I am writing a Java EE application which is supposed to consume SAP BAPIs/RFC using JCo and expose them as web-services to other downstream systems. The application needs to scale to huge volumes in scale of tens of thousands and thousands of simultaneous users.
I would like to have suggestions on how to design this application so that it can meet the required volume.
Its good that you are thinking of scalability right from the design phase. Martin Abbott and Michael Fisher (PayPal/eBay fame) layout a framework called AKF Scale for scaling web apps. The main principle is to scale your app in 3 axis.
X-axis: Cloning of services/ data such that work can be easily distributed across instances. For a web app, this implies ability to add more web servers (clustering).
Y-axis: separation of work responsibility, action or data. So for example in your case, you could have different API calls on different servers.
Z-Axis: separation of work by customer or requester. In your case you could say, requesters from region 1 will access Server 1, requesters from region 2 will access Server 2, etc.
Design your system so that you can follow all 3 above if you need to. But when you initially deploy, you may not need to use all three methods.
You can checkout the book "The Art of Scalability" by the above authors. http://amzn.to/oSQGHb
A final answer is not possible, but based on the information you provided this does not seem to be a problem as long as your application is stateless so that it only forwards requests to SAP and returns the responses. In this case it does not maintain any state at all. If it comes to e.g. asynchronous message handling, temporary database storage or session state management it becomes more complex. If this is true and there is no need to maintain state you can easily scale-out your application to dozens of application servers without changing your application architecture.
In my experience this is not necessarily the case when it comes to SAP integration, think of a shopping cart you want to fill based on products available in SAP. You may want to maintain this cart in your application and only submit the final cart to SAP. Otherwise you end up building an e-commerce application inside your backend.
Most important is that you reduce CPU utilization in your application to avoid a 'too-large' cluster and to reduce all kinds of I/O wherever possible, e.g. small SOAP messages to reduce network I/O.
Furthermore, I recommend to design a proper abstraction layer on top of JCo including the JCO.PoolManager for connection pooling. You may also need a well-thought-out authorization concept if you work with a connection pool managed by only one technical user.
Just some (not well structured) thoughts...
Related
I'm building a microservices architecture that should deal with:
Direct database access
Call to external legacy services
I can think about 2 caching strategies, but can't figure out what is the best considering that I will not have control on what other people could do across the layers.
Caching at application level (#Cacheable)
I only provide a caching feature that everyone can use, by enforcing the spring.cache.redis.key-prefix to the microservice name to limit conflicting keys.
PRO: most flexible way
CONS:
No control over cache except for maximum space: people would just create new cache entries
No control over cache invalidation: we don't know what kind of data is actually stored so if, for example, a legacy system needs to be reloaded we cannot empty some cache keys
Possible redundancy: as caching is at application layer it could happen that microservices store about the same data. While I could have control on the database (one MS should own its own db or at least a subset of tables) I can't guarantee about the legacy SOAP layer
Caching at service layer (connectors)
I don't provide a caching feature but I provide custom soap connectors that will/will not cache response based on a configuration that I will provide (could also be a blacklist/whitelist)
PROS:
cache is controlled
easy to invalidate
CONS:
need to update connectors each time a cache policy changes
dependency between development and architecture
edit: I need suggestion about the theoretical approach, not about a specific technology.
I suppose you should build different microservices (apis) to deal with different set of responsibilities. Like , you can have a one microservice which deals with legacy and other one which deals with database. In order for these two microservices to communicate, you can have a message broker architecture like apache kafka (hazelcast being cost effective or Rabbit MQ).
Communication between these two microservices can be event driven as well.
Once you decide this, then you can finalize where to place your cache.
You will need to place cache at application level and not service level if there is an UI where you are showing these values.
It seems that in the traditional microservice architecture, each service gets its own database with a different understanding of the data (described here). Sometimes it is considered permissible for databases to duplicate data. For instance, the "Users" service might know essentially everything about a user, whereas the "Posts" service might just store primary keys and usernames (so that the author of a post can have their name displayed, for instance). This page talks about eventual consistency, sources of truth, and other related concepts when data is duplicated. I understand that microservice architectures sometimes include a shared database, but most places I look suggest that this is a rare strategy.
As for why each service typically gets its own database, all I've seen so far is "so that each service owns its own resources," but I'm not convinced that a) the service layer in any way "owns" the persisted resources accessed through the database to begin with, or that b) services even need to own the resources they require rather than accessing necessary subsets of the master resources through a shared database.
So what are some of the justifications that each service in a microservice architecture should get its own database?
There are a few reasons why it does make sense to use a separate database per micro-service. Some of them are:
Scaling
Splitting your domain in micro-services is fine. You can scale your particular micro-service on the deployed web-server on demand or scale out as needed. That it obviously one of the benefits when using micro-services. More importantly you can have micro-service-1 running for example on 10 servers as it demands this traffic but micro-service-2 only requires 1 web-server so you deploy it on 1 server. The good thing is that you control this and you can manage your computing resources like in order to save money as Cloud providers are not cheap.
Considering this what about the database?
If you have one database for multiple services you could not do this. You could not scale the databases individually as they would be on one server.
Data partitioning to reduce size
Automatically as you split your domain in micro-services with each containing 1 database you split the amount of data that is stored in each database. Ideally if you do this you can have smaller database servers with less computing power and/or RAM.
In general paying for multiple small servers is cheaper then one large one.
So in this case you could make use of this fact and save some resources as well.
If it happens that the already spited by domain database have large amount of data techniques like data sharding or data partitioning could be applied additional, but this is another topic.
Which db technology fits the business requirement
This is very important pro fact for having multiple databases. It would allow you to pick the database technology which fits your Business requirement best in order to get the best performance or usage of it. For example some specific micro-service might have some Read-heavy operations with very complex filter options and a full text search requirement. Using Elastic Search in this case would be a good choice. Some other micro-service might use SQL Server as it requires SQL specific features like transnational behavior or similar. If for some reason you have one database for all services you would be stuck with the particular database technology which might not be so performant for those requirement. It is a compromise for sure.
Developer discipline
If for some reason you would have a couple micro-services which would share their database you would need to deal with the human factor. The developers would need to be disciplined to not cross domains and access/modify the other micro-services database(tables, collections and etc) which would be hard to achieve and control. In large organisations with a lot of developers this could be a serious problem. With a hard/physical split this is not an issue.
Summary
There are some arguments for having database per micro-service but also some against it. In general the guidelines and suggestions when using micro-services are to have the micro-service together with its data autonomous in order to work independent in Ideal case(this is not the case always). It is defiantly a compromise as well as using micro-services in general. As always the rule is the rule but there are exceptions to it. Micro-services architecture is flexible and very dependent of your Domain needs and requirements. If you and your team identify that it makes sense to merge multiple micro-service databases to 1 and that it solves a lot of your problems then go for it.
Microservices
Microservices advocate design constraints where each service is developed, deployed and scaled independently. This philosophy is only possible if you have database per service. How can i continue my business if i have DB failure and what steps i can take to mitigate this?DB is essential part of any enterprise application. I agree there are different number of challenges when services has its own databases.
Why Independent database?
Unlike other approaches this approach not only keeps your code-base clean and extendable but you truly omit the single point of failure in your business. To achieve this services sometimes can have duplicated data as well, as long as my service is autonomous and services can only be autonomous if i have database per service.
From business point of view, Lets take eCommerce application. you have microserivces like Booking, Order, Payment, Recommendation , search and so on. Database is shared. What happens if the DB is down ? All your services are down ! and there is no point using Microservies architecture other than you have clean code base.
If you have each service having it's own database , i don't mind if my recommendation service is not working but i can still search and book the order and i haven't lost the customer. that's the whole point.
It comes at cost and challenges, but in longer run it pays off.
SQL / NoSQL
Each service has it's own needs. To get the best performance I can use SQL for payment service (transaction) and I can use (I should) NoSQL for recommendation service. Shared database wouldn't help me in this case. In modern cloud Architectures like CQRS, Event Sourcing, Materialized views, we sometimes use 2 different databases for same service to get the performance out of it.
Again Database per service is not only about resources or how much data should it own. But we really have to see the bigger picture. Yes we have certain practices how much data and duplication is good or bad but that's another debate.
Hope that helps !
I'm porting a huge application to Windows Azure. It will have a web service frontend and a processing backend. So far I thought I would use web roles for servicing client requests and worker roles for backend processing.
Managing two kinds of roles seems problematic - I'll need to decide how to scale two kinds of roles and also I'll need several (at least two) instances of each to ensure reasonable fault tolerance and this will slightly increase operational costs. Also in my application client requests are rather lightweight and backend processing is heavyweight, so I'd expect that backend processing would consume far more processing power than servicing client requests.
This is why I'm thinking of using web roles for everything - just spawn threads and do both servicing requests and backend processing in each instance. This will make the role more complicated but will I guess simplify management. I'll have more instances of a uniform role and better fault tolerance.
Is it a good idea to reuse web roles for backend processing? What drawbacks should I expect?
Sounds like you already have a pretty good idea of what to think about when using multiple roles:
Cost for 2 instances to meet SLA (although some background tasks really don't need SLA if the end user doesn't see the impact)
Separate scale units
However: If you run everything in one role, then everything scales together. If, say, you have an administrative web site on port 8000, you might have difficulty reaching it if your user base is slamming the main site on port 80 with traffic.
I blogged about combining web and worker roles, here, which goes into a bit more detail along what we're discussing here. Also, as of some time in March, the restriction of 5 endpoints per role was lifted - see my blog post here for just how far you can push endpoints now. Having this less-restrictive endpoint model really opens up new possibilities for single-role deployments.
From what I understand your are asking if it makes sense to consolidate service layers so that you only have to deal with a single layer. At a high level, I think that makes sense. The simpler the better, as long as it's not so simple that you can't meet your primary objectives.
If your primary objective is performance, and the calls to your services are inline (meaning that the caller is waiting for an answer), then consolidating the layers may help you in achieving greater performance because you won't have to deal with the overhead of additional network latency of additional physical layers. You can use the Task Parallel Library (TPL) to implement your threading logic.
If your primary objective is scalability, and the calls to your services are out-of-band (meaning that the caller implements a fire-and-forget pattern), then using processing queues and worker roles may make more sense. One of the tenets of cloud computing is loosely coupled services. While you have more maintenance work, you also have more flexibility to grow your layers independendly. Your worker roles could also use the TPL mentioned above so that you can deploy your worker roles on larger VMs (say with 4CPUs, or 8), which would keep the number of instances deployed to a minimum.
My 2 cents. :)
I would suggest you to develop them as separated roles: a web role and a worker role, and then just combine them into a single web role.
this way, in the future you can easaly convert to real separated roles, if needed.
for more details:
http://www.31a2ba2a-b718-11dc-8314-0800200c9a66.com/2010/12/how-to-combine-worker-and-web-role-in.html
http://wely-lau.net/2011/02/25/combining-web-and-worker-role-by-utilizing-worker-role-concept/
I use several loadtesting tools (Loadrunner, JMeter, NeoLoad) to performance test different applications. Im wondering if it is possible to monitor all layers of an application stack so for example. Say i have the following data chain.
Loadbalancer <-x-> Application Server <-x-> RMI <-x-> Java Application <-x-> MQ <-x-> Legacy application <-x-> Database
Where i have marked the x in the chain i am interested in monitoring, for example avg responsetimes.
Obviously we could simply create a wrapper on all endpoints which would gather the statistics for us and maybe we could import it into loadrunner or other loadtesting tools and sideline hem with the tools inbuilt performance statistics, but maybe there is tools/applications which already does this?
If not, how should we proceed, in order to gather this kind of statistics?
The standard for this was supposed to be Application Response Measurement (ARM). It was a cross language set of APIs that did just what you were looking for. The issue is that the products that implement this spec all tend to be big, expensive "enterprise" level monitoring tools. Think multi-week installs, consultants, more infrastructure and lots of buzzwords.
Still, if this is a mission critical app with a mission critical budget, this may be what you need. But you may be able to build your own that does just enough without too much effort. A quick search turns up at least one open source ARM implementation if you still want to use that API.
Another option is to simply to have transactions you can run against each tier of the system to check general responsiveness. For example you can have a static web page on the LB, a no-op tx on the app server, a "hello" servlet on the Java app, put a message directly on the queue, etc. During a performance / load test, these could be hit directly by the load testing tool or you could write a wrapper servlet / application call that does this as a single HTTP (RMI?) call. Running these a few times a minute won't add too much load to the system, but it should help you pinpoint which tier is slower. The nice thing about this approach is that it also works in production, just watch out for security issues.
For single user kind of test, where you know you have problem (e.g. this tx is "slow"), I have also had pretty good luck with network tracing. It's very tedious, but when you aren't sure what tier is slow, starting up a network trace on a few machines and running a single tx usually gives a good idea of what the system is doing.
I have handled this decomposition a number of ways in the past. The first is at a very low level using protocol analyzer dumped data to find the time points where a conversation leaves tier X and enters tier Y. The second method is through the use of log examination for the various tiers. Something that can make your examination quite usefule in this case is a common log server for all of your components (syslog, Rsyslog, etc....) and a nice log parsing tool, such as the freely available Microsoft Logparser. The third method utilization of the audit trail for an application stored in the database. You may find this when working on enterprise services bus style applications which have a consumer/producer model and a bus to pass information rather than a direct connection. The audit trails I have seen are typically stored in a database and allow the tracking of an individual transaction through the entire application infrastructure. Your Load balancer, as a network device, may be out of the hunt on this one.
Note, if you go the protocol analyzer or log route, then be sure and synchronize all of your source information devices to a common time server. Having one of your collectors (analyzer, app log) off on a time stamp basis can really be a hair pulling experience when you get into the analysis phase.
As to how you move from your collected data into LoadRunner, that part is very mechanical. The Analysis program supports an interface to import external datapoints. The format is very specific and is documented in both help and the online docs. This import process works very well, as I often have to use it for collection of statistics from hosts which I do not have direct monitoring access to, but which need to be included as a part of the monitored test infrastructure.
James Pulley
Moderator (YahooGroups LoadRunner, Advanced-Loadrunner; GoogleGroups lr-LoadRunner; Linkedin LoadRunner, LoadRunnerByTheHour; SQAForums LoadRunner, WinRunner)
We have a new project for a web app that will display banners ads on websites (as a network) and our estimate is for it to handle 20 to 40 billion impressions a month.
Our current language is in ASP...but are moving to PHP. Does PHP 5 has its limit with scaling web application? Or, should I have our team invest in picking up JSP?
Or, is it a matter of the app server and/or DB? We plan to use Oracle 10g as the database.
No offense, but I strongly suspect you're vastly overestimating how many impressions you'll serve.
That said:
PHP or other languages used in the application tier really have little to do with scalability. Since the application tier delegates it's state to the database or equivalent, it's straightforward to add as much capacity as you need behind appropriate load balancing. Choice of language does influence per server efficiency and hence costs, but that's different than scalability.
It's scaling the state/data storage that gets more complicated.
For your app, you have three basic jobs:
what ad do we show?
serving the add
logging the impression
Each of these will require thought and likely different tools.
The second, serving the add, is most simple: use a CDN. If you actually serve the volume you claim, you should be able to negotiate favorable rates.
Deciding which ad to show is going to be very specific to your network. It may be as simple as reading a few rows from a database that give ad placements for a given property for a given calendar period. Or it may be complex contextual advertising like google. Assuming it's more the former, and that the database of placements is small, then this is the simple task of scaling database reads. You can use replication trees or alternately a caching layer like memcached.
The last will ultimately be the most difficult: how to scale the writes. A common approach would be to still use databases, but to adopt a sharding scaling strategy. More exotic options might be to use a key/value store supporting counter instructions, such as Redis, or a scalable OLAP database such as Vertica.
All of the above assumes that you're able to secure data center space and network provisioning capable of serving this load, which is not trivial at the numbers you're talking.
You do realize that 40 billion per month is roughly 15,500 per second, right?
Scaling isn't going to be your problem - infrastructure period is going to be your problem. No matter what technology stack you choose, you are going to need an enormous amount of hardware - as others have said in the form of a farm or cloud.
This question (and the entire subject) is a bit subjective. You can write a dog slow program in any language, and host it on anything.
I think your best bet is to see how your current implementation works under load. Maybe just a few tweaks will make things work for you - but changing your underlying framework seems a bit much.
That being said - your infrastructure team will also have to be involved as it seems you have some serious load requirements.
Good luck!
I think that it is not matter of language, but it can be be a matter of database speed as CPU processing speed. Have you considered a web farm? In this way you can have more than one machine serving your application. There are some ways to implement this solution. You can start with two server and add more server as the app request more processing volume.
In other point, Oracle 10g is a very good database server, in my humble opinion you only need a stand alone Oracle server to commit the volume of request. Remember that a SQL server is faster as the people request more or less the same things each time and it happens in web application if you plan your database schema carefully.
You also have to check all the Ad Server application solutions and there are a very good ones, just try Google with "Open Source AD servers".
PHP will be capable of serving your needs. However, as others have said, your first limits will be your network infrastructure.
But your second limits will be writing scalable code. You will need good abstraction and isolation so that resources can easily be added at any level. Things like a fast data-object mapper, multiple data caching mechanisms, separate configuration files, and so on.