because of some problem with the contents of the request. A fault is request failure that is caused by a fault either in the communication chain or the implementation of the service. It is important to separate these out because as a service owner you have little to no control of the errors because they are due to a mistake in the client that calls you. Faults however do reflect your availability and are not dependent of mistakes made in the calling client. Worth noting though is that even though errors generally do not count against availability, they can if they represent errors that should not happen because a bug in your code. It is worth having some visibility into having an unusually high error rate.

If you are using HTTP to implement your API, errors should be any response status code between 400 and 499, faults are any status codes over 500 (Inclusive). Make sure that you implement your service to follow this pattern (Basically, do not invent your own usage pattern for the HTTP status codes). If you are using Amazon API Gateway, you get a metric for 4xx responses and a separate metric for 5xx responses. If you need better visibility into exactly what kind of error you are receiving, you can also set up a Amazon CloudWatch Logs Metric Filter on the request log from Amazon API Gateway.

How to calculate availability

Usually, availability is calculated as a percentage. This percentage represents the amount of traffic that is not faulty compared to the total request count. How exactly this percentage is calculated though is not as easy as it might sound and more on that in a bit.

When it comes to picking a goal for availability it is up to you as an engineer to come up with a goal that you are comfortable with. Another common pattern is that once you have implemented a proper availability goal and have good visibility into it on an ongoing basis you can always strive for higher by improving your goal incrementally. As an example, most Amazon Web Services have an availability Service Level Agreement (SLA) of 99.95% or higher. Most services can probably make do with a lower goal if you implement appropriate retries in your clients.

Simple Availability

The most obvious and simple way of defining this is to just use the ratio of non-fault requests divided by the total number of requests. With this definition your if you have a goal of 99.95% availability means that you should only have at most 1 faulty request for every 2000 requests. The advantage of this approach is that the value generally comes right from your metrics and is super easy to monitor and calculate. Using Amazon API Gateway this availability can be calculated directly from metrics emitted to Amazon CloudWatch Metrics. This is also a metric that is suitable for putting on a graph over time to visualize availability.

Calculating Availability for a Service Level Agreement

This way of measuring availability has its issues though because with this definition if you have otherwise perfect availability you can have an almost 4.5-hour long outage without breaking your 99.95% available goal for a year. But if you have a background level of continuous availability that is not perfect this does not generally negatively affect your consumers significantly, but it will significantly reduce the time you can have an outage before you have broken your goal. This difference become increasingly important once you start having an actual Service Level Agreement (SLA) for your service.

One way of addressing the shortcoming of the previous definition is to define your availability in the number of minutes you are above a certain minimum availability. An example of this definition would be that you measure your availability in the number of minutes you had an average availability over 99.99%. You can now have an availability SLA of 99.95% and in this case if your availability normally stays over 99.99% you get to use the full 4.5 hours long outage before you start breaking that SLA over a year. The bad news is that there is no easy way of calculating this metric without looking at each individual availability data point for every minute during the period. The same method can also be used with any period other than a minute.

Optimizing for client experience

If you are looking for the best experience for your clients though the previous methods still has their shortcomings. To illustrate this let us take an example where you introduce a bug that makes 100% of calls fail for 1% of your clients. In this example the way your API is used clients normally make an initial list request followed by 25 detail requests. But for the 1% of clients that get failing calls the initial list call fails. So for clients for whom the service works they make on average 26 calls where the failing clients only make a single call. In this case the simple available is 99 * 26 successful requests for every total 99 * 26 + 1 requests which translates to a simple availability of 99.96%. However, this hides the fact that 1% of your clients can not use your service at all.

The way to measure availability to catch cases like this is to define your availability goal per time period and per client. As an example, you can define the availability as the number of minutes where 99.5% of your customers have more than 99.99% availability. In the example above only 99% of clients have any availability which means that every minute is an unavailable minute by this metric until the bug is fixed. The bad news is that there really is no way of calculating this kind of availability without processing all your requests per minute to determine if you are in breach. So, it is by far the most complicated and expensive way of calculating availability. This method of calculation could potentially save you money for your SLA refunds though since if you apply it to SLA calculations you can keep track on which clients your service has breached the SLA on a per client basis instead of the previous method which would apply equally to all clients once in breach.

How to detect network outages

There is a problem with measuring availability by simply instrumenting the boundary of the service and that is what if you encounter an issue outside of that boundary. If your internet service provider suffers an outage it would stop all incoming traffic to your service. Your availability would still be 100% because there are no failing requests that you are aware of because they fail before they even reach a point in the communication chain that you can measure.

The solution for this problem is to create a canary that makes at least a minimum number of requests to your service in a way that imitates real client scenarios as closely as possible. This can be as simple creating a Amazon CloudWatch Events that triggers a AWS Lambda that generate traffic to your service. On top of this you need to add monitoring that alerts you when there is no traffic coming into your site. Ideally as your service grow you can trim this alarm to alert you when the traffic pattern goes below anything that is abnormally low instead of close to 0. That way you can also detect partial outages that are normally out of your control to measure. Furthermore, make sure that your canary emits metrics on the success of the calls it is making. Your canary traffic metrics will represent a true measurement of availability and latency covering the entire communication chain. It does only represent a small portion of all traffic, but it does properly measure all potential failures that a real client could encounter.

Latency as an aspect of availability

Even though technically latency does not affect your availability, it is extremely important for a good client experience. Latency can be hard to visualize. You might be tempted to believe that just taking the average of your request times will give you a good idea of what the latency of your service looks like. However, latency tend to have a very long tail and using the average generally is not best practice for ensuring that your clients have a good experience. As an example, below is the latency charted for a week of a sample service where it is aggregated as average, median (p50), p90 and p99. If you are unfamiliar with the pXX notation it denotes the percentile. The p99 graph represents how long the worst 1% request took to process.

As you can see in the example above there is a big difference between how you measure latency. The graph for maximum is cut off and goes all the way to 29 seconds in the worst-case scenario. In any environment with software defined networking and a decently high load you will be seeing strange outliers, so the maximum measurement is usually not very useful. Similarly, as you can see the average measurement can also hide issues that affect a not insignificant amount of your traffic. Using the p99 measurement to visualize your latency performance is usually a good middle ground. It includes enough of your worst behaving requests to see if you have significant issues with outliers taking a long time, but also ignores some of the more egregious network blips that can give extremely rare, but very high measurements otherwise skewing your graph.

When measuring anything using p99 aggregation another thing that is very important is the period under which you aggregate. You want to make sure that during the period you are measuring you have at least 100 measurements or more. If you do not, then p99 will be the same as maximum which leads to undesirable results. If you have at least 100 requests during the time period, you get to remove at least 1 request that is an anomaly before it affects your p99 measurement. If you have a minimum call rate of 1 call per second you will need to use a measurement period of at least 1 minute and 40 seconds or you will fall into this trap. Usually, you would use 5 minutes though if you do not have enough traffic to measure p99 for 1 minute though.

Finally, it is worth pointing out that each point in your service architecture will add latency. Same as with availability, it is important to measure the latency as close to the client as possible. Apart from using canaries you can rarely measure it from the client, but usually the gateway is a good place to collect latency measurements that are a good representation of your general client experience.

Create an availability dashboard

Your goal should always be to strive for higher and higher availability. To reach for this goal though you need to have visibility into what your current availability actually is. At minimum this requires you to monitor at the following on a continuous basis.

• Availability - The number of faults divided by the total requests coming into your service.
• Error rate - The rate of invalid requests that you are receiving. Even though this can be a false alarm, it can be an indication of a faulty deployment causing existing traffic to now fail if you see an unexpected change in the rate.
• Transactions per second (TPS) - The number of requests coming into your service. The key thing you want to look at here is if there is a precipitous drop because that likely means a network failure that has occurred before you can measure it. A large, unexpected increase in traffic could also be an indication of a denial of service attack.
• Latency - You should have goals on your latency and strive to decrease it. The way to have and keep these goals is to put them on a dashboard to make sure that you are aware of any changes in trends. If your service has different classes of operations that have significantly different latency profiles, you might consider separating each one out as a separate graph.

Below is an example dashboard that you can implement if you are using Amazon API Gateway as the gateway for your API.

Here is the definition of this dashboard in Amazon CloudWatch Dashboards. All you need to do is change the metric dimension of ApiName from YourAwesomeApi to whatever your API is called and reuse it. You might also need to tweak your minimum TPS limit and error rate amounts to something suitable for your traffic patterns.

  {
    "widgets": [
      {
        "height": 6,
        "width": 12,
        "y": 0,
        "x": 0,
        "type": "metric",
        "properties": {
          "metrics": [
              [ { "expression": "100*(1-m1)", 
                  "label": "Availability",
                  "id": "e1", "region": "us-east-1" } ],
              [ "AWS/ApiGateway", "5XXError", 
                "ApiName", "YourAwesomeApi", 
                { "id": "m1", "visible": false } ]
          ],
          "view": "timeSeries",
          "stacked": false,
          "region": "us-east-1",
          "stat": "Average",
          "period": 60,
          "title": "API Availability",
          "yAxis": { "left": {
            "min": 99.7, "max": 100, "showUnits": false, "label": "%"
          } },
          "annotations": { "horizontal": [
            { "label": "Goal > 99.95%", "value": 99.95 }
          ] }
        }
      },
      {
        "height": 6,
        "width": 12,
        "y": 0,
        "x": 12,
        "type": "metric",
        "properties": {
          "metrics": [
              [ { "expression": "m1 * 100", 
                  "label": "Error Rate", 
                  "id": "e1", "region": "us-east-1" } ],
              [ "AWS/ApiGateway", "4XXError", 
                "ApiName", "YourAwesomeApi", 
                { "id": "m1", "visible": false } ]
          ],
          "view": "timeSeries",
          "stacked": false,
          "region": "us-east-1",
          "stat": "Average",
          "period": 60,
          "title": "Error Rate",
          "yAxis": { "left": {
             "min": 0, "max": 10, "label": "%"
          } },
          "annotations": { "horizontal": [
            { "label": "Error Rate < 5%", "value": 5 }
          ] }
        }
      },
      {
        "type": "metric",
        "x": 0,
        "y": 6,
        "width": 12,
        "height": 6,
        "properties": {
          "metrics": [
              [ { "expression": "m1 / PERIOD(m1)", 
                  "label": "TPS", "id": "e1" } ],
              [ "AWS/ApiGateway", "Count", 
                "ApiName", "YourAwesomeApi",
                { "id": "m1", "period": 60, "visible": false } ]
          ],
          "view": "timeSeries",
          "stacked": false,
          "region": "us-east-1",
          "stat": "Sum",
          "period": 300,
          "title": "Request Rate",
          "yAxis": { "left": {
            "min": 0, "showUnits": false
          } },
          "annotations": { "horizontal": [
            { "label": "TPS > 20", "value": 20 }
          ] }
        }
      },
      {
        "type": "metric",
        "x": 12,
        "y": 6,
        "width": 12,
        "height": 6,
        "properties": {
          "metrics": [
              [ "AWS/ApiGateway", "Latency", 
                "ApiName", "YourAwesomeApi", 
                { "label": "p99 Latency" } ]
          ],
          "view": "timeSeries",
          "stacked": false,
          "region": "us-east-1",
          "stat": "p99",
          "period": 60,
          "start": "-P7D",
          "end": "P0D",
          "title": "Latency",
          "yAxis": { "left": {
            "min": 0, "label": "Milliseconds", "showUnits": false
          } },
          "annotations": { "horizontal": [
            { "label": "Latency < 1s", "value": 1000 }
          ] }
        }
      }
    ]
  }

Summary

Do:

• Count faults against your availability
• Have a canary to always have some traffic
• Measure availability and latency as close to the client as possible
• Have a dashboard that shows at minimum faults, errors, requests over time, and p99 latency

Don't:

• Count errors against your availability
• Aggregate latency on average, max or median.
• Measure availability or latency from your service implementation

Building for high availability: Security

Courtesy www.bluecoat.com I have a plan on doing a series of concerning things to think about when designing, building, and operating systems and services with reliability and high availability in mind. I will focus specifically on building services on a cloud services and my examples will generally be AWS, because that is what I know best. But most of the general principles should translate to any cloud provider of sufficient minimum functionality.

It is worth pointing out that the advice here is specifically for reliability and high availability. If, for instance, your goal is to be able to do rapid prototyping or being able to quickly go to market the advice would be very different (Perhaps I will do another series of posts on that once I am done with this topic). Sometimes it can be hard to explain to your Product Manager that even though somebody created a working prototype of something in less than a week it will still take 2 months to create the real thing, and this is one of the reasons why. As a preview of the difference between the two is that you can skip this entire section if you are only creating a prototype because security really does not matter for that (But be wary of the risk of the prototype making it to production, because then you would not have wanted to have skipped it).

There are many different things that can affect the availability of a service or site that you are building but probably the first and most important one is to make sure that your site is secure. Other failures, although severe would not result in the kind of disaster that a security failure could lead to. Not only could your entire service be taken offline or deleted, but all data you have stored could also be let loose on the dark web.

Defense in depth

The key for designing for security is defense in depth. You should not assume that you can establish a perimeter around your service and trust everything inside the service. Instead, you should consider how you can make each subcomponent as secure as possible. This will mean that if one of your components do get compromised it will not necessarily mean that your entire service or all your data is compromised. Additionally, by having each component always validating and logging access appropriately also means that any potential breach in one component can be detected earlier when an attacker unsuccessfully tries to extend the breach to other components.

The Least Privilege Principle

Each component should only have the minimum privileges needed to perform its job. If you have a component that needs permission to read a specific S3 bucket to perform its job, only grant read access to that specific bucket and not any bucket in your account nor allow it to do anything but reading from S3. Same thing goes to database access. This way if a component does get compromised only the data available to that component is potentially put at risk instead of all the data in your service.

Avoid fixed credentials

In AWS, most services allow you to grant permissions based on your execution environment such as EC2, ECS or Lambda execution roles without the need to distribute any credentials. This is a great feature that avoids the possibility of any credentials being lost in the wind and turning up in the wrong places.

If you do have to use fixed credentials such as to a RDBMS, then make sure that these credentials are automatically rotated often so that for instance ex-employees will not accidentally retain credentials to your systems.

In the case of AWS make sure you take advantage of the strong authentication options for the AWS console. And heed the advice of never using the root credentials for anything.

Limit your attack surface

Do not have any component of your service available from the internet that does not absolutely need to. Usually this would mean only your public API and your website being accessible through the internal.

Make sure that all your internal components are only available to the other internal components that need to communicate with them. In AWS you can accomplish this either through internal API:s inside of a VPC, or you can use AWS secured primitives to communicate between components such as queues or event buses.

If you need to be able to get access to the internal network for operational reasons make sure that all this access goes through a Bastion hosts that is truly locked. In AWS consider not using a Bastion host at all and instead rely on the System Manager Run & ECS Exec functionality to avoid the bastion host all together.

Avoid managing your own infrastructure and have a patching strategy

Using managed versions of almost any services means that when there is a problem with that service it is not your problem to fix it anymore, instead there is a specialist team available to handle the issue and you can just sit back and wait for the issue to be resolved. Granted, it does mean that you lose some control. But general the headache of needing to have a specialist on hand for every component you use in a complex system. It also means that for every component you have you need to have a comprehensive upgrade and patching strategy. In today's environment you must be prepared to be able to patch within hours of a critical vulnerability if not sooner or risk complete compromise of that component as evidenced most recently in the massive Exchange Service hack that has compromised at least 30k corporate email servers. If you are using managed services for your components the headache of patching, especially security vulnerabilities, is entirely handled for you.

This also extends to trying to use alternative methods of compute such as AWS Fargate and AWS Lambda to remove the burden of patching any OS that you are deploying your code on. That said, you are still responsible for patching your own code and making sure you are not relying on libraries that have known vulnerabilities in them. Using the Github code repository will provide you with automated vulnerability scanning for your code though if you are using standard dependency managers.

Encrypt everything

Always encrypt everything you save both in transit and at rest. Any intra component communication should always use TLS. Almost all AWS primitives that store data will have an option to encrypt data at rest using your own provided KMS key or at least a service owned key. Quite often though this functionality does need to be turned on explicitly, make sure you do this. Furthermore, make sure that the access to the keys for data that is sensitive is only provided to the components that need it. This is an extension of the Least Privilege Principle above. If an adversary does break into your system, this is another way that you can minimize the amount of data that is accessible and exfiltratable.

Pick the right tool for the job

When building a new system, it is important to pick the right language and framework because some are simply safer by design that others.

The first kind of language that is unsuitable is any language that contains unchecked primitives for direct memory access. This group includes languages such as C, C++ and obviously assembly language. The main danger with these kinds of languages is that it is just too easy to make a mistake and create a buffer overflow issue.

The second kind of language and or framework to avoid are languages that do too much "magic" to help you be productive. Most frameworks that involve Ruby or PHP fall in this category in my opinion. Not only do these languages lead to hard to maintain code, because it is very hard to understand the real ramifications of a change. Because so much is happening underneath the hood that you as a developer are probably not aware of, it is very hard to ensure that this "magic" is not doing something that will also lead to a security vulnerability.

Languages that I generally find suitable for building internet facing services include Java, C#, Python and Typescript. This is not an exhaustive list though and there are many more.

Avoid SQL

This is really a special case to call out in this section. The tip to avoid RDBMS:s will come up repeatedly during this serious of blog posts because they are generally not suitable for building high availability systems for many reasons. However, this specific tip is not specifically about RDBMS:s but about using any kind of database with the SQL query language. Regarding security, probably the most common reason for security breaches today is still SQL injection attacks and this kind of attack is only possible if your underlying database access language is SQL. There are almost always better choices for databases than SQL for your specific use case. Educate yourself on your options and pick anything that is not SQL. By doing this you also have the added benefit of removing even the possibility of being the target of this entire class of attacks.

Various other security related tips and tricks

This section contains some additional tips and tricks that might be more AWS specific for helping you to build secure services.

Be wary of deleting

Some cloud primitives such as S3 related to storing data allow you to not be able to delete or overwrite data. If you enable versioning in S3 and remove the permission to delete data, all together and instead use life cycle rules to expire data you can remove the threat of ransomware all together from that portion of your system. Similarly enable deletion protection to all other aspects of your infrastructure if available such as Cloud Formation stacks. This will protect you both from intentional vandal acts, but also unintentional accidents that could potentially take down your service by accidentally deleting critical infrastructure.

Safety of a crowd

When implementing your service perimeter take advantage of a managed components that sit between your service and the internet to protect yourself against both carefully crafted payloads designed to attack your service and also being able to weather the massive load of a DDOS attack. Examples of these kinds of services is not just AWS WAF, but also services such as Amazon S3, Amazon CloudFront and Amazon API Gateway. This does not include services that are simple load balancers though as these generally are provisioned to handle a single routing task explicitly and even though they do scale, it is at a slower rate and they also generally do not protect you against any kind of malicious payloads as the other services might.

Limit internet access from your components

Assuming the worst, that an adversary has broken into your system, one way that you can limit the damage that can be done is to remove access to the internet from inside your system. Quite often a service only needs to be accessible from the internet through a load balancer and all the internal components only really need to talk to other services of your cloud provider. If this is the case for you, using AWS PrivateLink for accessing the AWS services needed and otherwise have no internet connectivity from your internal service network will greatly increase the difficulty of any attacker to exfiltrate any data that they may have gained access to.

Summary

Do:

• Implement defense in depth
• Encrypt everything
• Limit attack surface
• Use the right language and framework

Don't:

• Manage your own infrastructure if you can avoid it
• Use fixed credentials
• Use SQL