Four Reasons Why Symantec Cluster File System Continues to Get the Nod over NAS with NFS v4
A few months ago I wrote a blog entry that took a look at how enhancements in the NFS v4 standard have opened the door for NAS file service appliances to be used with more highly available (HA) transaction-oriented applications such as the TIBCO Enterprise Service Bus (ESB). But before organizations adopt a NAS solution that uses NFS v4 for real-time failover, they should be aware of features offered by the Veritas Cluster File System (CFS) solution that NAS with NFS v4 alone does not yet provide.
To summarize a few of the points that I made in that previous blog entry: One of the biggest improvements in NFS v4 relates to file management, a notable enhancement from prior versions as it applies to clustered application servers configured to access the same file via NFS.
Previously using pre-NFS v4, if the application failed and was not able to release the lock, the application on the standby server could not start processing as the file would still be locked. The fix to this typically required administrative intervention for a manual lock release.
This deficiency in client-server lease management is addressed in NFS v4. NFS v4 NAS configurations introduce a new feature that allows a client to check-in with the NAS appliance every 45 seconds or the lock on the file being accessed is released. This new feature - a default time-out setting in NFS v4 - now allows sub-minute failover of TIBCO ESB in NAS environments. However this improvement is prompting NAS providers such as NetApp to claim they can deliver HA configurations for messaging applications such as TIBCO ESB.
So the question becomes: why would organizations select or remain with a clustered file system such as Symantec's Veritas Cluster File System when they can conceivably start to use a NAS appliance that offers NFS v4?
In looking at this matter, I see four reasons why Veritas Cluster File System remains a preferable choice over a NAS appliance with NFS v4 for HA.
First, just because NAS with NFS v4 now supports HA does not necessarily mean it is optimized for HA. TIBCO ESB environments need continuous availability and often cannot tolerate more than a few seconds of downtime during a failover, especially when they are running trading desks or call logging systems.
Veritas CFS, on the other hand, can run in an Active - Hot Standby mode so the Hot Standby TIBCO server can monitor all activity on the Active TICBO server in real-time. In the event the Active TIBCO ESB server goes down, the Hot Standby TIBCO server takes over processing within a few seconds whereas using a NAS appliance with NFS v4 configuration the failover can take up to a minute.
Second, organizations cannot automatically assume that, in the event of a failover, the TIBCO ESB application server will be able to resume processing in under a minute on the NAS appliance with NFS v4 as the NAS appliance automatically holds the lock for 45 seconds.
So when a TIBCO ESB server is accessing its files on a NAS appliance with NFS v4, the standby server first has to detect that the primary server has failed before it can take over processing. If and when the production TIBCO ESB server fails, the standby server will have to wait at least 45 seconds for the NAS appliance to release the lock before it can restart the TIBCO application in order to build its own view of the TIBCO EBS application.
Third, Symantec CFS still has a significant edge over NAS with NFS v4 in the area of performance. Because Symantec CFS sits atop Veritas Storage Foundation, it can communicate with the underlying storage using block-based protocols instead of file-based protocols. This enables Veritas CFS to perform faster than NAS with NFS v4 as there is more server and network overhead associated with processing NFS.
Fourth, Veritas Cluster File System combined with Veritas Cluster Server (VCS) HA provides the ability to manage and monitor both the file system as well as the TIBCO ESB. A typical HA configuration requires two steps; bring up the file system; then bring up the application. With Veritas Cluster File System HA, you get application-level HA as well as a highly available file system.
The recent introduction of HA capabilities into NFS v4 means that more organizations can examine supporting HA applications on NAS appliances. However, moving to NAS with NFS v4 is not without its downsides as a failover occurs more slowly, application re-builds may first need to occur and performance will likely degrade. It is for reasons like these that many organizations will continue to give preference to Veritas Cluster File System HA as a proven and preferred HA solution.
To summarize a few of the points that I made in that previous blog entry: One of the biggest improvements in NFS v4 relates to file management, a notable enhancement from prior versions as it applies to clustered application servers configured to access the same file via NFS.
Previously using pre-NFS v4, if the application failed and was not able to release the lock, the application on the standby server could not start processing as the file would still be locked. The fix to this typically required administrative intervention for a manual lock release.
This deficiency in client-server lease management is addressed in NFS v4. NFS v4 NAS configurations introduce a new feature that allows a client to check-in with the NAS appliance every 45 seconds or the lock on the file being accessed is released. This new feature - a default time-out setting in NFS v4 - now allows sub-minute failover of TIBCO ESB in NAS environments. However this improvement is prompting NAS providers such as NetApp to claim they can deliver HA configurations for messaging applications such as TIBCO ESB.
So the question becomes: why would organizations select or remain with a clustered file system such as Symantec's Veritas Cluster File System when they can conceivably start to use a NAS appliance that offers NFS v4?
In looking at this matter, I see four reasons why Veritas Cluster File System remains a preferable choice over a NAS appliance with NFS v4 for HA.
First, just because NAS with NFS v4 now supports HA does not necessarily mean it is optimized for HA. TIBCO ESB environments need continuous availability and often cannot tolerate more than a few seconds of downtime during a failover, especially when they are running trading desks or call logging systems.
Veritas CFS, on the other hand, can run in an Active - Hot Standby mode so the Hot Standby TIBCO server can monitor all activity on the Active TICBO server in real-time. In the event the Active TIBCO ESB server goes down, the Hot Standby TIBCO server takes over processing within a few seconds whereas using a NAS appliance with NFS v4 configuration the failover can take up to a minute.
Second, organizations cannot automatically assume that, in the event of a failover, the TIBCO ESB application server will be able to resume processing in under a minute on the NAS appliance with NFS v4 as the NAS appliance automatically holds the lock for 45 seconds.
So when a TIBCO ESB server is accessing its files on a NAS appliance with NFS v4, the standby server first has to detect that the primary server has failed before it can take over processing. If and when the production TIBCO ESB server fails, the standby server will have to wait at least 45 seconds for the NAS appliance to release the lock before it can restart the TIBCO application in order to build its own view of the TIBCO EBS application.
Third, Symantec CFS still has a significant edge over NAS with NFS v4 in the area of performance. Because Symantec CFS sits atop Veritas Storage Foundation, it can communicate with the underlying storage using block-based protocols instead of file-based protocols. This enables Veritas CFS to perform faster than NAS with NFS v4 as there is more server and network overhead associated with processing NFS.
Fourth, Veritas Cluster File System combined with Veritas Cluster Server (VCS) HA provides the ability to manage and monitor both the file system as well as the TIBCO ESB. A typical HA configuration requires two steps; bring up the file system; then bring up the application. With Veritas Cluster File System HA, you get application-level HA as well as a highly available file system.
The recent introduction of HA capabilities into NFS v4 means that more organizations can examine supporting HA applications on NAS appliances. However, moving to NAS with NFS v4 is not without its downsides as a failover occurs more slowly, application re-builds may first need to occur and performance will likely degrade. It is for reasons like these that many organizations will continue to give preference to Veritas Cluster File System HA as a proven and preferred HA solution.
Newsletter Signup
Nice deployment in the Huawei Symantec N8000 Clustered NAS solutions. Yet to hit the U.S. but has arrived on European shores.