Better performance at 1/8th the price-per-query. If you’re facing another expensive RISC/UNIX upgrade, there’s a better alternative. Break out of the RISC lock-in! You can now get the performance and reliability your critical workloads demand while taking advantage of the economics only x86 can deliver. Shift to industry standard HP ProLiant servers running Red Hat Enterprise Linux.

These days, when wearing my “Linux planner” hat, and with Virtualization being the “phrase that pays”, I’m often asked to help provide guidance on how to best take advantage of the technology included in our 8-socket HP ProLiant server offerings for Linux based virtualization solutions like Red Hat Enterprise Virtualization or Suse Linux Enterprise Server Xen (there’s a plethora of information out there about VMware ESX/ESXi 3.5.x and vSphere 4.0, so I’m not going to talk about that, this time around.)

The problem I’ve had, until recently, was providing actual – objective - data as a means to help illustrate my points.  For instance, I could not clearly illustrate how a snoop filter on the CPU interconnect can improve the linearity of the workload scalability in a virtualized environment (see Fig. 1).

 Fig. 1: Average response time with pinned vs. un-pinned processors

I was unable to demonstrate benefits of the NUMA aware scheduler that the Linux kernel uses and how it does improve performance. (In figures 2 and 3, it’s represented by the improvement in average response times from the web-servers included in the workload) when your workloads run with memory interleaving disabled in the system BIOS – see Fig. 2. Unless, for support reasons, your application vendor explicitly tells you otherwise.

Fig. 2: Average Response Times - Non-interleaved Memory Config

Fig. 3: Average Response Times - Interleaved memory

I also used to have a hard time explaining how and why to tune the Linux kernel for these systems. For instance, I only suspected how little (none) tuning of the host platform is required in order to drive pretty significant numbers of guests  (98) in these environments - see Fig. 4. But, if you engage in some very minor tuning activities of the network stack, how those very same workload performance results can be extended even further (to 256 guests) – see Fig. 5:

Fig. 4: The system has not been tuned beyond it's "out of the box" state.

Fig. 5: System is tuned and exhibiting linear scalability to 256 KVM guests

As part of a joint documentation effort with Red Hat, all of the data collected has been brought together in a Reference Architecture document  - “Scaling RHEL 5.4 + KVM up to 256 Guests" available for free from Red Hat’s website.

We obviously picked the guest density to prove a point about the platform, however it’s worth mentioning that 256 guests does not represent the upper bound for the platform. It only represents where we thought the density went (far) beyond what is reasonable to expect in a production environment this day in age.

Contributed by Thomas Sjolshagen (Strategic Planner for Linux and Virtualization on scale-up x86 servers)

As I mentioned in one my previous blogs, as x86
processors are getting more powerful, 64 bit architecture is becoming
more mature, memory DIMMs are getting bigger & cost effective, more
and more scalable x86 software applications are becoming available,
scale-up x86 servers are becoming ideal choice for large business
intelligence and decision support system deployment at a cost that no
one imagined a few years ago.

Normal
0




false
false
false

EN-US
X-NONE
X-NONE














MicrosoftInternetExplorer4





























































































































































The latest world record performance
result of HP ProLiant DL785 G6, 102,375.3 QphH at $3.63 USD/QphH, on TPC-H @ 1000 GB benchmark running Sybase IQ 15.1 database and Red Hat Enterprise Linux 5.3
operating system is an excellent proof point. This #1 non-clustered performance benchmark result demonstrates
that customers can deploy large business intelligence solutions at an
aggressive TCO on high-performance 8‑socket x86 servers running. In addition to holding
the #1 non-clustered x86 performance result, the DL785 G6 offers outstanding
price/performance maintaining #1 8P price /performance record in the TPC-H @
1000 GB benchmark category.

The
DL785 G6 with the six-core AMD Opteron™ processors has been designed as
an excellent database server. Its balanced architecture with ample I/O
and memory make it an ideal platform for decision support and business
intelligence  processes.

Hundreds of customers run their database applications on the DL785
server.

The TPC Benchmark™H (TPC-H)  is a decision support benchmark, with components that are intended to be relevant to customers who deploy decision support systems as part of their business intelligence solution. The benchmark is comprised of a suite of business oriented ad-hoc queries and concurrent data modifications that examine large volumes of data and execute highly complex queries. Many
businesses find this type of benchmark useful in determining what
servers to utilize because the TPC-H benchmark illustrates decision
support systems that examine large volumes of data, execute queries with
a high degree of complexity, and give answers to critical business
questions.

Tweet this!