Monitoring the Performance to SaaS & Cloud Services
What is a Path?
A Path is the network connection from a source to a target, specifically measuring the maximum end to end performance achievable. For example a Path from our office here in Boston to the Salesforce.com instance we use (NA2) located outside of Chicago consists of 12 layer 3 network hops.
We have a 100Mbps symmetric internet connection to the internet, so that means we have 100Mbps to Salesforce, right? Unlikely.
We are getting between 80Mbps and 90Mbps Total Capacity end-to-end this morning, which is reasonable. But why aren’t we getting all 100Mbps we are paying for? Great question, which leads us to our next topic:
AppNeta is giving me different performance measures than my SNMP tool. Why?
Because they are measuring different things. As you remember, there are 12 layer 3 network hops from us to Salesforce, and an SNMP tool is measuring the throughput at hop 1, our network firewall. This is a valid measure of the traffic volume and rate at the egress point of your network, but remember that PathView is measuring the end-to-end performance all the way to the target, including hops 2 through 12 which are shared with general internet traffic and some portion of the 82,000 other Salesforce customers.
So the SNMP measure is accurate but is not close to telling the complete picture, especially when what you care about is the performance to SaaS provider, Cloud service or your remote offices.
How do we measure performance over 3rd Party Networks?
This is where it gets interesting (for network geeks anyway). The AppNeta service uses active performance monitoring, so we send and receive a small amount of network traffic, and by very precisely analyzing that traffic we can determine all of these key metrics.
Where this becomes most valuable for Cloud and Hosted Services, you do not need to own or install anything on the target you are measuring the performance of. This means that you can measure the performance of Salesforce.com, hosted Microsoft Lync, and your own Apps running in AWS without any changes to your environment or the hosting providers environment.
If the Performance Issue is in “the Internet” or my Cloud / Saas Provider, what can I do about it?
Let them know, chances are that if the issue is still occurring the service provider does not know about the issue and they want their service to work as well as possible for all users. AppNeta has helped thousands of customers diagnose these issues with many, many providers and while the Network or Cloud / SaaS Provider will deny up and down that the issue is occurring, the problem will magically get solved without admitting fault. Funny how that happens.
What Application SLA’s should I be measuring against with my provider?
The performance requirements are determined by your application, and while you should dig into the details of your specific applications, here are some good guidelines for common Cloud / Saas apps:
|Application||Available Capacity||End-to End QoS||Maximum Jitter / RTT||Maximum Packet Loss|
|Standard Web Applications||125Kbps*||Optional||120ms RTT||1.50%|
|Rich Internet Applications||400kbps*||Optional||80ms RTT||0.50%|
|Hosted VoIP||80kbps*||Optional||40ms Jitter||1%|
|*Available end-to-end Bandwidth per concurrent Session|