Most amateur database programmers will answer you that any kind of application “basically only needs one big table”. Well, Twitter has proved them wrong. Just throwing more resources at a problem isn’t necessarily the best idea if your design is already flawed. That is also true for cloud-based testing, or even cloud-based testing of your cloud-based applications.
ran into similar problems when testing my deployment of 100 Asterisk servers last night and loading them with about 100,000 calls per hour. Lori Macvittie has a great summary on ports, load balancing, browsers, and polocies; here an excerpt (and read the full article!):
|
You can follow the comments for this article with the RSS 2.0 feed.
You make some good points here, Sachin. I guess the lesson learned is that a lot of people who have little previous experience with large deployments also have little experience with the right testing environments. In effect they forgo basic testing principles and procedures that are “old” and well known and instead think that more resources can solve their problem instead of the right design – which is exactly my point.
On the other hand, if she would have 100 local PCs in her network, a DHCP router would’ve assigned 100 IP addresses, which in return would’ve avoided triggering flooding policies. Bad virtual machine setup, on the other hand, could lead to a single IP address for 100s of machines, resulting in her aforementioned triggering of flooding policies.
And yes, I’m looking forward to the order entry due to “please rewrite my applications” requests :)
http://www.kemptechnologies.com/?utm_source=blog&utm_medium=pv&utm_content=zs&utm_campaign=home
For the small-medium businesses who are struggling with server issues, think about investing in a load balancer.
why *exactly*? When i’m struggling with server issues, I buy a load balancer – in general? Could you elaborate a little bit, please? or are you simply advertising that you have a load balancer at kemptechnologies? ;)
Please let me know, I’m genuinely interested where you see that connection, for what business need, and for what technology problem…
Content © Playout Intelligence
Proudly powered by WordPress
Theme designed by Artisan Themes
Hi Thorsten,
The issue Lori brings up is a network provisioning (lack of IP addresses) rather than a cloud-specific one. It would be the same if she had 100s of (non-cloud) physical machines; unless she supplied non-NATed PCs to test, there would still be the confused web-apps. So I would contend that the lack of resources (IP addresses) caused the issues she described, rather than cloud computing. In my opinion the design assumptions of the applications being deployed over cloud infrastructures are not in line with the new reality of execution environments (e.g. assumptions of non-NATed machines).
Theres a whole bunch of applications that need to be re-written due to (1) Cloud computing execution platforms (2) Multi-core processing environments. Thats good news for developer firms!