Skip to main content

Why Do Internet Services Fail, and What Can Be Done About It?

The number and popularity of large-scale Internet services such as Google, MSN, and Yahoo! have grown significantly in recent years. Such services are poised to increase further in importance as they become the repository for data in ubiquitous computing systems and the platform upon which new global-scale services and applications are built. These services' large scale and need for 24x7 operation have led their designers to incorporate a number of techniques for achieving high availability. Nonetheless, failures still occur.  Although the architects and operators of these services might see such problems as failures on their part, these system failures provide important lessons for the systems community about why large-scale systems fail, and what techniques could prevent failures. In an attempt to answer the question "Why do Internet services fail, and what can be done about it?" we have studied over a hundred post-mortem reports of user-visible failures from three large-scale Internet services. In this paper we
  • identify which service components are most failure-prone and have the highest Time to Repair (TTR), so that service operators and researchers can know what areas most need improvement;
  • discuss in detail several instructive failure case studies;
  • examine the applicability of a number of failure mitigation techniques to the actual failures we studied; and
  • highlight the need for improved operator tools and systems, collection of industry-wide failure data, and creation of service-level benchmarks.
Read the paper here.

Ron - the problem management guy


Popular posts from this blog

Why Madge Networks, the token-ring company, went titsup

There I was shooting the breeze with an old mate. The conversation turned to why Madge Networks which I wrote about here went titsup. My analysis is that Madge Networks had a solution and decided to go out and find a problem. They deferred to more incorrect strategic technology choices. The truth of the matter is that when something goes titsup, its not because of one reason only, but a myriad of them all contributing to the negative consequence. There are the immediate or visual ones, which are underpinned by intermediate ones and finally after digging right down, there are the root causes. There is never a singular root cause for anything but I'll present my opinion and encourage everyone else to chip in. All of them together are more likely the reason the company went titsup. As far as technology brainfarts go there is no better example than Kodak . They invented the digital camera that killed them. However, they were so focused on milking people in their leg

Flawed "ITIL aligned"​ Incident Management

Many "ITIL aligned" service desk tools have flawed incident management. The reason is that incidents are logged with a time association and some related fields to type in some gobbledygook. The expanded incident life cycle is not enforced and as a result trending and problem management is not possible. Here is a fictitious log of an incident at PFS, a financial services company, which uses CGTSD, an “ITIL-aligned” service desk tool. Here is the log of an incident record from this system: Monday, 12 August: 09:03am (Bob, the service desk guy): Alice (customer in retail banking) phoned in. Logged an issue. Unable to assist over the phone (there goes our FCR), will escalate to second line. 09:04am (Bob, the service desk guy): Escalate the incident to Charles in second line support. 09:05am (Charles, technical support): Open incident. 09:05am (Charles, technical support): Delayed incident by 1 day. Tuesday, 13 August: 10:11am (Charles, technical support): Phoned Alice.

Updated: Articles by Ron Bartels published on iot for all

  These are articles that I published during the course of the past year on one of the popular international Internet of Things publishing sites, iot for all .  These are articles that I published during the course of the past year on one of the popular international Internet of Things publishing sites, iot for all . Improving Data Center Reliability With IoT Reliability and availability are essential to data centers. IoT can enable better issue tracking and data collection, leading to greater stability. Doing the Work Right in Data Centers With Checklists Data centers are complex. Modern economies rely upon their continuous operation. IoT solutions paired with this data center checklist can help! IoT Optimi