IT Operations
A place for everything, --especially xkoto’s GRIDSCALE database virtualization
Wednesday, December 17, 200802:20 pm EDT
Vertica’s founder, Mike Stonebraker, had some interesting observations about the database world, and more specifically Oracle, in this interview. As one of the industry’s pioneers, he takes some jabs at Oracle RAC and the company’s monolithic view of data, which he calls their “one size fits all” assertion. In the past, I have weighed in on the downsides of RAC’s shared everything approach, but I’ll let Mike’s observations on the inherent scalability limitations speak for me this time.
Clearly, one size does not fit all. For example, innovative Column databases such as Vertica have significant advantages in many of the analytics-heavy use cases that make them shine. Mainstream row-oriented databases are particularly useful for those relational join gymnastics required by reporting and decision support. Oracle’s Exadata, Microsoft’s DATAllegro, and IBM’s Balanced Warehouse advance the cause of data warehouse appliances. Each technology has its place—one size does not fit all.
There is also an important place for our GRIDSCALE database virtualization product. Today I was speaking to one of the world’s biggest ISVs in the telco OSS space, which is weighing how to address the need to improve database access for its data-intensive applications. This is that important place for xkoto’s GRIDSCALE database virtualization software. We see GRIDSCALE as the database “shim” that allows an application provider to turn its single instance database into an active/active database pool without a major coding exercise for the ISV. This way, the ISV can focus on its strengths and leverage xkoto’s strengths as a software appliance in its stack. A place for everything and everything in its place.
full story comments (0)
xkoto GRIDSCALE - A New and Better Way to Provide Database Disaster Recovery
Friday, December 05, 200802:12 pm EDT
Day 3 at the 2008 Gartner Data Center Conference in Las Vegas - crowds are becoming predictably thinner as the week progresses and even as each hour progresses, although some sessions focused on critical issues such as disaster recovery are still packing them in. I heard one attendee complaining about how many miles he had to walk to get from his hotel to the conference - one of the hazards of being in the land of the mega-hotels.
I was on my way to the final session of the day, 4:45-5:45pm, looking for the location for “The Impact of Virtual Technology on Vendor Licensing” (something that strikes fear in the hearts of legacy software vendors); as I turned the corner I came upon a line about 40 people deep. Fearing that this was the lineup for my session I went to the front of the line and saw the display with the name of the session: “Data Replication Architectures for Disaster Recovery.” Suddenly curious, I ditched my designated session and joined the long queue.
The room filled with more than 200 people, standing room only - must be giving away a car or something big is what I reasoned - wrong, no giveaway. Instead, I discovered that these attendees all had a lot of pain around replication. Gartner analyst Stanley Zaffos laid out the strengths/weaknesses of the basic types of replication - server, application, database, storage, and network-based. In a poll of the audience, 35% rely on storage based replication, likely because of they want a mature solution to support disaster recovery. On the downside, Zaffos pointed out that storage based replication (like EMC SRDF) is expensive, consumes a lot of bandwidth, and suffers from vendor lock-in.
After the session, it was crystal clear to me how xkoto GRIDSCALE offers a new, practical, effective solution to the DR problem. As a database virtualization solution, it does not fit nor require traditional database log shipping replication or application replication. As a software appliance, it often does not require much change to the software stack (i.e. the application or the database), unlike server-based replication. It is storage-agnostic like network-based replication. GRIDSCALE provides a new and better way to solve an old and vexing problem.
full story comments (0)
Dealing with Disaster Recovery
Wednesday, December 03, 200811:18 pm EDT
We have just passed one of the biggest shopping days in the US (Black Friday), which to our retail customers also represents a time when most IT organizations have locked down their production environments until after the holiday season. Given that research analysts Donna Scott and Bill Malik at this week’s 2008 Gartner Data Center Conference have cited that 40% of unplanned downtime is caused by operator error, it makes a lot of sense to have these IT blackout periods for disaster prevention.
Regarding the related topic of disaster recovery, there was a focus on the RTO (recovery time objectives - how long you can wait until service is resumed after a disruption) & RPO (recovery point objectives - how much data loss can be tolerated) in determining service level and hence what technologies need to be in place to mitigate disaster. The inability of storage replication technologies to satisfy an RTO of zero was singled out because after storage bits have been replicated to a DR site the remote database would still need to apply those changes to be ready to serve data.
xkoto’s GRIDSCALE technology will be helpful in addressing stringent RTO and RPO requirements due to its active/active database management architecture, which does not rely on third party storage replication. Further, as Gartner analyst Tom Morency observed in his research, data recovery procedures are a key component to successful DR maturity and readiness. With GRIDSCALE, data recovery becomes dramatically simplified since its active/active approach ensures that data is always up-to-date, consistent, and usable in every location.
full story comments (0)
Gartner Data Center Conference 2008
Tuesday, December 02, 200808:09 pm EDT
The MGM Grand in Las Vegas is top-heavy this week with many of Gartner’s top research analysts converging for the annual Data Center Conference. The keynotes and breakouts cover the spectrum from green IT to the cloud. By far the most popular topic is virtualization. In the first presentation server virtualization was predicted to grow to touch up to 50% of all workloads by 2012.
According to Gartner analyst Thomas Bittman, in the past most of these workloads were for dev & test, but now up to 70% of virtualized workloads today are in production environments. In Bittman’s response to the question of whether databases could be virtualized, he noted that the concerns over I/O intensive loads can now be addressed with well-defined data paths, careful application/database design, and adherence to best practices.
From an xkoto perspective, we are seeing more customers moving their databases to VMs. Some of the database best practices that make this transition easier are the same rules that have governed good database design for years, like enforcing referential integrity with primary keys on each table, clustering table and index data together, separating logs from data, making full use of the database engine’s optimizer, and carefully tuning buffer pool sizing.
The steady march of server virtualization is moving from delivering consolidation to agililty to data center management across the stack. It is likely that most workloads including databases will by necessity end up being managed by virtualization technologies.
full story comments (0)
London Stock Exchange - “Highly Reliable Times”?
Tuesday, September 09, 200804:27 pm EDT
The outage at the London Stock Exchange Sept. 8, 2008 may have been triggered due to high trading volumes. Whatever the post-mortem analysis ends up uncovering, the result is another painful example of how fragile IT infrastructure can be and the damaging impact that can be inflicted…
full story comments (0)
