After you make an investment, your first step should be to decide how you’re going to protect it. With the cloud, the best protection you can ensure is making good use of your cloud management tools.
So when it comes time to map your cloud strategy, you should have a plan and arsenal of tools at hand to measure and monitor the following: performance, application management, security, storage, disaster recovery, and more.
The good news is that there are now cloud management tools that address almost every performance task you can think of. The bad news is that most of them aren’t widely known.
That said, I was excited when I was asked to present during the breakout session, “Successful Cloud Management Strategies,” at this year’s All About the Cloud, our industry’s leading ISV conference.
I’ll be speaking on Wednesday, May 9, and I’ll be in good company. Moderator John Rowell, Co-Founder and Chief Technology Officer, OpSource, Inc. will lead the discussion among my fellow panelists: Mat Ellis, CEO and Co-Founder of Cloudability; Jamal Mazhar, CEO and Founder of Kaavo; and Ray Solnik, President of Appnomic Systems, Inc.
There’s still time to register for the event, so if you haven’t already, be sure to sign up.
How did you first map out your cloud strategy? Were management tools the top priority? Let us know below in a comment.
If you aren’t able to attend the conference, follow us on Twitter @apicasystems. I’ll be live tweeting during the entire show and you can catch some highlights from my session there as well. You can also keep up with the conference by following @SIIASoftware and #AATC, the official hashtag of All About the Cloud.
Need to access your customer information in the cloud at all times? There’s an app for that. Need to see live production numbers? There’s an app for that. Want to start a business in the cloud? Now there’s an entire marketplace for that. And Apica is in it.
This is big news in the cloud space, so it’s likely you heard about it by now. But in case you missed it: Last week, Amazon Web Services (AWS) launched AWS Marketplace, an online store that makes it easy for people to find, compare, and begin using the software and technical services they need to build products or run their businesses on the AWS cloud. It’s a one-stop shop that features myriad products in software infrastructure, business software, and developer tools.
It’s that last category we’re most excited about, because that’s where Apica LoadTest, our newest addition to AWS is featured. LoadTest follows Apica WebOverload and ApicaWatch to our list of solutions on AWS.
Moving your business to the cloud is a smart move for companies of any size, but only if you’ve properly tested to ensure you’ll be able to perform at the same capacity or higher.
It’s also a smart move for developers, who are increasingly targeting the cloud as their platform of choice. Unfortunately, the cloud is a new vista for most programmers, and many programming teams lack the requisite skills and cloud-savvy tools to do the capacity planning and load testing in the cloud. It’s an entirely new set of variables. And that’s where Apica can help.
If you’re a cloud developer, if your company’s moving to the cloud, or if you’re having performance problems in the cloud, hit the new AWS Marketplace to learn more about Apica LoadTest. And be sure to stay tuned to this blog where we’ll be posting a case study in the next few days that shows a real-world example of load testing in the cloud.
Good performance is a result of a long term commitment to details.
Websites today are struggling to maintain high quality, good performance and still be able to deploy new versions and updated functionality with the speed the market is demand. The problem when you have a website development project is that functional validations often introduce separate stage environments which typically don’t include an identical setup. Even if most of the server setup is similar, infrastructure components like a load balancer/Firewall and backend database will differ. So if possible, always load test in the production environment in least the final stage of the web project at the very least.
By relying solely on functional testing you are treading a very dangerous path if performance and stability is the goal. Knowing your applications performance and different levels of performance zones is a must in today’s fast changing web and cloud environment. In order to optimize your capacity and prepare for different types of load patterns, you need to schedule accordingly:
Load patterns:
- Low – during night
- High – during busy hours
- Peak – during events or incidents
A load test will give you a clear picture of how your system performs during the different load patterns – from Low to Peak - before you actually will be challenged during real production
One way of describing and communicating performance results is to define load levels as zones, then correlate them to response time, system utilization application setup and performance goals.
Load pattern Zones:
- Basic Zone: Low load - night time, low to none, start point for the load test
- Comfort Zone: Medium load - stable response time, low standard deviation, low – zero level of application errors - baseline
- Stress Zone: High Load - when the maximum throughput of the system is reached (performance knee) - response time starts to escalate when the performance knee is passed
- Breakdown Zone: Peak load 2-10 times more than the max number of users/load during max throughput
In the graphs above you see the different zones marked. Please note that the Breakdown zone has two typical setups.
#1: Throughput has reached the maximum but is still stable – straight line, response times goes up drastically, but system still responding.
#2: Throughput is reduced due to a system problem, so your system is actually collapsing due to overload. You might typically have Database / Web Server that is not responding and therefore the response time increases exponentially and the throughput goes drastically down.
In case #1 the system response times will be reduced as soon as the load goes down. If you have case #2 setup, you probably need to reboot the full system, even if the load goes down
This should be able to give you a clear view of the capacity of the system (response time / number of users) during the different load patterns. The result gives you guidance if you need to extend the comfort zone, introduce scaling or have performance problems right out of the Get-Go during low load. Repeated system updates need to be plotted in this diagram so that you can track the impact of component and application changes in a true DevOp style of release management.
Do you know your comfort zone? When do you tread into your stress zone?
Like in real life stress planning ahead of time (always perform load test before release) is a good way to avoid stressful situations.
Apica has made it our mission to help companies optimize their web and cloud performance, but often the clients aren’t even aware of the problem until it’s too late. To illustrate this we used the holidays to perform a monitoring cycle in order to determine the response time Index for a few high performance brands within the automotive industry.
The brands we chose where Audi, BMW, Ferrari, Jaguar, Lamborghini, Mercedes-Benz and Porsche, all well known for their high quality cars and prestige linked to their brands. We performed the monitoring using one of our services called ApicaWatch™. ApicaWatch™ mimic real user traffic to the first page of each company site in order to determine the user experience when entering and watching the site. This is what we in the industry call a "check". We performed the checks during a 7 day period over the holidays and all of the sites monitored where “.com”.
The checks provides us with two kinds of important information, the first is how the page is organized by showing all the elements it consists of, and in which order they are rendered (made visible) on the site. The other thing is how long load time every single element have detailed down to URL-level. All this is shown in graphs in order to simplify to identification of the problem areas. Today’s standard is that the site should take no longer than 3000 ms (3 sec) to be completely rendered (visible).
Results
The site with the best average where Jaguar (www.jaguar.com) that hade and average load time of 2.480 seconds, in second place came Porsche (http://www.porsche.com/) with 4.085 seconds. The bronze medal went to BMW (http://www.bmw.com/) that had the highest average, 4.422 seconds during the test period. We can see that out of these three only one where within the 3 second web standard and with almost half a second to spare.
Then we had Audi (http://www.audi.com/) with an average of 6.208 seconds, following them came Lamborghini (http://www.lamborghini.com) with 6.576 seconds, that’s almost 3 second slower than for the Murciélago to reach 100 km/h. Next on the list where their Italian brother Ferrari (http://www.ferrari.com) with 8.959 seconds. Mercedes-Benz (http://www3.mercedes-benz.com) had the highest average, 9.126 seconds, during the test period.
This gives us a total average of 5.872 across the different sites we tested. Thus the average load time where almost 3 seconds above todays “accepted” average for websites. This may not sound like much, but reports from amazon have shown that for every 0.1 second increase in load time you’ll drop 1% in sales. This translates into an average 30% drop of sales for the brands we tested during the test period.
So take it from us, continue focus on speed and performance, but not only on the road. You might just be overtaken by a competitor right at the finish line.
Reuters estimated this holiday season brought in over US $35B in online shopping sales worldwide, the highest amount ever recorded. Significant spikes in website traffic can bring serious consequences to unprepared e-Retailers. Slow page load times, performance issues and even downed sites, can effectively put e-Retailers out of business.
“Research shows 40 percent of online shoppers will abandon a site after waiting three seconds for a page to load. Frustration with online shopping due to performance issues was recounted first hand in many of the creative entries that were submitted for our Twitter contest. From the comical to the unusual, submissions spanned the spectrum, making selecting our winners a tough decision. A huge thank you to all who participated,” said Sven Hammar, CEO of Apica.
Winners were selected based on creativity, humor and also relevance. And the winners are…drum roll please…
Grand Prize ($1,000 Amazon Gift Card)
Kaizeneer Kaizeneer:
Buying bling like Lil John/Check’n status-BAD log on/Order went 2 Kuala Lumpur/Mad like gunshots at Blk Friday store @apicasystems #apicarap
First Prize ($100 Amazon Gift Card)
jadenruby jadenruby: @apicasystems #apicarap Online Ease-I'll take it Please/No Bump and Grind-of the annoying kind/Next year all shopping thru computer queue!
Second Prize ($50 Amazon Gift Card)
JCris_08 Julius Cristobal:
Hate da long line so I jus hop online/ Nothin can beat shoppin from my computer seat #apicarap @apicasystems
Failure to deliver performance in an application will have a significant impact on the business. Apica’s solutions test, monitor and optimize web, cloud and mobile applications and give businesses a true understanding of how their applications will and are performing. Using Apica’s technology, users can reduce the risk of project failure.
Keep on rhyming!
We have so far seen a good number of tweets from different participants since the start of the competition at Black Friday. Marking the first point of the competition, Cyber Monday, we can se that the holiday shopping this year is as intense as ever. Presenting the Top 10 Submissions in the contest, so far...
Whats happening?
To help kick off the 2011 e-Retail holiday shopping season, we announced a fun Twitter contest where consumers can share their best or worst e-Retail story as a rap or rhyme. The most creative submissions will be chosen for a chance to win a US$1,000 gift card to Amazon and other cash prizes. To be a part of this contest, entries must be submitted in the form of a 140 character rap or rhyme via Twitter to @apicasystems and include the hash tag #apicarap.
Top 10 Submissions to Rap/Rhyme Twitter Contest
jadenruby jadenruby
@apicasystems #apicarap Shopping cart, shopping cart, shopping cart is full, Oh what fun, it is to shop, With a credit card on-line! :)
Monnie311 Monica Young
@apicasystems I get so angry when the pages won't load, just want to check out but feel like I'm going to explode! #apicarap
Kaizeneer Kaizeneer
Buying bling like Lil John/Check’n status-BAD log on/Order went 2 Kuala Lumpur/Mad like gunshots at Blk Friday store @apicasystems #apicarap
ChesterAlArthur Chester
@apicasystems #apicarap Money was donated to charity with my online transaction, my praise to the e-retailer for such a heartfelt action!
roschler Robert Oschler
@apicasystems - Better pray, pepper spray, doesn't get in your eye. Or lose your place, in the race, to get an Xbox or die! #apicarap
katmagick Melanie
@apicasystems #apicarap Shoppin online really frees up time tap tap tap All the shoppins done went out to have some fun Easy as tap tap tap
PERiWiNKLEluv hisini
@apicasystems #apicarap no time to browse, big darn crowd, every1 was loud, can't say the manager was proud should of been better my bounds.
Kaizeneer Kaizeneer
HiJack was nimble/HiJack was quik/hacked firewall so slick/sent my e-order/over da border/sayn what a bad boy am I! @apicasystems #apicarap
mycripley Mychogg Ripley
@apicasystems after 4 seconds i had 2 refresh with 1 item left in my cart im frettin I might not get it w no credit #apicarap
jadenruby jadenruby
#apicarap @apicasystems Shoppin mall shoppin mall/parkin brawls, aisles crawl/kids R wired, moms R tired, workers R fired/4 Xmas cheer! NOT!
A panel of judges will select the winners based on creativity, humor and relevance. Winners will be notified on January 4, 2012. For official rules and prize information, please visit: http://www3.apicasystem.com/twittercontest11/
Good luck and happy rappin!
Consumers continue to embrace the web for holiday shopping this season. Statistics show traffic and sales for both “Black Friday” and “Cyber Monday” increased significantly over last year -- many e-Retailers saw an increase of 20 percent or more. In total, more than $40 billion is expected to be spent in e-Commerce sales this holiday season.
With this flood of additional web traffic comes the potential for significant performance issues and potentially lost revenues if retailers aren’t ready. Research by Forrester has shown that 40 percent of online shoppers will abandon a site after waiting three seconds for a page to load. Despite months of preparation, performance satisfaction for the top 50 U.S. retailers was down more than 12 percent on Cyber Monday from non-holiday baseline levels. We offer unmatched website optimization expertise to our clients. Our unique technology keeps e-Retailers in business and making money.
To help kick off the 2011 e-Retail holiday shopping season, we are pleased to announce a fun Twitter contest where consumers can share their best or worst e-Retail story as a rap or rhyme. The most creative submissions will be chosen for a chance to win a US$1,000 gift card to Amazon and other cash prizes. To be a part of this contest, entries must be submitted in the form of a 140 character rap or rhyme via Twitter to @apicasystems and include the hash tag #apicarap.
Here are a few examples to get your creativity ‘flowin’:
@apicasystems #apicarap Got the holiday blues when I couldn’t buy some shoes…web page wouldn’t load. eRetailers, get a clue!
@apicasystems #apicarap Holiday shopping, oh what fun! Click… click… click… Thanks to the Internet, all my shopping is done!
A panel of judges will select the winners based on creativity, humor and relevance. Winners will be notified on January 4, 2012. For official rules and prize information, please visit: http://www3.apicasystem.com/twittercontest11/
Good luck and happy rappin!
What do Virtualization and Cloud executives think about 2012? Find out in this VMblog.com series exclusive.
2012 - The Year Performance Pays Off
Contributed Article by Sven Hammar, CEO, Apica
The explosion of web applications and websites moving to the cloud has, in some respect, caused a fair amount of problems with regards to capacity. Given this, not all services have fully kept pace with demand in terms of performance and reliability. But, in 2012, the market will take performance and stability to a new level.
Prediction #1
The new web standard, HTML 5, will become a huge splash and introduce a new and much better way to work with web applications. Most new applications will take advantage of HTML 5 features, local storage, computing capabilities and also WebSocket support.
Prediction #2
Adobe Flash will take a big step back. With HTML 5 moving forward, Flash will, in many respects, become obsolete. Apple has not been shy about their hard resistance to this technology and Microsoft has announced they will not support Flash in Metro. So, if you are developing a new application, Flash isn’t the best way to go in the new year. It might be too soon to claim the death of Flash, but 2012 should be a defined year in the trend that development will move away from Flash in huge volumes.
Prediction #3
It’s interesting that there are no cloud providers out there that can give you a true indication of how good the performance of your applications will actually be. There are no guaranties for load times or transfer speeds. So, let’s be bold and say that one of the big 5 cloud providers will step forward in the new year and provide both guarantees and also independent validation of performance as new offers to its clients. This trend will then escalate to application and services. In short, just “uptime” will not be enough.
Prediction #4
Anyone serious about achieving exceptional performance will start to use WebSocket for transaction and data-intense applications. AJAX calls are just not good enough to transfer data in client server speed. Updating real-time graphs and dealing with large volumes of transaction will greatly benefit from WebSocket instead of working the state less challenge to send data back and forth via HTTP is great for content, but will come up short for frequent data updates.
Prediction #5
With social media setting the stage for more and more of the Internet’s overall web traffic, a lot of the typical social manifestation will be moved to .Net in 2012. Product demonstrations and such from companies will become a much more frequent event on the web.
And consider the role Facebook played during the rising in the Arabic world. DDOS attacks used to be the main problem, now you can be shut down just by a Twitter feed to your website in a snap – if the activating person has 10 million followers on Twitter or Facebook. You can easily get 20-30 Gb/s load towards your site.
Load testing your applications and monitoring your site on-going to tune scaling and other queuing systems to withstand extreme traffic will become a must for any successful website in 2012.
##
About the Author
Sven Hammar, founder and CEO of Apica, is an expert in application testing and performance monitoring. He has both business and technical experience with start-ups and has established companies in the United States and in Europe. Sven has served in many positions throughout his career, including CEO, VP of Sales and VP of Marketing. Sven's expertise includes presenting and receiving major venture capital funding, establishing distribution and partnerships with leading industry players and leading mergers and acquisition transactions.
Published Thursday, December 01, 2011 6:00 AM by
David Marshall Filed under:
VMBlog Info,
Prediction 2012
Cloud Scaling Based on User Response Time
Today, scaling functionality is typically very one-dimensional and based almost exclusively on server CPU, not on the performance of applications, such as response times, transaction time for purchases, login etc. When a heavy load hits a site site owners need to be certain that it responds and scales with optimal performance in mind. Solely measuring CPU is not necessarily the best way to identify the optimal response. Ideal would be if you could sync the performance monitoring with the internal cloud scaling, either through an API or directly through the monitoring application.
This week, Apica unveiled the industry’s first cloud performance scaling solution with our new service, Apica AzureCheck. With it, you can monitor inside vital sign, as well as outside performance through the Apica Network. Designed specifically for Microsoft Azure applications, Apica AzureCheck enables Azure scaling, or any cloud scaling function, to be integrated with the Apica Application Performance Monitoring system. The new technology is being offered at no cost for a 30 day trail.
So, let us take full advantage of clouds possibilities. Do yourself a favor and make sure to set the scaling setup to identify application performance, not just the server CPU.
You have successfully relocated your application to the cloud. You know the cloud is elastic and you can increase what you need when you need it. This is an insurance in itself but what you should ask yourself is; when should you actually increase whatever it is that drives your application – or an even better question is – how do you get notified when you need to increase?
When that time comes - do you double the resources, triple or make it tenfold?
If you don’t have a plan you will be in for a surprise when there is a sudden rush of traffic to your site - resource usage goes up, you are not prepared for it and the site goes down.
Depending on which cloud you have your application in - you have different integrated options (if any) to monitor the health and manage your applications elasticity. You are often limited to what is already there and it is difficult to correlate resource usage with actual
outside response times.
Apica are currently working on new products that will address these issues and actually provide the way of being kept “in the loop” of your applications health, along with outside monitoring. When resource usage is intensive, you could have your application automatically scale up an extra web depending on what rules you have defined. Likewise – when the application is idle, why not let it scale down and cruise on just one web?