Member Avatar for newsguy
newsguy 30 The News Guy

John Tu is the President and Co-founder of Kingston Technology which just happens to be one of the biggest providers of memory products on the planet, so when he talks about trends and developments in the memory market a lot of people start listening. Especially when he does so against a backdrop of a global economic downturn that has hit the tech industry hard. "Too many companies producing memory chips have caused prices to fall over the last months and consolidation should help the industry correct itself" Tu admits.

However, when it comes to those companies that can weather the economic storm, things are looking surprisingly good through Tu's rose tinted crystal ball. "DDR3 memory is set to gain more ground during 2009" he insists, and this will be boosted by the introduction and adoption of triple-channel and possibly quad-channel memory technologies. "We are also seeing an increased demand for higher speed memory products by prosumers, who need higher performances in applications like video/audio content creation to improve application speeds” Tu adds. Then there is Flash which just continues to go from strength to strength. 2009 is likely to be no exception it seems, especially with regards to high capacity and high performance SD cards for HD-based video and photography use.

What about virtualisation, something which has both benefited and suffered from a bad case of being over-hyped? Tu is upbeat about this saying that "virtualisation is now over the Hype Cycle... IT managers are likely to start looking at ways to get increased ROI and improved TCO from their virtualisation projects. This is likely to come from increasing the memory capacity for virtual servers. Higher memory capacity also enables improved virtual machine performance and scalability”.

And finally, what has Tu got to say about new industry standards for SSD as we move into 2009? After all, there were plenty of reliability issues with SSD usage, mainly thanks to early-design growing pains it would appear, but have refinements been made and problems solved? Tu insists that SSD products from tier one makers are now market worthy but adds "as SSDs gain traction in the market it is likely that we will see a level of new industry standards to ensure the high performance and high quality of this technology”.

  • 1 Contributor
  • 0 Replies
  • 232 Views

Dani AI

Generated

This thread (posted Nov 24, 2008) by raises useful points about memory and storage for virtualised environments. Below is practical, vendor-agnostic guidance that remains applicable years later: focus first on measurement, then on conservative configuration and monitoring so upgrades deliver predictable ROI rather than chasing specs.

Start with a proper baseline. Collect representative samples over normal and peak periods (3–7 days if possible). Key signals are guest working set, guest page faults and swap, host free memory and host swapping, balloon-driver activity, and I/O latency. Example quick checks on Linux:

free -m
vmstat 1 5
cat /proc/meminfo
iostat -x 1 5

On Windows, use Performance Monitor counters (Memory\Available MBytes, Memory\Pages/sec) and hypervisor-specific VM counters. Correlate host-level and guest-level metrics before making sizing decisions.

When sizing, add realistic headroom. Sum peak working sets for the co-located VMs, then reserve 10–20% of host RAM for hypervisor/management overhead and transient peaks. Treat host-level swapping as a service breaker: if it happens, performance is unpredictable. Use memory reclamation features (ballooning, large pages, shared-page techniques) only after testing with the same workload; each technique trades consolidation for latency or visibility. On NUMA systems, align vCPU/memory placement to NUMA nodes for consistent latency.

For SSDs, prefer enterprise-class devices for VM datastores. Key specs to check are endurance (TBW or DWPD), power-loss protection, and controller/firmware maturity. For write-heavy or mixed workloads, plan extra overprovisioning and monitor SMART/health metrics. Protect data with appropriate redundancy (RAID or replication) and keep regular backups—snapshots are not a substitute. Finally, quantify ROI: measure latency and throughput before and after RAM or SSD changes so purchases are justified by measurable benefits.

Be a part of the DaniWeb community

We're a friendly, industry-focused community of developers, IT pros, digital marketers, and technology enthusiasts meeting, networking, learning, and sharing knowledge.