If you want to understand Complexity and Contradiction in IT Architecture and struggling to manage non-adaptive and dysfunctional systems, you don't want to miss this. one day Certified conference by John Zachman in Dubai on 23rd March 2009. For more details visit http://www.EnterpriseArchitectureLive.com For registrations, group discounts or further details please contact Caroline.email@example.com
John Zachman (Father of enterprise architecture) Given this renascent interest, who better to explain the principles behind Enterprise Architecture than the man himself, John Zachman, the originator of the " Zachman Framework for Enterprise Architecture" Join this workshop in Johannesburg 25th Mar 09 and Cape town in 27th March 09 and Mr.Zachman will explain how and why Enterprise Architecture provides measure, such an implementation is a daunting task with opportunities to fail lurking in many places. For more details visit http://www.ITArchitectureSummit.com For registrations, group discounts or further details please contact Caroline.firstname.lastname@example.org
Jurriaan Persyn is a Lead Web Developer at Netlog, a social portal site that gets 50 million unique visitors and 5+ billion page views per month. In this paper Jurriaan goes into a lot of excellent nuts and bolts details about how they used sharding to scale their system. If you are pondering sharding as a solution to your scaling problems you'll want to read this paper. As the paper is quite well organized there's no reason to write a summary, but I especially liked this part from the conclusion:
If you can do with simpler solutions (better hardware, more hardware, server tweaking and tuning, vertical partitioning, sql query optimization, ...) that require less development cost, why invest lots of effort in sharding? On the other hand, when your visitor statistics really start blowing through the roof, it is a good direction to go. After all, it worked for us.
Garry Tan, cofounder of Posterous, lists 12 lessons for scaling that apply to more than just Rails.
Update 6:: Back to the Future for Data Storage. We are in the middle of a renaissance in data storage with the application of many new ideas and techniques; there's huge potential for breaking out of thinking about data storage in just one way.
Update 5: Building Scalable Web Applications with Google App Engine by Brett Slatkin.
Update 4: Why Google App Engine is broken and what Google must do to fix it by Aral Balkan. We don't care that it can scale. We care that it does scale. And that it scales when you need it the most. Issues: 1MB limit on data structures; 1MB limit on data structures; the short-term high CPU quota; quotas in general; Admin? What's that?
Update 3: BigTable Blues. Catherine Devlin couldn't port an application to GAE because it can't do basic filtering and can't search 5,000 records without timing out: "Querying from 5000 records - too much for the mighty BigTable, apparently." Followup: not the future database. "90% of the work of this project has been trying to figure out workarounds and kludges for its bizzare limitations."
Update 2: Having doubts about AppEngine. Excellent and surprisingly civil debate on if GAE is a viable delivery platform for real applications. Concerns swirl over poor performance, lack of a roadmap, perpetual beta status, poor support, and a quota system as torture chamber model of scalability. GAE is obviously part of Google's grand plan (browser, gears, android, etc) to emasculate Microsoft, so the future looks bright, but is GAE a good choice now?
Update: Here are a few experience reports of developers using GAE. Diwaker Gupta likes how easy it is to get started on the good documentation. Doesn't like all the limits and poor performance. James here and here also likes the ease of use but finds the data model takes some getting used to and is concerned the API limits won't scale for a real site. He doesn't like how external connections are handled and wants a database where the schema is easier to manage. These posts mirror some of my own concerns. GAE is scalable for Google, but it may not be scalable for my application.
It's been a few days now since GAE (Google App Engine) was released and we had our First Look. It's high time for a retrospective. Too soon? Hey, this is Internet time baby. So how is GAE doing? I did get an invite so hopefully I'll have a more experience grounded take a little later. I don't know Python and being the more methodical type it may take me a while. To perform our retrospective we'll take a look at the three sources of information available to us: actual applications in the AppGallery, blogspew, and developer issues in the forum.
The result: a cautious thumbs up. The biggest issue so far seems to be the change in mindset needed by developers to use GAE. BigTable is not MySQL. The runtime environment is not a VM. A service based approach is not the same as using libraries. A scalable architecture is not the same as one based on optimizing speed. A different approach is needed, but as of yet Google doesn't give you all the tools you need to fully embrace the red pill vision.
I think this quote by Brandon Smith in a thread on how to best implement sessions in GAE nicely sums up the new perspective:
Consider the lack of your daddy's sessions a feature. It's what will make your app scale on Google's infrastructure.
In other words: when in Rome. But how do we know what the Romans do when the Romans do what they do?
Brett Morgan expands our cultural education in a thread on slow GAE databases performance when he talks about why MySQL thinking won't work on BigTable:
It might look almost look like a sql db when you squint, but it's
optimized for a totally different goal. If you think that each
different entity you retrieve could be retrieving a different disk
block from a different machine in the cluster, then suddenly things
start to make sense. avg() over a column in a sql server makes sense,
because the disk accesses are pulling blocks in a row from the same
disk (hopefully), or even better, all from the same ram on the one
computer. With DataStore, which is built on top of BigTable, which is
built on top of GFS, there ain't no such promise. Each entity in
DataStore is quite possibly a different file in gfs.
So if you build things such that web requests are only ever pulling a
single entity from DataStore - by always precomputing everything -
then your app will fly on all the read requests. In fact, if that
single entity gets hot - is highly utilized across the cluster - then
it will be replicated across the cluster.
Yes, this means that everything that we think we know about building
web applications is suddenly wrong. But this is actually a good thing.
Having been on the wrong side of trying to scale up web app code, I
can honestly say it is better to push the requirements of scaling into
the face of us developers so that we do the right thing from the
beginning. It's easier to solve the issues at the start, than try and
retrofit hacks at the end of the development cycle.
A truly excellent explanation of the differences between MySQL thinking and GAE thinking.
Now, if you can't use MySQL's avg feature, how can an average be calculated using BigTable? Brett advises:
Instead of calculating the results at query time, calculate them when
you are adding the records. This means that displaying the results is
just a lookup, and that the calculation costs are amortized over each
Clearly this is more work for the programmer and at first blush doesn't seem worth the effort, especially when you are used to the convenience of MySQL. That's why in the same thread Barry Hunter insightfully comments that GAE may not be for everyone:
This might be a very naive observation, but I perhaps wonder then if
GAE is the tool for you.
As I see it the App Engine is for applications that are meant to
scale, scale and really scale. Sounds like an application with a few
hundred hits daily could easily run on traditional hosting platforms.
It's a completely different mindset.
Again maybe I am missing something, but the DataStore isn't designed to
be super fast at the small scale, but rather handle large amounts of
data, and be distributed (and because its distributed it can appear
very fast at large scale).
So you break down your database access into very simple processes.
Assume your database access is VERY slow, and rethink how you do
things. (Of course the piece in the puzzle 'we' are missing is
MapReduce! - the 'processing' part of the BigTable mindset)
Before developers can take full advantage of GAE these types of lessons need to be extracted and popularized with the same ferocity the multi-tier RDBMS framework has been marketed. It will be a long difficult transition.
Interestingly, many lessons from AWS are not transferable to GAE. AWS has a VM model whereas GAE has an application centric model. They are inverses of each other.
In AWS you have a bag of lowish level components out of which you architect your application. You can write all the fine low level implementations bits you desire. A service layer is then put in front of everything to hide the components. In GAE you have a high level application component and you build out your application using services. You can't build any low level components in GAE. In AWS the goal is to drive load to the CPU because CPU and bandwidth are plentiful. In GAE you get very limitted CPU, certainly none to burn on useless activities like summing up an average over a whole slice of data returned from SimpleDB. And in GAE the amount of data returnable from the database is small so your architecture needs to be very smart about how data is stored and accessed.
Very different approaches that lead to very different applications.
The number of applications has exploded. I am always amazed at how enthusiastic and productive people can be when they are actually interested in what they are doing. It happens so rarely. True, most applications aren't even up to Facebook standards yet, but it's early days. What's impressive is how fast they were created and deployed. That speaks volumes about the efficacy of the application centric development model.Will it be as effective delivering "real" apps? That's a question I'm not sure about.
So far application performance is acceptable. Certainly nothing spectacular. What can you do about it? Nada.
I like the sketch application because people immediately and quite predictably drew lewd depictions of various body parts. I also like this early incarnation of a forum app. A forum is one of the ideas I thought might work well on AppEngine because the scalable storage problem is solved. I do wonder how the performance will be with a fine tuned caching layer? Vorby is a movie quote site showing a more realistic level of complexity. It has tabs, long lists of text, some graphical elements, some more complex screens, and ratings. It shows you can make applications you wouldn't mind people using.
An option I'd like to see in the App Gallery is a view source link. Developers could indicate when adding an application if others can view their application source. Then when browsing the gallery we could all learn by looking at real working code. This is how html spread so quickly. Anyone could view the source for any page, copy paste, and you're on your way! With an application centric model the view source viral spread approach would also work.
As expected there's lots of blog activity on GAE:
A lot has been made of the risk of lock-in. I don't really agree with this as everything is based around services, which you can port to another infrastructure. What's more the problem is developers will be acquiring a sort of learned helplessness. It's not that developers can't port to another environment, they simply won't know how to anymore because they will have never had to do it themselves. Their system design and infrastructure muscles will have atrophied so much from disuse that they'll no longer be able to walk without the aide of their Google crutches. More in another post.
Developer ForumThe best way to figure out how a system is doing is to read the developer support forum. What problems and successes are real developers experiencing trying to get real work done? The forum is a hoppin'. As of this writing over 1300 developers have registered and nearly 400 topics are active. What are developers talking about?
Many "how do I" questions come up because of the requirement for service level interfaces. For example, something as simple as a hostname to IP mapping can't be done because you don't have socket level access. Someone, somewhere must make a service out of it. Make an external service is a common response to problems. You must make a service external to the GAE environment to get things to work which means you have to develop in multiple environments. This sort of sucks. To get cron functionality do I really need to create an external service outside of GAE?
The outcome of all this is probably an accelerated servicifaction of everything. What were once simple library calls must now be exposed with service level interfaces. It's not that I think HTTP is too heavy, but as development model it is extremely painful. You are constantly hitting road blocks instead of getting stuff done.
Share the experience of hosting highly scalable/reliable GIS based application which involves Map Server, Spatially enabled database, j2ee, Routing Applications etc.
Hi, We are running a backup solution that uploads every night the files our clients worked on during the day (Cabonite-like). We have currently about 10GB of data per night, via http PUT requests (1 per file), and the files are written as-is on a NAS. Our architecture is basically compound of a load balancer (hardware, sticky sessions), 5 servers (Tomcat under RHEL4/5, ) and a NAS (nfs 3). Since our number of clients is rising, (as is our system load) how would you recommend we could scale our infrastructure? hardware and software? Should we go towards NAS sharding, more servers, NIO on tomcat...? Thanks for your inputs!
Google AppEngine Numbers
This group of numbers is from Brett Slatkin in Building Scalable Web Apps with Google App Engine.
Writes are expensive!
* The size and shape of your data
* Doing work in batches (batch puts and gets)
Reads are cheap!
* For a 1MB entity, that's 4000 fetches/sec
Numbers MiscellaneousThis group of numbers is from a presentation Jeff Dean gave at a Engineering All-Hands Meeting at Google.
The TechniquesKeep in mind these are from a Google AppEngine perspective, but the ideas are generally applicable.
Sharded CountersWe always seem to want to keep count of things. But BigTable doesn't keep a count of entities because it's a key-value store. It's very good at getting data by keys, it's not interested in how many you have. So the job of keeping counts is shifted to you.
The naive counter implementation is to lock-read-increment-write. This is fine if there a low number of writes. But if there are frequent updates there's high contention. Given the the number of writes that can be made per second is so limited, a high write load serializes and slows down the whole process.
The solution is to shard counters. This means:
This approach seems counter-intuitive because we are used to a counter being a single incrementable variable. Reads are cheap so we replace having a single easily read counter with having to make multiple reads to recover the actual count. Frequently updated shared variables are expensive so we shard and parallelize those writes.
With a centralized database letting the database be the source of sequence numbers is doable. But to scale writes you need to partition and once you partition it becomes difficult to keep any shared state like counters. You might argue that so common a feature should be provided by GAE and I would agree 100 percent, but it's the ideas that count (pun intended).
Paging Through Comments
How can comments be stored such that they can be paged through
in roughly the order they were entered?
Under a high write load situation this is a surprisingly hard question to answer. Obviously what you want is just a counter. As a comment is made you get a sequence number and that's the order comments are displayed. But as we saw in the last section shared state like a single counter won't scale in high write environments.
A sharded counter won't work in this situation either because summing the shared counters isn't transactional. There's no way to guarantee each comment will get back the sequence number it allocated so we could have duplicates.
Searches in BigTable return data in alphabetical order. So what is needed for a key is something unique and alphabetical so when searching through comments you can go forward and backward using only keys.
A lot of paging algorithms use counts. Give me records 1-20, 21-30, etc. SQL makes this easy, but it doesn't work for BigTable. BigTable knows how to get things by keys so you must make keys that return data in the proper order.
In the grand old tradition of making unique keys we just keep appending stuff until it becomes unique. The suggested key for GAE is: time stamp + user ID + user comment ID.
Ordering by date is obvious. The good thing is getting a time stamp is a local decision, it doesn't rely on writes and is scalable. The problem is timestamps are not unique, especially with a lot of users.
So we can add the user name to the key to distinguish it from all other comments made at the same time. We already have the user name so this too is a cheap call.
Theoretically even time stamps for a single user aren't sufficient. What we need then is a sequence number for each user's comments.
And this is where the GAE solution turns into something totally unexpected. Our goal is to remove write contention so we want to parallelize writes. And we have a lot available storage so we don't have to worry about that.
With these forces in mind, the idea is to create a counter per user. When a user adds a comment it's added to a user's comment list and a sequence number is allocated. Comments are added in a transactional context on a per user basis using Entity Groups. So each comment add is guaranteed to be unique because updates in an Entity Group are serialized.
The resulting key is guaranteed unique and sorts properly in alphabetical order. When paging a query is made across entity groups using the ID index. The results will be in the correct order. Paging is a matter of getting the previous and next keys in the query for the current page. These keys can then be used to move through index.
I certainly would have never thought of this approach. The idea of keeping per user comment indexes is out there. But it cleverly follows the rules of scaling in a distributed system. Writes and reads are done in parallel and that's the goal. Write contention is removed.
Moshe Kaplan of RockeTier shows the life cycle of an affiliate marketing system that starts off as a cub handling one million events per day and ends up a lion handling 200 million to even one billion events per day. The resulting system uses ten commodity servers at a cost of $35,000. Mr. Kaplan's paper is especially interesting because it documents a system architecture evolution we may see a lot more of in the future: database centric --> cache centric --> memory grid. As scaling and performance requirements for complicated operations increase, leaving the entire system in memory starts to make a great deal of sense. Why use cache at all? Why shouldn't your system be all in memory from the start?
General Approach to Evolving the System to Scale
One Million Event Per Day System
2.5 Million Event Per Day System
20 Million Event Per Day System
200 Million Event Per Day System
Joe Stump, Lead Architect at Digg, gave this presentation at the Web 2.0 Expo. I couldn't find the actual presentation, but fortunately Kris Jordan took some great notes. That's how key moments in history are accidentally captured forever. Joe was also kind enough to respond to my email questions with a phone call. In this first part of the post Joe shares some timeless wisdom that you may or may not have read before. I of course take some pains to extract all the wit from the original presentation in favor of simple rules. What really struck me however was how Joe thought MemcacheDB Will be the biggest new kid on the block in scaling. MemcacheDB has been around for a little while and I've never thought of it in that way. Well learn why Joe is so excited by MemcacheDB at the end of the post.