Burst versus Swap Memory

OpenVZ vs XEN

Burst versus swap discussions are normally – comparisons of OpenVZ versus Xen memory management. There are advantages and disadvantages to each, with burst associated with VZ and swap to Xen. Assuming everything else is equal, which would you consider is the better option – 1024MB guaranteed (SLM) or 768MB guaranteed (burstable to 2048MB)?

I recently read of an experiment – pairing the two technologies under load conditions. The test involved an OpenVZ container (512MB dedicated RAM with 512MB burst) versus a XEN based VPS (768MB with swap). A basic Debian 5 64 bit OS was installed on each with LAMP, email and Jabber/XMPP. The total OpenVZ memory usage equaled 445MB versus XEN’s 296MB.

So what can we take away from this test? If you’re running Java or other applications that tend to allocate large amounts of virtual memory, OpenVZ containers (without burst) will have less guaranteed memory available. Enter burst .. OpenVZ needs an appropriate amount of burst (usually 1.5 to 2 times the guaranteed).

Beyond guaranteed to burst and swap – access latency on RAM (burst) is measured in nanoseconds whereas swap (HDD) is measured in milliseconds (much, much slower). Swap is slow, so you want ample guaranteed (SLM) to handle your applications. Plus, your system cannot execute stuff out of swap. It has to be in active RAM to execute. With both burst and swap, you want just enough allocation to handle inactive data, keeping as much active data in guaranteed RAM as possible.

Overselling can be – an issue with OpenVZ, e.g., killing your application if it bursts beyond the guaranteed resource allocation, instead of simply slowing down as with a true dedicated server. The kernel of Virtuozzo cannot be modified as it’s based on an OS template. By contrast (using default settings), Xen doesn’t allow overselling, so a physical server with 16GB of RAM would be limited to 64 x 256MB virtual machines. Plus, XEN allows you to fully modify the kernel, so advanced users can compile and load their own modules, resulting in properly virtualized memory, IO and scheduler that’s stable and predictable.

ProlimeHost

Brought to you by ProlimeHost

We’ve been in the web hosting industry for over a decade, helping hundreds of clients succeed in what they do best and that’s running their business. We specialize in Virtual Private Servers (VPS) and dedicated servers, with data centers in Los Angeles, Denver & Singapore.

VPS Services: Lightning Fast SSD Virtual Servers

Our Virtual Private Servers all feature high performance Xeon processors and SSD storage in a RAID10 configuration to optimize your server’s performance, which dramatically enhances visitor experiences on your site.

That speed is backed by unparalleled 24/7 support, featuring both outstanding response AND resolution times to maximize your uptime.

Now is the time to join the ProlimeHost virtual private server revolution.

Dedicated Servers: Backed by a 99.9% SLA network uptime guarantee

We only use enterprise-class hardware in our dedicated servers and offer a four (4) hour hardware replacement. Throw in IPMI for remote management, support for public and private networks, free operating system (OS) re-installs, and SATA, SAS & SSD (including NVMe) storage. Call +1 877 477 9454 or email us at Sales@ProlimeHost.com. We’re here to help.

Steve
Latest posts by Steve (see all)

Leave a Reply

two + 8 =