How Zero Nines Ensures Uptime
By Audrey Rasmussen - Enterprise Management Associates
April 23, 2003
Many management products kick in after a problem occurs, requiring some lag time before users can resume their work. In some cases, it could take minutes, but sometimes data recovery could take days. Can your company have an application that's down for days? Obviously, it depends on the application, but for mission-critical applications, there can be serious impacts on the business.
I recently became acquainted with a product from Zero Nines that can help IT to avoid downtime altogether, rather than reacting once it happens. It's particularly useful for applications that can't go down or can tolerate only a minimal amount of downtime.
Here's how it works. Zero Nines uses a specially designed switch that takes user application requests and multicasts them to several application servers, creating multiple application streams running in parallel. Zero Nines bases its recommendations on statistical analysis that shows that multicasting to three servers enables ultrahigh availability, and multicasting to two servers provides five nines (99.999%) of availability.
Each of the receiving application servers thinks it is the primary server processing the request. If one of the servers goes down in the middle of processing the transaction, one of the other servers takes over immediately, with no lag time perceived by the user. All of this happens without the user knowing that the application server has gone down, because the application is still running just like it's supposed to. Zero Nines handles the processing sequences to ensure that the transactions are handled in the correct order.
In addition, once the server is online again, it automatically "catches up" with the processing that it missed during the outage, so that it's up to date. Failover is not necessary because the application is running in three places, in real time. If a server goes down, there are two other servers that can still service the user.
An application can always be available, assuming that all three servers running it are not disabled at once. If your sites are carefully planned, statistically, the likelihood that all three sites are rendered inoperable simultaneously is extremely small.
The application servers can reside within the corporate network in disparate locations, or companies can take advantage of Zero Nines' partnership with AT∓T, which offers the capability to leverage external telecommunications infrastructure for added resiliency.
Interestingly, Zero Nines can handle most platforms for the target systems, including older operating systems like Windows 98. It can handle heterogeneous target systems, heterogeneous both in hardware and in operating system. So companies can leverage the equipment that they already own, old and new, rather than having to buy identical systems, as some high-availability products require.
What makes this product different from others is that it aims to provide business continuity at the application level. Others focus on redundancy or recovery at the hardware level or at the data level. And some products focus on what can be done once a problem occurs. However, Zero Nines' focus is to keep an application running, where the infrastructure is resilient enough to continue delivering the application service, despite specific problems in the infrastructure.
During a demonstration of the technology, and unlike most demonstrations, the Zero Nines people invited me to pull out any cables and try to break it. To borrow a phrase from old Timex commercials, "It takes a licking but keeps on ticking." Zero Nines is an interesting product for applications that need high availability.
Copyright Network World, Inc., 2003