Let's say you've got a fully hypermedia driven API. Consumers have to navigate three reources, via following hypermedia, until they can get to the resource they want. Is there any reason a client could not cache these steps temporarily and go directly to the resource they want?
I know the goal of REST is to decouple clients and servers, but if you've got 5 web requests going on behind the scenes the user experience could be poor waiting for all this to happen.
The worst case I can think of is that a cached URL gets changed. And so the client will just start from the entrypoint again and cache the steps.
Caching on the client side is going to be very important for a lot of well performing Hypermedia clients. Here is some more specific guidance straight from Fielding's dissertation:
The advantage of adding cache constraints is that they have the potential to partially or completely eliminate some interactions, improving efficiency, scalability, and user-perceived performance by reducing the average latency of a series of interactions. The trade-off, however, is that a cache can decrease reliability if stale data within the cache differs significantly from the data that would have been obtained had the request been sent directly to the server.
The are trade offs but event a short time frame for caching will greatly improve performance. Ideally the Hypermedia API will provide caching guidance. This could be done in the same manner that HTML caching works with the browser and Expires and Cache-Control headers.
Also if the resource has moved then the API should inform you with the proper 301 Moved Permanently response.
Related
While designing rest API's I time to time have challenge to deal with batch operations (e.g. delete or update many entities at once) to reduce overhead of many tcp client connections. And in particular situation problem usually solves by adding custom api method for specific operation (e.g. POST /files/batchDelete which accepts ids at request body) which doesn't look pretty from point of view of rest api design principles but do the job.
But for me general solution for the problem still desirable. Recently I found Google Cloud Storage JSON API batching documentation which for me looks like pretty general solution. I mean similar format may be used for any http api, not just google cloud storage. So my question is - does anybody know kind of general standard (standard or it's draft, guideline, community effort or so) of making multiple API calls combined into one HTTP request?
I'm aware of capabilities of http/2 which include usage of single tcp connection for http requests but my question is addressed to application level. Which in my opinion still make sense because despite of ability to use http/2 taking that on application level seems like the only way to guarantee that for any client including http/1 which is currently the most used version of http.
TL;DR
REST nor HTTP are ideal for batch operations.
Usually caching, which is one of RESTs constraints, which is not optional but mandatory, prevents batch processing in some form.
It might be beneficial to not expose the data to update or remove in batch as own resources but as data elements within a single resource, like a data table in a HTML page. Here updating or removing all or parts of the entries should be straight forward.
If the system in general is write-intensive it is probably better to think of other solutions such as exposing the DB directly to those clients to spare a further level of indirection and complexity.
Utilization of caching may prevent a lot of workload on the server and even spare unnecessary connecctions
To start with, REST nor HTTP are ideal for batch operations. As Jim Webber pointed out the application domain of HTTP is the transfer of documents over the Web. This is what HTTP does and this is what it is good at. However, any business rules we conclude are just a side effect of the document management and we have to come up with solutions to turn this document management side effects to something useful.
As REST is just a generalization of the concepts used in the browsable Web, it is no miracle that the same concepts that apply to Web development also apply to REST development in some form. Thereby a question like how something should be done in REST usually resolves around answering how something should be done on the Web.
As mentioned before, HTTP isn't ideal in terms of batch processing actions. Sure, a GET request may retrieve multiple results, though in reality you obtain one response containing links to further resources. The creation of resources has, according to the HTTP specification, to be indicated with a Location header that points to the newly created resource. POST is defined as an all purpose method that allows to perform tasks according to server-specific semantics. So you could basically use it to create multiple resources at once. However, the HTTP spec clearly lacks support for indicating the creation of multiple resources at once as the Location header may only appear once per response as well as define only one URI in it. So how can a server indicate the creation of multiple resources to the server?
A further indication that HTTP isn't ideal for batch processing is that a URI must reference a single resource. That resource may change over time, though the URI can't ever point to multiple resources at once. The URI itself is, more or less, used as key by caches which store a cacheable response representation for that URI. As a URI may only ever reference one single resource, a cache will also only ever store the representation of one resource for that URI. A cache will invalidate a stored representation for a URI if an unsafe operation is performed on that URI. In case of a DELETE operation, which is by nature unsafe, the representation for the URI the DELETE is performed on will be removed. If you now "redirect" the DELETE operation to remove multiple backing resources at once, how should a cache take notice of that? It only operates on the URI invoked. Hence even when you delete multiple resources in one go via DELETE a cache might still serve clients with outdated information as it simply didn't take notice of the removal yet and its freshness value would still indicate a fresh-enough state. Unless you disable caching by default, which somehow violates one of REST's constraints, or reduce the time period a representation is considered fresh enough to a very low value, clients will probably get served with outdated information. You could of course perform an unsafe operation on each of these URIs then to "clear" the cache, though in that case you could have invoked the DELETE operation on each resource you wanted to batch delete itself to start with.
It gets a bit easier though if the batch of data you want to remove is not explicitly captured via their own resources but as data of a single resource. Think of a data-table on a Web page where you have certain form-elements, such as a checkbox you can click on to mark an entry as delete candidate and then after invoking the submit button send the respective selected elements to the server which performs the removal of these items. Here only the state of one resource is updated and thus a simple POST, PUT or even PATCH operation can be performed on that resource URI. This also goes well with caching as outlined before as only one resource has to be altered, which through the usage of unsafe operations on that URI will automatically lead to an invalidation of any stored representation for the given URI.
The above mentioned usage of form-elements to mark certain elements for removal depends however on the media-type issued. In the case of HTML its forms section specifies the available components and their affordances. An affordance is the knowledge what you can and should do with certain objects. I.e. a button or link may want to be pushed, a text field may expect numeric or alphanumeric input which further may be length limited and so on. Other media types, such as hal-forms, halform or ion, attempt to provide form representations and components for a JSON based notation, however, support for such media-types is still quite limited.
As one of your concerns are the number of client connections to your service, I assume you have a write-intensive scenario as in read-intensive cases caching would probably take away a good chunk of load from your server. I.e. BBC once reported that they could reduce the load on their servers drastically just by introducing a one minute caching interval for recently requested resources. This mainly affected their start page and the linked articles as people clicked on the latest news more often than on old news. On receiving a couple of thousands, if not hundred thousands, request per minute they could, as mentioned before, reduce the number of requests actually reaching the server significantly and therefore take away a huge load on their servers.
Write intensive use-cases however can't take benefit of caching as much as read-intensive cases as the cache would get invalidated quite often and the actual request being forward to the server for processing. If the API is more or less used to perform CRUD operations, as so many "REST" APIs do in reality, it is questionable if it wouldn't be preferable to expose the database directly to the clients. Almost all modern database vendors ship with sophisticated user-right management options and allow to create views that can be exposed to certain users. The "REST API" on top of it basically just adds a further level of indirection and complexity in such a case. By exposing the DB directly, performing batch updates or deletions shouldn't be an issue at all as through the respective query languages support for such operations should already be build into the DB layer.
In regards to the number of connections clients create: HTTP from 1.0 on allows the reusage of connections via the Connection: keep-alive header directive. In HTTP/1.1 persistent connections are used by default if not explicitly requested to close via the respective Connection: close header directive. HTTP/2 introduced full-duplex connections that allow many channels and therefore requests to reuse the same connections at the same time. This is more or less a fix for the connection limitation suggested in RFC 2626 which plenty of Web developers avoided by using CDN and similar stuff. Currently most implementations use a maximum limit of 100 channels and therefore simultaneous downloads via a single connections AFAIK.
Usually opening and closing a connection takes a bit of time and server resources and the more open connections a server has to deal with the more a system may suffer. Though open connections with hardly any traffic aren't a big issue for most servers. While the connection creation was usually considered to be the costly part, through the usage of persistent connections that factor moved now towards the number of requests issued, hence the request for sending out batch-requests, which HTTP is not really made for. Again, as mentioned throughout the post, through the smart utilization of caching plenty of requests may never reach the server at all, if possible. This is probably one of the best optimization strategies to reduce the number of simultaneous requests, as probably plenty of requests might never reach the server at all. Probably the best advice to give is in such a case to have a look at what kind of resources are requested frequently, which requests take up a lot of processing capacity and which ones can easily get responded with by utilizing caching options.
reduce overhead of many tcp client connections
If this is the crux of the issue, the easiest way to solve this is to switch to HTTP/2
In a way, HTTP/2 does exactly what you want. You open 1 connection, and using that collection you can send many HTTP requests in parallel. Unlike batching in a single HTTP request, it's mostly transparent for clients and response and requests can be processed out of order.
Ultimately batching multiple operations in a single HTTP request is always a network hack.
HTTP/2 is widely available. If HTTP/1.1 is still the most used version (this might be true, but gap is closing), this has more to do with servers not yet being set up for it, not clients.
Should we use http caching only for static stuff?
Or also in API responses could be using caching headers if data from API is not static? it can be changed by application's users.
Caching is needed to gain performance but at the same time it increases the likelihood of the data being outdated. It's true for static resources as well. So if your app is under high load and you want to increase the speed - you may sacrifice up-to-date data for gain in performance.
Note, though, that client side needs to respect caching headers. We often work with browsers - they have it all figured out, but if our client is another service, then you need to ensure that it doesn't ignore the headers. This won't be for free - code will need to be written for this to happen.
Your cache may also be public or private. If it's public (any client is allowed to see the content), you may configure a reverse proxy (like nginx) between your server and the clients. Nginx can be set up to cache results (it also understands cache headers). So it may take off some load from your application by not letting requests through and instead returning cached copies.
TL;DR : scroll down to the last paragraph.
There is a lot of talk about best practices when defining RESTful APIs: what HTTP methods to support, which HTTP method to use in each case, which HTTP status code to return, when to pass parameters in the query string vs. in the path vs. in the content body vs. in the headers, how to do versioning, result set limiting, pagination, etc.
If you are already determined to make use of best practices, there are lots of questions and answers out there about what is the best practice for doing any given thing. Unfortunately, there appears to be no question (nor answer) as to why use best practices in the first place.
Most of the best practice guidelines direct developers to follow the principle of least surprise, which, under normal circumstances, would be a good enough reason to follow them. Unfortunately, REST-over-HTTP is a capricious standard, the best practices of which are impossible to implement without becoming intimately involved with it, and the drawback of intimate involvement is that you tend to end up with your application being very tightly bound to a particular transport mechanism. So, some people (like me) are debating whether the benefit of "least surprise" justifies the drawback of littering the application with REST-over-HTTP concerns.
A different approach examined as an alternative to best practices suggests that our involvement with HTTP should be limited to the bare minimum necessary in order to get an application-defined payload from point A to point B. According to this approach, you only use a single REST entry point URL in your entire application, you never use any HTTP method other than HTTP POST, never return any HTTP status code other than HTTP 200 OK, and never pass any parameter in any way other than within the application-specific payload of the request. The request will either fail to be delivered, in which case it is the responsibility of the web server to return an "HTTP 404 Not Found" to the client, or it will be successfully delivered, in which case the delivery of the request was "HTTP 200 OK" as far as the transport protocol is concerned, and anything else that might go wrong from that point on is exclusively an application concern, and none of the transport protocol's business. Obviously, this approach is kind of like saying "let me show you where to stick your best practices".
Now, there are other voices that say that things are not that simple, and that if you do not follow the RESTful best practices, things will break.
The story goes that for example, in the event of unauthorized access, you should return an actual "HTTP 401 Unauthorized" (instead of a successful response containing a json-serialized UnauthorizedException) because upon receiving the 401, the browser will prompt the user of credentials. Of course this does not really hold any water, because REST requests are not issued by browsers being used by human users.
Another, more sophisticated way the story goes is that usually, between the client and the server exist proxies, and these proxies inspect HTTP requests and responses, and try to make sense out of them, so as to handle different requests differently. For example, they say, somewhere between the client and the server there may be a caching proxy, which may treat all requests to the exact same URL as identical and therefore cacheable. So, path parameters are necessary to differentiate between different resources, otherwise the caching proxy might only ever forward a request to the server once, and return cached responses to all clients thereafter. Furthermore, this caching proxy may need to know that a certain request-response exchange resulted in a failure due to a particular error such as "Permission Denied", so as to again not cache the response, otherwise a request resulting in a temporary error may be answered with a cached error response forever.
So, my questions are:
Besides "familiarity" and "least surprise", what other good reasons are there for following REST best practices? Are these concerns about proxies real? Are caching proxies really so dumb as to cache REST responses? Is it hard to configure the proxies to behave in less dumb ways? Are there drawbacks in configuring the proxies to behave in less dumb ways?
It's worth considering that what you're suggesting is the way that HTTP APIs used to be designed for a good 15 years or so. API designers are tending to move away from that approach these days. They really do have their reasons.
Some points to consider if you want to avoid using ReST over HTTP:
ReST over HTTP is an efficient use of the HTTP/S transport mechanism. Avoiding the ReST paradigm runs the risk of every request / response being wrapped in verbose envelopes. SOAP is an example of this.
ReST encourages client and server decoupling by putting application semantics into standard mechanisms - HTTP and XML/JSON (or others data formats). These protocols and standards are well supported by standard libraries and have been built up over years of experience. Sure, you can create your own 'unauthorized' response body with a 200 status code, but ReST frameworks just make it unnecessary so why bother?
ReST is a design approach which encourages a view of your distributed system which focuses on data rather than functionality, and this has a proven a useful mechanism for building distributed systems. Avoiding ReST runs the risk of focusing on very RPC-like mechanisms which have some risks of their own:
they can become very fine-grained and 'chatty'
which can be an inefficient use of network bandwidth
which can tightly couple client and server, through introducing stateful-ness and temporal coupling beteween requests.
and can be difficult to scale horizontally
Note: there are times when an RPC approach is actually a better way of breaking down a distributed system than a resource-oriented approach, but they tend to be the exceptions rather than the rule.
existing tools for developers make debugging / investigations of ReSTful APIs easier. It's easy to use a browser to do a simple GET, for example. And tools such as Postman or RestClient already exist for more complex ReST-style queries. In extreme situations tcpdump is very useful, as are browser debugging tools such as firebug. If every API call has application layer semantics built on top of HTTP (e.g. special response types for particular error situations) then you immediately lose some value from some of this tooling. Building SOAP envelopes in PostMan is a pain. As is reading SOAP response envelopes.
network infrastructure around caching really can be as dumb as you're asking. It's possible to get around this but you really do have to think about it and it will inevitably involve increased network traffic in some situations where it's unnecessary. And caching responses for repeated queries is one way in which APIs scale out, so you'll likely need to 'solve' the problem yourself (i.e. reinvent the wheel) of how to cache repeated queries.
Having said all that, if you want to look into a pure message-passing design for your distributed system rather than a ReSTful one, why consider HTTP at all? Why not simply use some message-oriented middleware (e.g. RabbitMQ) to build your application, possibly with some sort of HTTP bridge somewhere for Internet-based clients? Using HTTP as a pure transport mechanism involving a simple 'message accepted / not accepted' semantics seems overkill.
REST is intended for long-lived network-based applications that span multiple organizations. If you don’t see a need for the constraints, then don’t use them. -- Roy T Fielding
Unfortunately, there appears to be no question (nor answer) as to why use best practices in the first place.
When in doubt, go back to the source
Fielding's dissertation really does quite a good job at explaining how the REST architectural constraints ensure that you don't destroy the properties those constraints are designed to protect.
Keep in mind - before the web (which is the reference application for REST), "web scale" wasn't a thing; the notion of a generic client (the browers) that could discover and consume thousands of customized applications (provided by web servers) had not previously been realized.
According to this approach, you only use a single REST entry point URL in your entire application, you never use any HTTP method other than HTTP POST, never return any HTTP status code other than HTTP 200 OK, and never pass any parameter in any way other than within the application-specific payload of the request.
Yup - that's a thing, it's called RPC; you are effectively taking the web, and stripping it down to a bare message transport application that just happens to tunnel through port 80.
In doing so, you have stripped away the uniform interface -- you've lost the ability to use commodity parts in your deployment, because nobody can participate in the conversation unless they share the same interpretation of the message data.
Note: that's doesn't at all imply that RPC is "broken"; architecture is about tradeoffs. The RPC approach gives up some of the value derived from the properties guarded by REST, but that doesn't mean it doesn't pick up value somewhere else. Horses for courses.
Besides "familiarity" and "least surprise", what other good reasons are there for following REST best practices?
Cheap scaling of reads - as your offering becomes more popular, you can service more clients by installing a farm of commodity reverse-proxies that will serve cached representations where available, and only put load on the server when no fresh representation is available.
Prefetching - if you are adhering to the safety provisions of the interface, agents (and intermediaries) know that they can download representations at their own discretion without concern that the operators will be liable for loss of capital. AKA - your resources can be crawled (and cached)
Similarly, use of idempotent methods (where appropriate) communicates to agents (and intermediaries) that retrying the send of an unacknowledged message causes no harm (for instance, in the event of a network outage).
Independent innovation of clients and servers, especially cross organizations. Mosaic is a museum piece, Netscape vanished long ago, but the web is still going strong.
Of course this does not really hold any water, because REST requests are not issued by browsers being used by human users.
Of course they are -- where do you think you are reading this answer?
So far, REST works really well at exposing capabilities to human agents; which is to say that the server side is so ubiquitous at this point that we hardly think about it any more. The notion that you -- the human operator -- can use the same application to order pizza, run diagnostics on your house, and remote start your car is as normal as air.
But you are absolutely right that replacing the human still seems a long ways off; there are various standards and media types for communicating semantic content of data -- the automated client can look at markup, identify a phone number element, and provide a customized array of menu options from it -- but building into agents the sorts of fuzzy intelligence needed to align offered capabilities with goals, or to recover from error conditions, seems to be a ways off.
Assume the following scenario A web application serves up resources through a RESTful API. A number of clients consume this API. The goal is to keep the data on the clients synchronized with the web application (in both directions).
The easiest way to do this is to ask the API if any of the resources have changed since the client last synchronized with the API. This means that the client needs to ask the API for the appropriate resources accompanied by timestamp (to see if the data needs to be updated). This seems to me like the approach with the least overhead in terms of needless consumption of bandwidth.
However, I have the feeling that this approach has a few downsides in terms of design and responsibilities. For example, the API shouldn't have to deal with checking whether the resources are out of date. It seems that the only responsibility of the API should be to serve up the resources when asked without having to deal with the updating aspect. By following this second approach, the client would ask for a lot of data every time it wants to update its data to keep it synchronized with the web application. In other words, the client would check whether the data it got back is newer than the locally stored data. If this process takes place every few minutes, this might become a significant burden for the system.
Am I seeing this correctly or is there a middle road that I am overlooking?
This is a pretty common problem, and a RESTful approach can help you solve it. HTTP (the application protocol typically used to build RESTful services) supports a variety of techniques that can be used to keep API clients in sync with the data on the server side.
If the client receives a Last-Modified or E-Tag header in a HTTP response, it may use that information to make conditional GET calls in the future. This allows the server to quickly indicate with a 304 – Not Modified response that the client’s previously stored representation of the resource is still valid and accurate. This will allow the server (or even better, an intermediate proxy or cache server) to be as efficient as possible in how it responds to the client’s requests, potentially reducing costly round-trips to a back-end data store.
If a response contains a Last-Modified header and the client wishes to take advantage of the performance optimization available with it, they must include an If-Modified-Since directive in a subsequent GET call to the same URI, passing in the same timestamp value it received. This instructs the server to only GET the information from the authoritative back-end source if it knows it has changed since that time. Your server will have to be built to support this technique, of course.
A similar principle applies to E-Tag headers. An E-Tag is a simple hash code representing a specific state of the resource at a particular point in time. If the resource changes in any way, so does its E-Tag value. If the client sees an E-Tag in a response it should pass it in subsequent GET requests to the same URI, thereby allowing the server to quickly determine if the client has an up-to-date representation of the resource.
Finally, you should probably look at the long polling technique to reduce the number of repeated GET requests issued by your clients to the server. In essence, the trick is to issue very long GET requests to the server to watch for server data changes. The GET will not return a response until either the data has changed or the very long timeout fires. If the latter, the client just re-issues the same long-lived request to watch for changes again. See also topics like Comet and Web Sockets which are similar in approach.
I've been considering the advantages of REST services, the whole statelessness and session affinity "stuff". What strikes me is that if you have multiple deployed versions of your service on a number of machines in your infrastructure, and they all act on a given resource, where is the state of that resource stored?
Would it make sense to have a single host in the infrastructre that utilises a distributed cache, and any state that is change inside a service, it simply fetches/puts to the cache? This would allow any number of deployed services for loading balancing reasons to all see the same state views of resources.
If you're designing a system for high load (which usually implies high reliability), having a single point of failure is never a good idea. If the service providing the consistent view goes down, at best your performance decreases drastically as the database is queried for everything and at worst, your whole application stops working.
In your question, you seem to be worried about consistency. If there's something to be learned about eBay's architecture, it's that there is a trade-off to be made between availability/redundancy/performance vs consistency. You may find 100% consistency is not required and you can get away with a little "chaos".
A distributed cache (like memcache) can be used as a backing for a distributed hashtable which have been used extensively to create scalable infrastructures. If implemented correctly, caches can be redundant and caches can join and leave the ring dynamically.
REST is also inherently cacheable as the HTTP layer can be cached with the appropriate use of headers (ETags) and software (e.g. Squid proxy as a Reverse proxy). The one drawback of specifying caching through headers is that it relies on the client interpreting and respecting them.
However, to paraphrase Phil Karlton, caching is hard. You really have to be selective about the data that you cache, when you cache it and how you invalidate that cache. Invalidating can be done in the following ways:
Through a timer based means (cache for 2 mins, then reload)
When an update comes in, invalidating all caches containing the relevant data.
I'm partial to the timer based approach as its simpler to implement and you can say with relative certainty how long stale data will live in the system (e.g. Company details will be updated in 2 hours, Stock prices will be updated in 10 seconds).
Finally, high load also depends on your use case and depending on the amount of transactions none of this may apply. A methodology (if you will) may be the following:
Make sure the system is functional without caching (Does it work)
Does it meet performance criteria (e.g. requests/sec, uptime goals)
Optimize the bottlenecks
Implement caching where required
After all, you may not have a performance problem in the first place and you may able to get away with a single database and a good back up strategy.
I think the more traditional view of load balancing web applications is that you would have your REST service on multiple application servers and they would retrieve resource data from single database server.
However, with the use of hypermedia, REST services can easily vertically partition the application so that some resources come from one service and some from another service on a different server. This would allow you to scale to some extent, depending on your domain, without have a single data store. Obviously with REST you would not be able to do transactional updates across these services, but there are definitely scenarios where this partitioning is valuable.
If you are looking at architectures that need to really scale then I would suggest looking at Greg Young's stuff on CQS Architecture (video) before attempting to tackle the problems of a distributed cache.